Hypercritical

94: Shrink, Shrink, Shrink!

 

  you can't thread this thing through [TS]

  until the needle comes down but you got [TS]

  to make sure the needle is all the way [TS]

  down before you grab it then you what [TS]

  you have to start and stop a piece of [TS]

  machinery from moving it's there is a [TS]

  limit how fast a sewing machine can [TS]

  operate this is hypercritical weekly [TS]

  talk show ruminating on exactly what's [TS]

  wrong we'll have apple related [TS]

  technologies and businesses nothing is [TS]

  so perfect that it can't be complained [TS]

  about by my friend and yours John [TS]

  siracusa I'm Dan Benjamin today is [TS]

  Friday November 16 2012 this is episode [TS]

  number 94 of our beloved hypercritical [TS]

  show we would like to say thank you very [TS]

  much to our sponsors today squarespace [TS]

  com hover com and lynda.com who are [TS]

  collaborating to make this show possible [TS]

  we also want to thank mutual mobile [TS]

  these guys are super passionate about [TS]

  iOS as a platform they have a large [TS]

  dedicated team all these guys do is [TS]

  conceived deliver and support mobile [TS]

  solutions for global brands and you know [TS]

  what they want some help they want some [TS]

  help they trust your engineers freedom [TS]

  to make technical decisions by hiring [TS]

  smart talented people to their team and [TS]

  that's you they want you you need to [TS]

  come in and once you do iOS work with [TS]

  them they're a great company they [TS]

  support the iOS community they do [TS]

  meetups they do hackathons to do all the [TS]

  cool stuff and just by going and looking [TS]

  at this job you can support this show [TS]

  careers dot mutual mobile com such [TS]

  five-by-five go there to learn more and [TS]

  of course we'd be irresponsible and [TS]

  potentially liable if we did not mention [TS]

  the bandwidth for November's brought to [TS]

  you by mind node mind mapping app for [TS]

  mac and iOS and brainstorm your project [TS]

  you can organize your life you plan your [TS]

  vacation don't matter they late collect [TS]

  structure expand your ideas they have [TS]

  icloud sharing keep your mind maps with [TS]

  you very very cool app beloved mind node [TS]

  dot-com how are you doing today John [TS]

  siracusa I'm just fine you still the [TS]

  same end time we got to be out of here [TS]

  by noon well what time is our thing over [TS]

  to Evernote 1145 and we need to prep for [TS]

  it a little bit and I'd say fifteen [TS]

  minutes to get over there yeah I mean if [TS]

  we go if we go 15 minutes past it's fine [TS]

  so 1215 all right well we gotta get [TS]

  started then let's do it boom yeah we [TS]

  got [TS]

  one major topic of follow-up and then [TS]

  one main topic for the show so maybe [TS]

  we'll fit it in we'll see first let's [TS]

  start with some little tiny follow-up [TS]

  first one is from one canepa CA and the [TS]

  EPA okay I did on that while back we [TS]

  talked about the microsoft surface and [TS]

  how much free space was available on it [TS]

  right up because the OS and bail office [TS]

  took up a lot of room yes and they don't [TS]

  have a 16 gig model daily up 32 and the [TS]

  last show I think I mentioned that one [TS]

  of the big outs that you get with [TS]

  Microsoft services that it's got or the [TS]

  surface RT anyways you got a place to [TS]

  plug in an SD card and SD cards are [TS]

  cheap by third-party SD card shoved in [TS]

  there by 64 gig SD card and suddenly [TS]

  your surface has tons of space well one [TS]

  writes in to tell me and the world that [TS]

  that's all well and good but apparently [TS]

  you can't install applications on the SD [TS]

  card so you can put photos and videos [TS]

  and things that are actually big but the [TS]

  applications have to be on the main [TS]

  system I don't think that's a big [TS]

  limiting factor but it does explain how [TS]

  they manage to deal with like what [TS]

  happens if I yank up the SD card do a [TS]

  bunch of my apps disappear well no [TS]

  because you can't put any apps on uh-huh [TS]

  so that's interesting compromise and i [TS]

  believe is only applies to the RT when [TS]

  the surface pro comes that I think you [TS]

  will be able to put apps there but we'll [TS]

  see alright another little tidbit from [TS]

  Ali Hawkins last week he talked about [TS]

  putting music behind ads you actually [TS]

  talked about this on build and analyzed [TS]

  to how you had put the music track [TS]

  behind that ad read you did when Marco [TS]

  had the ad on Howard Stern Show and I [TS]

  asked you what the name of that thing [TS]

  was called you didn't quite know a [TS]

  couple people wrote in including Ali to [TS]

  say that that's called a bed apparently [TS]

  yeah I that's just something that I [TS]

  should have been able to respond to I've [TS]

  heard it referred to as a couple [TS]

  different things bed is the correct term [TS]

  music bed if you want to get specific [TS]

  about it there you go but there you go [TS]

  now the world knows all right now on to [TS]

  the main topic of follow-up which [TS]

  believe it or not actually shouldn't be [TS]

  surprising as one of the small things we [TS]

  talked about last week was a voting [TS]

  technology [TS]

  this is not a political show John that's [TS]

  right I was happy to see that we didn't [TS]

  get any feedback as far as I saw from [TS]

  people who are angry about politics [TS]

  because we didn't talk about politics we [TS]

  just talked about voting technology but [TS]

  there were lots of people who are not [TS]

  angry but engaged on the topic of voting [TS]

  technology I'm going to read a couple of [TS]

  snippets of feedback here and you tell [TS]

  me if you would notice a theme mostly [TS]

  because I cherry picked these and it [TS]

  makes it look like this a theme but now [TS]

  this really was the theme of a lot of [TS]

  feedback to the first is from andreas [TS]

  hartl he says paper-based method without [TS]

  anything more high-tech than a ballpoint [TS]

  pen does work vote counting can be [TS]

  verified by anybody without [TS]

  sophisticated technical or mathematical [TS]

  skills it's only downside is that you [TS]

  have to wait a little bit longer for the [TS]

  result and he gave me a link to the 25th [TS]

  chaos communication congress website [TS]

  where they had a paper and a [TS]

  presentation on a voting technology i [TS]

  put them both in the show notes so you [TS]

  can read the paper and also watch the [TS]

  video which is kind of long and boring [TS]

  but it's all there and the the paper is [TS]

  avoiding after any da p which is a [TS]

  german voting system and digital pin [TS]

  which is another kind of electronic [TS]

  voting and subtitle is why cryptography [TS]

  might not fix the issue of transparent [TS]

  elections and is a little bit from the [TS]

  abstract these methods introduce a level [TS]

  of complexity into elections which [TS]

  prevents most voters from understanding [TS]

  the election process and its [TS]

  verification where elections are [TS]

  currently controlled by the people trust [TS]

  in the ability of experts is required [TS]

  when cryptographic methods are [TS]

  introduced the next bit of feedback is [TS]

  from Leonard if I'm pronouncing isn't [TS]

  incorrectly because it's felt a little [TS]

  bit differently there's one thing that [TS]

  paper voting system has going for that [TS]

  no computer system will ever get it's [TS]

  entirely comprehensible observable and [TS]

  provable by the simplest educated [TS]

  persons this is essential for trust in [TS]

  the voting process what are you saying [TS]

  is that that anybody with just the basic [TS]

  understanding of making a mark on a [TS]

  paper in a certain place can verify this [TS]

  right versus the electronic and [TS]

  cryptographic ones require master's [TS]

  level computer science degree of you'll [TS]

  figure it out all right he says contrast [TS]

  this to making a cross in a circle [TS]

  putting it in a box and they [TS]

  counting the votes and two piles it's so [TS]

  simple that even a six-year-old can get [TS]

  it a six-year-old could observe me doing [TS]

  it and testified cheatham a six-year-old [TS]

  can verify my result he does not need to [TS]

  trust me and he said any computer system [TS]

  has the problem that is basically magic [TS]

  to normal people he also says they have [TS]

  a word in German about this but I'm not [TS]

  going to try to pronounce it I'll come [TS]

  on no come on out but apparently is it [TS]

  means roughly that a ruling group has [TS]

  now it has a knowledge advantage that [TS]

  they could use to control the uneducated [TS]

  let's see next one from john berg mayor [TS]

  any system needs to be capable of paper [TS]

  recounts a lot of the theoretically very [TS]

  secure systems you mentioned doing do [TS]

  you have audible papal trails but the [TS]

  point is this but this point is so [TS]

  amazingly important that many of the [TS]

  tech nerds talk about voting misted it [TS]

  needs to be possible to conduct a [TS]

  complete hand recount without any [TS]

  computers without electricity even that [TS]

  no system ever to the end of time no [TS]

  matter what technology is ever invented [TS]

  that does not have the capability the [TS]

  capability doing a recount that can be [TS]

  conducted by candlelight without [TS]

  computers is a suitable choice next one [TS]

  is from Mike Musel I talks about what [TS]

  they have in Germany is that in Germany [TS]

  in 2009 our highest court ruled that [TS]

  voting computers we had up until then [TS]

  did violate the principles of our [TS]

  constitutions elections have certain [TS]

  requirements that must be fulfilled in [TS]

  order to be called democratic the [TS]

  manganese is the need for the election [TS]

  to be transparent to the voters and [TS]

  verifiable with our good old crosses on [TS]

  paper ballots even a ninety-year-old [TS]

  non-tech lady can stay at the polling [TS]

  station after the election is closed [TS]

  watch the ballots to be correctly [TS]

  counted and note the sum of the votes [TS]

  for each party even if you had a [TS]

  cryptographically totally secure [TS]

  voting's computer setup how could that [TS]

  nine-year-old non-tech lady know that [TS]

  everything is correct she can't she [TS]

  would have to trust the tech people how [TS]

  can she know that those people one [TS]

  hundred percent trustworthy and [TS]

  competent again she can't there's no way [TS]

  let's see if there's anything more on [TS]

  this one I just about up anyone can go [TS]

  and count the votes she has a good point [TS]

  about smartphone voting remotely or [TS]

  whatever it says even for technically [TS]

  feasible it would pose a risk that for [TS]

  example husband trusts the control house [TS]

  wife boat pilot the security print [TS]

  principal Democrat elections [TS]

  if you're not doing it in a controlled [TS]

  pulling place anonymously people who try [TS]

  to control across your boat so we did [TS]

  remote voting Omri our Beave rights in [TS]

  with a link to NPR on the media podcast [TS]

  segment about online voting and it's a [TS]

  short segment like five minutes long and [TS]

  it's about how is the country of Estonia [TS]

  has been has been doing online voting [TS]

  since 2005 they have government issued [TS]

  identification cards with the password [TS]

  and that's the only real security [TS]

  measure taken when they just basically [TS]

  you know over the internet do their [TS]

  votes so you want to hear how that's [TS]

  working out in a country that's actually [TS]

  doing it and the pros and cons of that [TS]

  approach take a look at that also in the [TS]

  show notes so show notes five by five [TS]

  dot TV such hypercritical / 94 yes many [TS]

  many links we've already just burned [TS]

  through in this quick follow-up so the [TS]

  theme here as you can see is the idea [TS]

  that you need to have a you need to have [TS]

  a way to both understand like you know [TS]

  non-technical birth needs to be able to [TS]

  understand what's going on the voting [TS]

  system and also the boring system still [TS]

  needs to work ah by candlelight with no [TS]

  electricity hmm now the first thing imma [TS]

  say on this topic is that a secure you [TS]

  know and inaudible electronic voting [TS]

  system doesn't preclude any of these [TS]

  things in fact some of these systems [TS]

  that were linked in that in the past [TS]

  shows show notes showed that yes the you [TS]

  know just because you use cryptography [TS]

  and electronic things to make the thing [TS]

  auditable and secure and all that stuff [TS]

  doesn't mean that you don't also have [TS]

  okay so say the power was out no one has [TS]

  any computers because it's like the [TS]

  terrible premise of the show revolution [TS]

  you still end up with a bunch of paper [TS]

  artifacts that you could then hand count [TS]

  manually right so it's not you know [TS]

  various voting systems passed or failed [TS]

  this thing but there's nothing inherent [TS]

  about electronic voting or [TS]

  cryptographically secure voting that [TS]

  doesn't say also you have you know it [TS]

  doesn't preclude you having this [TS]

  auditable paper trail right ah so most [TS]

  of these feedback you know acknowledge [TS]

  that but the other thing is to said you [TS]

  know all right so you can make it secure [TS]

  an electronic or whatever better you [TS]

  also have to sort of degrade gracefully [TS]

  there has to be a paper trail an [TS]

  artifact in the ability to do a Henry [TS]

  count but that aside I don't think any [TS]

  of these requirements you know eliminate [TS]

  electronic voting another thing in [TS]

  Germany Bailey had to eliminate the [TS]

  cruddy electronic voting machines they [TS]

  had and by the way several people from [TS]

  different countries wrote in to say that [TS]

  they have similar problems the US where [TS]

  when electronic voting machines were [TS]

  rolled out they were worse in all [TS]

  possible ways than the things that they [TS]

  were replacing we have the same problem [TS]

  the United States pointed last showbiz [TS]

  that that doesn't need to be the case [TS]

  that we have the smarts to make this [TS]

  better but it is important to recognize [TS]

  it just making electronic usually seems [TS]

  to go terribly wrong because of just you [TS]

  know corruption and competence and all [TS]

  the usual things that happen in [TS]

  government but I want to think about [TS]

  this idea that the the election both [TS]

  these idea is that that an election [TS]

  system needs to work by candlelight and [TS]

  the people need to understand so first [TS]

  the candlelight one people I understand [TS]

  people fearing comforted by the idea [TS]

  that if a giant EMP goes often destroys [TS]

  all electricity producing things or [TS]

  whatever crazy scenario you want to come [TS]

  out with that we can you know you a [TS]

  bunch of old ladies could do the recount [TS]

  like you don't need to understand [TS]

  anything you just need to look at pieces [TS]

  of paper and put them into piles and [TS]

  then count the things in the pile and [TS]

  you have 20 different people do this [TS]

  it's completely comprehensible the [TS]

  comfort of that is obvious like you [TS]

  don't like something that you don't [TS]

  understand but the comfort of that I [TS]

  think is mostly theoretical because [TS]

  there's no way that any single person is [TS]

  going to be able to hand count all the [TS]

  votes for any country or a significant [TS]

  size all right a group of people can [TS]

  hand count them in a distributed manner [TS]

  and then they add up their sums and you [TS]

  have people repeated or whatever but in [TS]

  general it you know you can never have [TS]

  every single person in the country get [TS]

  their hands on all the paper ballots [TS]

  like say you took all the paper ballots [TS]

  and the country the size of the US and [TS]

  each person the entire country got a [TS]

  chance to hand count them to verify for [TS]

  themselves well that's just obviously [TS]

  not going to work right so in all cases [TS]

  you're trusting that some group of [TS]

  people that is not you is trustworthy in [TS]

  some way or at least it there that their [TS]

  biases are counter to each other and [TS]

  balance each other out or whatever [TS]

  but people like to just think that you [TS]

  know okay well I'm not going to hang out [TS]

  the ballots and no 10 is going to Hank [TS]

  out the balance but I do though that 17 [TS]

  people hand counted the ballots and half [TS]

  of them were for party a and half of [TS]

  them from a party B and if they keep [TS]

  coming up with the same numbers i trust [TS]

  those number right I don't quite see the [TS]

  connection between that distributed [TS]

  trust where you didn't do it your friend [TS]

  didn't do it but eventually there's some [TS]

  group of people who actually had [TS]

  encountered the things and anyone can [TS]

  volunteer to count them liking all [TS]

  countries like Jeremy and stuff anyone [TS]

  can go down there and count them and [TS]

  satisfy themselves but everybody can't [TS]

  go down account so there is some remote [TS]

  trust and the idea that okay but I [TS]

  understand what they're doing they're [TS]

  just looking at pieces of paper and [TS]

  counting them sorting them into piles I [TS]

  understand that process I don't think [TS]

  those things are as connected as people [TS]

  seem to think they are ah that because [TS]

  you understand the system therefore this [TS]

  indirect trust is more trustworthy right [TS]

  because indirect trust for set work for [TS]

  when you're saying I don't you know [TS]

  these mathematicians assure me that it's [TS]

  correct but I don't know these [TS]

  mathematicians right that indirect dress [TS]

  is no good but the indirect trust of [TS]

  seven people away from you a bunch of [TS]

  people counted a piece of paper that's [TS]

  better because you're like okay not only [TS]

  am I trusting these people who are not [TS]

  me but I understand what they're doing [TS]

  therefore I feel like I'm less likely to [TS]

  be cheated my argument would be that [TS]

  it's much easier to like if those people [TS]

  were cheating or there was some sort of [TS]

  organized cheating conspiracy there [TS]

  there is nothing that anyone else can do [TS]

  to verify that they're not cheating [TS]

  except for count them themselves whereas [TS]

  and this is I think you know this is may [TS]

  come back to people not understanding [TS]

  the the mathematical details of [TS]

  cryptography and stuff like that or I [TS]

  mean I don't understand the mathematical [TS]

  details but just like the consequences [TS]

  of it if you can imagine a system where [TS]

  uh you add up a bunch of numbers and you [TS]

  come up with a total like 12 or 13 right [TS]

  and what you want to know is did my vote [TS]

  contribute to that number 12 it seems [TS]

  impossible because you just thought up [TS]

  with the number 12 you have no idea if [TS]

  your vote was part of that 12 right how [TS]

  can how can you ever know that you know [TS]

  it was just 12 was made up of a bunch of [TS]

  votes maybe one of those was yours maybe [TS]

  it wasn't maybe your [TS]

  was counted maybe it wasn't you just [TS]

  know number 12 will imagine the system [TS]

  where you could come up with a total [TS]

  number of votes and you could be [TS]

  mathematically assured that your vote [TS]

  contributed to that total this is the [TS]

  getting into the magic thing it's like [TS]

  what do you mean that makes no sense you [TS]

  can't tell if your number contribute at [TS]

  all that's impossible well there are [TS]

  things that you can do mathematically to [TS]

  make that so that you can confirm that [TS]

  your vote really did contribute to the [TS]

  total and that if your vote wasn't [TS]

  counted towards a total in the correct [TS]

  manner the total would not validate in a [TS]

  particular way uh this is going to help [TS]

  anybody because this is even more [TS]

  complicated but if you ever go read up [TS]

  on Bitcoin which is this distributed [TS]

  fiat currency / scam / whatever you know [TS]

  you know I don't know how you want to [TS]

  talk about wait why are you calling a [TS]

  scam that's not a scam but like it's [TS]

  it's an economic thing where it's kind [TS]

  of like gambling but not really but the [TS]

  currency and the imaginary nature of [TS]

  money is a topic that's beyond the scope [TS]

  of this podcast we're talking right now [TS]

  but the one thing Bitcoin does have [TS]

  going for it is that it has the system [TS]

  whereby everybody who's part of the [TS]

  Bitcoin system can tell whether a [TS]

  transaction is valid or not like you [TS]

  can't sneak in an invalid transaction [TS]

  that you know puts more money into one [TS]

  account that it took then it took out of [TS]

  the other or whatever because this is [TS]

  consensus type of algorithm and those [TS]

  things will be invalidated and it's [TS]

  basically possible to tell whether the [TS]

  money that you put it in correctly [TS]

  transferred to over there ah [TS]

  mathematically and so that's that you [TS]

  know that ignores all the whether it's a [TS]

  good idea to get into Bitcoin and then [TS]

  how much it's valued and stuff like that [TS]

  just simply the foundation of being able [TS]

  to tell that these inputs produce this [TS]

  output in this particular way and that [TS]

  everybody in the system has to agree on [TS]

  it for the system to move forward agree [TS]

  sounds like it's something anyway it's [TS]

  it's basically a much more complicated [TS]

  version of the thing that I was saying [TS]

  where if you have a total number you [TS]

  want to know that your vote counted [TS]

  towards it and that's something you can [TS]

  do with math there's always going to be [TS]

  a secret somewhere of some secret key or [TS]

  some other things being corrupted or [TS]

  stolen but trusting that there's a [TS]

  mathematical foundation to proving [TS]

  whether something happened or not [TS]

  I i think is better than trusting that [TS]

  people you don't know and never going to [TS]

  meet have hand counted these things [TS]

  simply because you think they're you [TS]

  know balancing each other off and if [TS]

  they were cheating they would catch each [TS]

  other whatever like there's nothing [TS]

  there's nothing to hang your hat on [TS]

  mathematically that you can use to to [TS]

  prove that those guys did the right [TS]

  thing you basically just have to trust [TS]

  them whereas if you have a [TS]

  cryptographically secure assistant [TS]

  whether you understand it or not uh [TS]

  there should be enough smart people in [TS]

  the world assuring you on both sides of [TS]

  the coin that like look if you do this [TS]

  you'll be able to tell whether your vote [TS]

  was counted and I do get the wrong ansel [TS]

  your note you know your vote wasn't [TS]

  counted and the things required cheat on [TS]

  this you know would be you know that [TS]

  they can lay out the exact parameters [TS]

  there's no there's no spy craft you [TS]

  could do to corrupt the system in a way [TS]

  that wouldn't be detectable except for [TS]

  you know stealing secret keys or [TS]

  something uh now just because I I I have [TS]

  more trust in mathematics combined with [TS]

  this remote trucks than people doing the [TS]

  other things like they'd rather have the [TS]

  comfort doesn't mean that this is that [TS]

  these people all don't have points that [TS]

  this is all well and good and you can [TS]

  try to like and put it on a blackboard [TS]

  and explain to people how it works and [TS]

  stuff but the bottom line is if they [TS]

  don't trust it it's not good because you [TS]

  know you want a system that everybody [TS]

  trusts whether people trust it is not [TS]

  necessarily connected to whether it is [TS]

  more or less trustworthy so I think [TS]

  their their point on it being able to be [TS]

  counted by candle light stands not so [TS]

  much on the merits of the technology but [TS]

  on the fact that most of the world [TS]

  probably agrees with them and in the end [TS]

  this is a system that everyone has to [TS]

  agree is fair for things to work well so [TS]

  if that's if that is unchangeable and [TS]

  you cannot convince the world that [TS]

  you're secure voting system actually is [TS]

  secure and you cannot convince them that [TS]

  you can show them how to secure it and [TS]

  that it's more secure than other [TS]

  alternatives it doesn't matter if it's [TS]

  really more secure not all it matters is [TS]

  what they think of course that's the [TS]

  ultimate election hack the fact that it [TS]

  doesn't really matter what actually [TS]

  happens in the election all that matters [TS]

  is that everyone involved or almost [TS]

  everyone involved agrees that it was [TS]

  fair whether it was really fair not [TS]

  almost immaterial if you were the best [TS]

  way to cheat an election is to do it in [TS]

  such a way that everyone agrees that the [TS]

  result is fair [TS]

  so that's kind of depressing but but [TS]

  anyway I think their point stands like I [TS]

  said I think any decent [TS]

  cryptographically secure voting system [TS]

  will also and must also include a paper [TS]

  trail and all the things that to be able [TS]

  to be hand countered and if the security [TS]

  is just extra on top of it oh and I also [TS]

  want to get back to one of the piece in [TS]

  the feedback of like the only thing you [TS]

  get from an electronic auditable roading [TS]

  record is that it's uh it's faster [TS]

  depart to tabulate the results that is [TS]

  not one of the goals of these systems or [TS]

  shouldn't be ready shortly speed it [TS]

  should be accuracy or not just accuracy [TS]

  but like it's an extra level [TS]

  verification like you've got all the [TS]

  same old verification with pieces of [TS]

  paper and people can hand recount them [TS]

  do a lot stuff but also this is extra [TS]

  layer of security on top of that that [TS]

  gives you that you didn't have before [TS]

  like for example before you didn't have [TS]

  the ability to like go to a website and [TS]

  enter in some secret code that you got [TS]

  in your little paper receipt and find [TS]

  out whether your vote was counted or not [TS]

  you may not understand that you may [TS]

  think it's all a lie and it's totally [TS]

  fabricated and not given you know any [TS]

  credence whatsoever that's fine because [TS]

  it's just on top of the existing thing [TS]

  that you do trust but in general I think [TS]

  you know that that's the case where your [TS]

  your disbelief the sort of Luddite [TS]

  disbelief that electronic scan cannot [TS]

  possibly add in the extra security [TS]

  doesn't mean that it doesn't actually [TS]

  add extra security and make it more [TS]

  difficult to to corrupt and stuff like [TS]

  that so I think that's like a bonus type [TS]

  system yeah I think I basically combine [TS]

  the two whether whether you know people [TS]

  understanding it and having the paper [TS]

  trail or two separate things for one [TS]

  they kind of meld into each other uh [TS]

  yeah III don't want to call people [TS]

  luddites because they don't trust [TS]

  electronic voting because if you look at [TS]

  the history of electronic voting you [TS]

  know like everyone who wrote in [TS]

  including you know everything's happened [TS]

  in the US we've just done it so terribly [TS]

  that how could it not have a bad [TS]

  reputation I mean it's got a deserve a [TS]

  terrible reputation and especially [TS]

  people who don't understand it all they [TS]

  know is every time someone tries to [TS]

  introduce electronic voting it ends up [TS]

  being a giant disaster therefore we [TS]

  should never do anything electronically [TS]

  and that's it I was also trying to think [TS]

  of other things in the world that [TS]

  average person doesn't understand but [TS]

  the net [TS]

  less people trust it's difficult to come [TS]

  with straight through analogies but [TS]

  there are plenty of them I the problem [TS]

  is they're not exactly the same thing [TS]

  like people understand how airplanes fly [TS]

  but they have faith that they'll be in [TS]

  the air and that you know something like [TS]

  they have really have no idea like her [TS]

  you know and people don't understand how [TS]

  government works in general and how you [TS]

  know just how the machinery of [TS]

  government works in actuality but they [TS]

  trust that all this law was passed and I [TS]

  trust it was passed in a way that didn't [TS]

  involve just a bunch of guys in the back [TS]

  room chomping cigars and saying let's do [TS]

  this you know there are many things in [TS]

  the world that people don't understand [TS]

  that nevertheless put their trust in so [TS]

  I'm not of the opinion that it would be [TS]

  impossible to get human beings to trust [TS]

  an end an auditable cryptographically [TS]

  secure electronic voting system I'm of [TS]

  the opinion that it would be possible to [TS]

  it maybe you have to wait for a couple [TS]

  generations people die off and gradually [TS]

  introduce it and actually do it right [TS]

  for once and stuff like that but it [TS]

  could happen and I think the people who [TS]

  are against it would eventually die off [TS]

  and everyone else would think it's crazy [TS]

  to do it in any other way I don't think [TS]

  voting is unlike every other part of the [TS]

  human condition where it's like well [TS]

  everything else we can advance [TS]

  technology and but voting we have to do [TS]

  with rocks until the end of my voting is [TS]

  different in some fundamental way than [TS]

  anything else human beings do right [TS]

  again out it cannot be improved upon [TS]

  with modern technology it must be the [TS]

  same that it was you know right hundreds [TS]

  and hundreds and hundreds of years ago [TS]

  and there are many ways you can shoot [TS]

  down any analogy flying a plane is not [TS]

  like an election don't you understand [TS]

  electron is how we govern ourselves it's [TS]

  like meta system it's so important ball [TS]

  blah but there is you know there's [TS]

  nothing in in life that is so special [TS]

  that technology cannot touch it ever and [TS]

  it almost sounds like the intro TV show [TS]

  no nothing is so perfect that it yeah I [TS]

  don't know it's it's kind of similar [TS]

  thing but anyway uh like I said I'm not [TS]

  I'm not arguing against paper trails I [TS]

  think they have to be there and I think [TS]

  there's no reason they shouldn't be but [TS]

  I am arguing for electronic voting [TS]

  systems done right and done right means [TS]

  you have all of the benefits of the [TS]

  existing system plus many extra benefits [TS]

  on top of that that people may not care [TS]

  about or anything but as long as the [TS]

  superset of the existing security [TS]

  mechanisms then so what people can [TS]

  pretend that's all magic and BS and it [TS]

  is equally secure to the best you know [TS]

  paper system with [TS]

  the hand recounts in all its stuff like [TS]

  all right that went a little long I [TS]

  think I was rambling there but I think [TS]

  it's alright sponsor sure hover com [TS]

  simplified domain management you've [TS]

  probably registered a domain from a [TS]

  company that they tried upsell you on [TS]

  stuff you go there you register domain [TS]

  name and the next thing you know you've [TS]

  signed up for it for five years and [TS]

  you've bought all kinds of additional [TS]

  you know back ordering things and stuff [TS]

  you didn't want or even if you are smart [TS]

  and savvy enough to disable all of those [TS]

  things as you're going through the [TS]

  checkout process they're still there and [TS]

  there's still kind of a nuisance well [TS]

  hover com does not do that these guys [TS]

  their goal is to provide you with a [TS]

  straightforward minimal clean honest [TS]

  process that makes it easy to go there [TS]

  register the domain that you want and go [TS]

  away that's what they do there's a [TS]

  search box you type in the domain name [TS]

  you want if it's available you register [TS]

  it if it's not available they'll show [TS]

  you some alternatives and that's it you [TS]

  get out there's a checkbox do you want [TS]

  to have the free who is privacy I [TS]

  believe that checked then you can go [TS]

  into your settings and make it so that [TS]

  there's like a checkbox in there that [TS]

  says that never never contact me you can [TS]

  do that if you want i don't mind them [TS]

  contacting me because they have some [TS]

  pretty cool stuff like their domain [TS]

  transfer valet service where if your [TS]

  transfer your domain from another [TS]

  registrar they'll do the entire thing [TS]

  like a human being that works there for [TS]

  free will do the whole process for you [TS]

  they have email hosting if you wanted [TS]

  they have built-in free dns which is [TS]

  what i'm pretty much using for almost [TS]

  everything now go check these guys out [TS]

  hover com / dan sent me this will get [TS]

  you ten percent off everything that you [TS]

  do or you can just use the coupon code [TS]

  dan sent me when you're buying something [TS]

  like if you're an existing customer if [TS]

  you're already there and you're [TS]

  transferring more domains there's no [TS]

  limit on how many times you can use that [TS]

  and i actually use that myself for my [TS]

  domain name registration problem my [TS]

  addiction and you can go there hover com [TS]

  such dance let me get ten percent [TS]

  hopefully you liked the experience as [TS]

  much as I do thanks very much to them [TS]

  for making the show possible what you [TS]

  got next John and before we move on one [TS]

  point that a lot of these bits of [TS]

  feedback oh I got a turned off all right [TS]

  a lot of these bits of feedback about [TS]

  voting kind of matter of factly state [TS]

  that the way they're voting system works [TS]

  in their country the voting system do [TS]

  they like usually it that involves [TS]

  government-issued ID right and from [TS]

  coming from a u.s. perspective the rest [TS]

  of the world probably thinks we're crazy [TS]

  but this just goes to show it like you [TS]

  know and in the same way that I don't [TS]

  think that the average person needs to [TS]

  understand the math involved for a [TS]

  cryptographic voting system to [TS]

  eventually be considered secure people [TS]

  and the rest of world don't think twice [TS]

  about having government-issued ID and [TS]

  having to show it at the polls to vote [TS]

  whereas an America that is a very very [TS]

  big issue and some of the same people is [TS]

  that you know don't you understand it [TS]

  has to be a paper ballot has to be hand [TS]

  candid and of course you have your [TS]

  government issue date I'm turn around [TS]

  said don't you understand you can't [TS]

  require government issued ID and say [TS]

  well what how can you vote without a [TS]

  government issue lady how do we even [TS]

  know who's what it again you know [TS]

  everyone has their personal hang-ups [TS]

  culturally personally then not all of [TS]

  them are based on logic so in the same [TS]

  in the same way that I'm saying to these [TS]

  people that owe you know you you know it [TS]

  doesn't need to be done with rocks when [TS]

  that they would say to us you know you [TS]

  have to have a government ID to vote how [TS]

  can you vote without that it's just [TS]

  stupid and I can't talk about this too [TS]

  much without getting into politics or US [TS]

  politics we just go show that the rest [TS]

  of the world is different like and I [TS]

  don't think most of the feedback even [TS]

  acknowledge that they were just like let [TS]

  me explain how it works in the civilized [TS]

  world we use rocks and pieces of paper [TS]

  and a government-issued ID it's like [TS]

  that's not gonna fly in the US that's no [TS]

  ah people are strange when you're a [TS]

  stranger is that it yes that is the [TS]

  correct Jim arson doors lyrics okay all [TS]

  right so let's move on to the main topic [TS]

  today and this could be short how many [TS]

  more sponsors do we have we have two [TS]

  more so we'll space them out alright so [TS]

  don't let me go too long alright gotta [TS]

  finish this topic is as teasing the last [TS]

  show it's about using ARM processors all [TS]

  caps a RM instead of Intel processors on [TS]

  max right and it is spawned by this [TS]

  business we guard [TS]

  from that a week ago titled Apple said [TS]

  to be exploring switch from Intel for [TS]

  Mac that is bad headline because it is [TS]

  awkward right here's a little synopsis [TS]

  thing Apple exploring ways to replace [TS]

  Intel processors in its Mac personal [TS]

  computers with a version of the chip [TS]

  technology uses in the iphone or ipad [TS]

  according to people familiar with the [TS]

  company's research people familiar with [TS]

  Apple engineers have grown confident [TS]

  that the chip designs used for its [TS]

  mobile devices will one day be powerful [TS]

  enough to run its desktops and laptops [TS]

  said three people knowledgeable of with [TS]

  knowledge of the work drank who has to [TS]

  remain anonymous because the plans are [TS]

  confidential alright so typically Fuzzle [TS]

  esource thing and mostly gonna use as [TS]

  articles a jumping-off point for a [TS]

  discussion about this idea and i'm going [TS]

  to give sort of a cliff notes background [TS]

  on what i think are the important issues [TS]

  behind this if you know anything about [TS]

  cpu technology and none of this is going [TS]

  to be new information and i apologize [TS]

  but most of the time i do a show like [TS]

  this i get positive feedback from the [TS]

  people who may not be you know nerds in [TS]

  this particular section of the topic so [TS]

  if you're a tech nerd you can tune out [TS]

  for a while and just wait till the end [TS]

  when i get to some conclusions and but [TS]

  if you're not i hope i will at least [TS]

  teach you something in broad strokes [TS]

  that you may not have known too much [TS]

  about before alright so let's start with [TS]

  the risk versus sisk you know those [TS]

  stand for damn i do not remember i used [TS]

  to know and back in college I there you [TS]

  go you'll be one of those people I was [TS]

  getting value out of the section that's [TS]

  round all right so these are all caps [TS]

  are is C and C is C rse is reduced [TS]

  instruction set computing on complex [TS]

  instruction set computing yeah you go [TS]

  pulling stuff up all right no as you as [TS]

  soon as you said the first and i [TS]

  remembered second yeah this is one of [TS]

  those i think it's pretty much like a [TS]

  backronym where the the acronym Sisk [TS]

  didn't invent didn't exist until the [TS]

  acronym risk was invented someone can [TS]

  send me reckons it out but it's kind of [TS]

  like oh I have an idea and I've given [TS]

  name and I'm also going to give a name [TS]

  to your idea because my idea is better [TS]

  than your idea and I without contrast [TS]

  them and yours as complex as my [TS]

  introduce anyway this was a hot debate I [TS]

  know decade ago longer yeah it's a [TS]

  couple other i would say longer because [TS]

  i remember the well this came this is [TS]

  the whole spark vs intel thing its like [TS]

  the 90's yeah mid late nineties even [TS]

  before that does there's a bunch of [TS]

  articles from ARS technica that are [TS]

  great to look at this prism the show [TS]

  notes one is called risk versus sisk the [TS]

  post risk era by John Stokes from 1999 [TS]

  that's a really old one and there's also [TS]

  risk versus Sisk in the mobile error [TS]

  also by John Stokes in 2008 if you want [TS]

  nerdier backgrounds on this stuff I [TS]

  suggest reading those articles all right [TS]

  but I'm going to give you the cliff [TS]

  notes version so first let's start with [TS]

  you know reduced instruction set [TS]

  computing complex instructions I [TS]

  computing what the heck is an [TS]

  instruction set instructions that is [TS]

  basically the list of things that you [TS]

  can ask a processor to do like you build [TS]

  this little piece of hardware and you [TS]

  have to talk to it and what can you tell [TS]

  it to do it is the API for the processor [TS]

  using the parlance of our time yeah cuz [TS]

  there's only you know what can you tell [TS]

  a processor to do this limited set of [TS]

  things you can tell it to do and you [TS]

  have to define them can you it can you [TS]

  tell the process of dad two numbers [TS]

  together probably right so how do you do [TS]

  that while you tell this instruction and [TS]

  you put this number here in this number [TS]

  that you know there's a list of things [TS]

  that you can tell it to do did you ever [TS]

  get assembly I sure did okay surely did [TS]

  yes too much assembly so how it works is [TS]

  that you know these days is source code [TS]

  is compiled the machine code which is [TS]

  like a list of instructions for the CPU [TS]

  and the CPU reads those instructions [TS]

  from memory and execute them right [TS]

  machine code is a turd term you don't [TS]

  hear thrown around too much today mostly [TS]

  because not that relevant but what it [TS]

  what it comes down to is just a bunch of [TS]

  numbers right it's going to read a bunch [TS]

  of numbers from memory and it's going to [TS]

  interpret them in a certain way that's [TS]

  like oh when I see the number five that [TS]

  means the add instruction and then the [TS]

  next number is the one thing I'm adding [TS]

  in the next number is the other thing [TS]

  I'm [TS]

  it's not really like that but that's [TS]

  that's the idea ah almost nobody writes [TS]

  machine code in the end these days uh [TS]

  maybe people who would like me and [TS]

  undergraduate course had to design their [TS]

  own CPUs and then program with machine [TS]

  code ended up writing machine code where [TS]

  you just literally write a series of [TS]

  numbers it is like the least human [TS]

  friendly way to program I guess toggle [TS]

  switches we worked so if you couldn't [TS]

  even type the numbers in you had to do [TS]

  them their binary representation with [TS]

  toggle switches that would be worse but [TS]

  no one does that right assembly which is [TS]

  a term you may have heard assembly code [TS]

  is kind of like a symbolic version a [TS]

  machine code instead of having my number [TS]

  five you decide to if you type a DD the [TS]

  word ad we're going to tell us the ad [TS]

  instructor so you don't have to remember [TS]

  that the add instruction is number five [TS]

  you just type in a DD and then the next [TS]

  thing you type in a bunch of letters it [TS]

  says where it's getting one number from [TS]

  and you know whatever it's much nicer [TS]

  for human beings to not have to type [TS]

  just a series of numbers they want words [TS]

  or and so they just basically map the [TS]

  words two numbers you take that assembly [TS]

  code you run it through an assembler it [TS]

  reads the little words it converts them [TS]

  into the actual numbers and then the [TS]

  Machine execute those numbers and those [TS]

  little words or those numbers that's [TS]

  your instruction set about you've got an [TS]

  add obstruction or subtract instruction [TS]

  and compare instructional also the stuff [TS]

  they have numbers assigned to them you [TS]

  write an assembly code the CPU is not [TS]

  executing assembly code your assemblers [TS]

  looking at the assembly code converting [TS]

  them pretty much one to one into their [TS]

  equivalent numbers putting them into [TS]

  memory than getting the CPU to slurp [TS]

  those instructions out of memory and [TS]

  then go through each one execute it and [TS]

  so and so forth right now people used to [TS]

  write programs in assembly a lot more [TS]

  than they do today now almost no one was [TS]

  writing a machine code in our lifetimes [TS]

  button assembly you know people used to [TS]

  write programs that huge parts of the [TS]

  original mac operating system were [TS]

  written in assembly not just the drivers [TS]

  or you know anything you would think [TS]

  it's like low-level code but just like [TS]

  huge parts of the actual operating is [TS]

  like parts of the mac toolbox that did [TS]

  the GUI like buttons and dialogues and [TS]

  menus big parts of that were in an [TS]

  assembly a lot of the reason they're [TS]

  written assembly is because assembly is [TS]

  like you know that's you talking almost [TS]

  directly to the cpu granted you're [TS]

  writing in a symbolic language thick is [TS]

  converted into numbers but it's very [TS]

  close to one to one correspondence is [TS]

  what's going on inside the CPU so if you [TS]

  want to wring every last piece of [TS]

  performance out soon as I look CPU I [TS]

  gotta tell you do this then you're going [TS]

  to do this and [TS]

  you're going to put the result there and [TS]

  the next instructor not going to take [TS]

  that result from there and put it over [TS]

  here like very low level direct you know [TS]

  not not a big layer of abstraction [TS]

  between you and telling the CPU what to [TS]

  do and assembly code can be very small [TS]

  and very clever and do interesting [TS]

  things so and memory was very tight back [TS]

  then so large parts of the back-up rings [TS]

  is written in assembly and if you're [TS]

  writing anything in assembly what you [TS]

  want is a a processor that has lots of [TS]

  powerful instructions for assembly right [TS]

  you wanted to be able to write an [TS]

  instruction to this cpu where do [TS]

  something useful for you all right so [TS]

  here's an example of what later became [TS]

  known as complex instruction set [TS]

  computer single command that says take a [TS]

  piece of memory from location a and move [TS]

  that piece of memory into a different [TS]

  place in memory location B that seems [TS]

  like the simplest possible instruction [TS]

  you can do but it's not it's actually [TS]

  pretty complicated because you were [TS]

  telling the CPU go into Ram fine address [TS]

  57 take the stuff that's there take a [TS]

  bite take a word dick a bubble or take [TS]

  some piece of memory and then move it to [TS]

  a different place in memory address 122 [TS]

  right that is actually relatively [TS]

  complicated because there's no way you [TS]

  know it's it has to you have to the CPU [TS]

  have to pull the thing out of memory [TS]

  hold it for a second and then stick it [TS]

  into another place but if you were [TS]

  writing if you were had to write [TS]

  assembly you would want to be able to [TS]

  write that you wouldn't want to have to [TS]

  write take the thing that's in memory [TS]

  location 57 pull it into the CPU take it [TS]

  out of the CPU and stick it into memory [TS]

  location 256 you just want to say just [TS]

  move it from 56 to 27 or whatever you [TS]

  just one instruction right that is an [TS]

  example of a complex instruction set it [TS]

  it's you can imagine the the individual [TS]

  steps that you want to do but you don't [TS]

  want to have to write those individual [TS]

  step you just want to sell cheap cell is [TS]

  super you look let's move this thing [TS]

  from one place to the other because if [TS]

  you didn't you have to write three [TS]

  instructions every time you just want it [TS]

  to move something from one thing it so [TS]

  the race back then was to make CPUs that [TS]

  had really fancy instructions so that if [TS]

  you're writing an assembler or even if [TS]

  you're writing a compiler or whatever [TS]

  you're like I don't need to handhold the [TS]

  CPU I can say [TS]

  here's what I want you to do you figure [TS]

  out how to do it I don't care how you do [TS]

  it but I have a very powerful [TS]

  complicated instruction and you do that [TS]

  the way you want it that's complex [TS]

  instruction set computing because that [TS]

  they all the CPU makers are trying to [TS]

  make you know my CPU has an instruction [TS]

  and lets you you know compared to [TS]

  strings to see whether they're equal [TS]

  just making this up but like like wow I [TS]

  don't even have to go through the [TS]

  individual steps of that the CPU will [TS]

  just do it itself that's a great CPU [TS]

  it's powerful it's awesome right who [TS]

  wouldn't like that John right alright so [TS]

  here's here this kind of thing falls [TS]

  down and I'm going to go into a crazy [TS]

  analogy that's not even remotely close [TS]

  to being correct but hopefully give you [TS]

  just what's going on right so I'm going [TS]

  to say that this the complex instruction [TS]

  set computing having very powerful [TS]

  instructions I can tell the computer to [TS]

  do this complicated thing is kind of [TS]

  like a sewing machine and there's a [TS]

  particular sewing machine animated gift [TS]

  that I tried to find on the web I found [TS]

  like eight thousand versions of it I put [TS]

  one in the chat room here I have you [TS]

  seen this one you're looking at chat [TS]

  room I am looking let me see oh yeah [TS]

  this is very essence old one that's [TS]

  great there's a glier ones too but for [TS]

  people who don't know how sewing [TS]

  machines work or have never used a [TS]

  sewing machine take a look of his [TS]

  animation you see all seen sewing [TS]

  machines it's like a big machine with [TS]

  like a needle and you put fabric down on [TS]

  a table and you slide it into needle [TS]

  needles up and down really fast up and [TS]

  down up and down up and down you feed [TS]

  the material through it and so does [TS]

  things together this animation shows you [TS]

  what's actually going on it also [TS]

  explains the mystery of the bobbin yeah [TS]

  like when the needle goes down in a song [TS]

  machine it's not just just punching a [TS]

  string through and pulling it back up [TS]

  because that wouldn't work we'll just [TS]

  make a series of holes all right there's [TS]

  a thread threaded through the needle and [TS]

  when it goes down through the fabric [TS]

  into the table below it is a little [TS]

  bobbin that basically takes some thread [TS]

  that's on the bobbin and weave it [TS]

  through the thread that came down with [TS]

  the needle and you get stitches so if [TS]

  you stare at this gif animation long [TS]

  enough you should get an idea of like [TS]

  the basics of how a sewing machine works [TS]

  and it's actually you see what it's [TS]

  doing it's like you know if a human [TS]

  being was doing it they wouldn't [TS]

  probably do it in this exactly this way [TS]

  but you end up getting a bunch of [TS]

  interlocked stitches with one set of [TS]

  thread going down the needle and the [TS]

  other set of thread on the bobbin and [TS]

  the interlock with each other and so two [TS]

  pieces of fabric together and so this is [TS]

  a pretty ingenious machine I also put in [TS]

  two links to a very old show called the [TS]

  secret [TS]

  life of machines oh yeah which was a [TS]

  british show most notable to my young [TS]

  American self as the program where i [TS]

  learned what sticky tape is as opposed [TS]

  to non sticky tape i guess that's what [TS]

  they call scotch tape and Scotch is the [TS]

  name brand anyway put a link to that in [TS]

  the show notes it's like The Secret Life [TS]

  machine is a sewing machine part 1 and [TS]

  part 2 I actually encourage you to watch [TS]

  these things yes they look like they're [TS]

  on terrible video from the 70s or 80s [TS]

  and the people have British accents and [TS]

  they're blurry but really watch it and I [TS]

  think by the end of watching these [TS]

  programs unlike by the end of this [TS]

  podcast you will actually understand how [TS]

  sewing machines work and I think it's [TS]

  important because you can see you know [TS]

  how how they figured out how to do this [TS]

  how to take this very complex series of [TS]

  motions having one piece of thread [TS]

  threaded through a hole in fabric and [TS]

  take another piece of thread and loop it [TS]

  through that and then come back out how [TS]

  can you do that like without having a [TS]

  human being threading things through a [TS]

  little hole and guiding it through and [TS]

  the Machine they came up with to do it [TS]

  looks almost nothing like it would look [TS]

  like if your hands did it but it it is [TS]

  you know basically the bobbin is being [TS]

  passed through not literally but you [TS]

  know effectively the threat on the [TS]

  bobbin is being passed through a loop [TS]

  brought down by the other needle and [TS]

  it's quite an interesting little machine [TS]

  right this is very close to [TS]

  philosophically what Sisk is like like [TS]

  if you had to explain someone how to sew [TS]

  two things together uh with a particular [TS]

  locking stitch that holds two pieces of [TS]

  fabric together you can do it but that's [TS]

  a lot of steps and you'll be like [TS]

  wouldn't it be better if I could just [TS]

  sort of feed a piece of fabric through a [TS]

  machine I say it machine look just do [TS]

  that stitch thing I don't wanna have to [TS]

  tell you all the individual steps so it [TS]

  basically comes down to is you gotta [TS]

  punch needle through and then something [TS]

  grabs little thing and then you gotta [TS]

  have a little bobbin it tucks its thread [TS]

  through there but I don't have typed I [TS]

  just want to have instruction called [TS]

  stitch and I hit a pedal on the floor [TS]

  anything goes stitch and I'm I just hold [TS]

  down that pedal go stitch stitch stitch [TS]

  stitch stage so I can feed the fabric [TS]

  through and I've made a sewing machine [TS]

  is awesome sisk processors were like [TS]

  sewing machines and the cooler steps [TS]

  they could do the more interesting [TS]

  stitches that they could do the more [TS]

  powerful those CPUs were but like a [TS]

  sewing machine like there's a limit on [TS]

  how fast the sewing machine can do what [TS]

  it does if you stare at that animation [TS]

  you're like okay how fast can this [TS]

  animation go can I do like five [TS]

  just a second ten stitches a second [TS]

  maybe tonight of 20 maybe but now you [TS]

  know my 20 stitches a second that's [TS]

  pretty darn fast can you make a machine [TS]

  that performs a series of complicated [TS]

  moves that best can I do 200,000 [TS]

  stitches a second probably not someone [TS]

  someone from the industrial sewing [TS]

  machine world will send me some feedback [TS]

  on this show and tell me how fast the [TS]

  fastest sewing machine in the world can [TS]

  go go but like at a certain point [TS]

  pressing the pedal to do one stitch and [TS]

  it's like you told me to do one stitch [TS]

  and I got into this series of 27 really [TS]

  complicated steps and the steps depend [TS]

  on each other the kind of interlocked [TS]

  and you know you get into issues of like [TS]

  physics of things hurtling past each [TS]

  other and you can't thread this thing [TS]

  through until the needle comes down but [TS]

  you got to make sure the needle is all [TS]

  the way down before you grab it then you [TS]

  but you have to start and stop a piece [TS]

  of machinery from moving it's there is a [TS]

  limit how fast a sewing machine can [TS]

  operate compare that limit of how fast [TS]

  can I make this little sewing machine [TS]

  thing go compare that limit to how fast [TS]

  an electric motor can spin just plain [TS]

  old straight barrel electric motor a [TS]

  sewing machine you know a thousand rpm [TS]

  100 rpm and you know stitches per minute [TS]

  rotations from it or whatever how many [TS]

  stitches can you make per second doesn't [TS]

  come close to an electric motor electric [TS]

  motor can do tens of thousands of [TS]

  rotations per minute a electric motor [TS]

  can spend very very fast you just you [TS]

  know it you put it in some really good [TS]

  ball bearings you have some magnets in a [TS]

  coil of wire and electricity maybe you [TS]

  even magnetically levitates spindles so [TS]

  there's no friction except for air that [TS]

  you could suck the air out of it to make [TS]

  it a vacuum you know you can make [TS]

  electric motor spin really fast because [TS]

  all does is spin there's no oscillating [TS]

  parts there's no thread going through a [TS]

  piece of fabric and one piece of thread [TS]

  being shoved through a nose but it's [TS]

  just spinning that's all does it rotates [TS]

  around again and again and you can do [TS]

  that really really really fast and if [TS]

  you want to make that luxury motor go [TS]

  really fast you know right where you [TS]

  have to concentrate what are the things [TS]

  stopping me from going too fast friction [TS]

  uh that the force inside of the thing [TS]

  I'm making it spin apart if it goes too [TS]

  fast you know that isn't the centripetal [TS]

  force on the outsides of the piece of [TS]

  metal causing to disintegrate but those [TS]

  limits are way higher than the limits of [TS]

  a sewing machine if you had a contest [TS]

  between the best engineers in the world [TS]

  how fast can you make a sewing machine [TS]

  go versus how fast can you make an [TS]

  electrical [TS]

  the electric motor guys would destroy [TS]

  the sewing machine and sew the risk [TS]

  philosophy was that they figured this [TS]

  out you know they said all right we're [TS]

  making these awesome cpu's with these [TS]

  really complicated instructions but geez [TS]

  every time we try to make these things [TS]

  go faster like how fast can we make that [TS]

  instruction to do some complicated [TS]

  operation I guess of all these steps [TS]

  involved and they're all interlocked [TS]

  with each other and this limit how fast [TS]

  I can go with that what if we make the [TS]

  CPU equivalent of that electric motor [TS]

  where what it does is really stupid and [TS]

  simple but we can do it really really [TS]

  fast uh and so that's that's a little [TS]

  definition of risk here from the the [TS]

  Wikipedia article CPU design strategy [TS]

  based on the inside that simplified as [TS]

  opposed to complex instructions can [TS]

  provide higher performance this [TS]

  simplicity enables much faster execution [TS]

  of each instruction so an example of a [TS]

  RISC architecture is the ARM [TS]

  architecture it's a emergency for acorn [TS]

  risc machines i believe and i forget [TS]

  what they we went through a [TS]

  senate-passed show I'm not gonna look it [TS]

  up again there's only the second we have [TS]

  feedback about if any way is a risk [TS]

  machine and they use called like load [TS]

  store architectures and the reason load [TS]

  store architecture comes up is because [TS]

  given that complex instruction operation [TS]

  from before take a piece of memory from [TS]

  location 57 and move it to 128 on the [TS]

  risc cpu that would be read the location [TS]

  57 into register one right location [TS]

  right register one out into location in [TS]

  256 so it would be two operations [TS]

  instead of one I may have gotten that [TS]

  wrong with the details so basically you [TS]

  can take any complicated operation and [TS]

  decompose it into simpler operations so [TS]

  the move from one place member to [TS]

  another it's going to have to go through [TS]

  the CPU anyway there is no instruction [TS]

  that says move it from the settle [TS]

  occasion the other one you have to issue [TS]

  the two taken from here put it to their [TS]

  take a rear porter there and that seems [TS]

  like a loss like wasn't it much better [TS]

  when you could just tell it to move from [TS]

  one place to the other but in actuality [TS]

  because those those risk instructions [TS]

  are so simple you can make them go much [TS]

  faster and you can interleave them with [TS]

  each other and duels all sorts of other [TS]

  things because the instructions [TS]

  themselves are small and so everything [TS]

  about risk machines was designed to be [TS]

  more like the electric motor and less [TS]

  like the sewing machine they [TS]

  won't have any special hardware special [TS]

  registers that only serve a single [TS]

  purpose when you divide two numbers the [TS]

  result always goes in this register they [TS]

  wanted to have a big very uniform [TS]

  register file with lots of what they [TS]

  call general purpose registers we're [TS]

  like when you divide two numbers the [TS]

  result doesn't always have to go into [TS]

  this register it can go in any of the [TS]

  registers in fact all the registers the [TS]

  same and we have a bunch of them because [TS]

  there's going to be a bunch of these [TS]

  little tiny instructions and flight at [TS]

  once and maybe we can put more than [TS]

  parallel and intro leave them with each [TS]

  other we just want to make it regular [TS]

  and uniform they did stuff like uh you [TS]

  know either discouraging or outright [TS]

  forbidding unaligned memory access you [TS]

  know old machines would be like we can [TS]

  read memory anywhere you don't have to [TS]

  start from a particular location you [TS]

  just tell me you can read you a bite at [TS]

  a time from anywhere in memory and then [TS]

  risk machines are more like now you can [TS]

  only do a line memory access you have to [TS]

  start on on a you know a multiple of 32 [TS]

  bits or multiple 16 bits or something [TS]

  some of the trunk like that if that's [TS]

  inconvenient for you tough luck take up [TS]

  the bigger piece bring it into the CPU [TS]

  chop it up how you want uh and it's [TS]

  basically you know making it much much [TS]

  less pleasant to program an assembly [TS]

  language because you're getting a bit [TS]

  much dumber simpler machine that has [TS]

  much more limitations and so many ones [TS]

  into assembly programming be like I [TS]

  don't want to program a risc cpu I got [TS]

  to tell it how to do everything and have [TS]

  instructions i wanted to aren't even [TS]

  there I can't even pull a byte of memory [TS]

  from here I gotta pull like 32 bits of [TS]

  memory and mask it to get the bytes I [TS]

  want and it's just this terrible right [TS]

  uh that's not what I was designed for it [TS]

  was designed for an age where compilers [TS]

  right that for you and optimize it for [TS]

  you purely because like the electric [TS]

  motor they could spin a risc cpu really [TS]

  really fast and in CPU world spinning [TS]

  really fast is the clock speed it's [TS]

  megahertz how many instructions per [TS]

  second or how many cycles per second can [TS]

  can happen oh the other thing that had [TS]

  done these on risky p uses they would [TS]

  try to make almost every single [TS]

  instruction executes in a single clock [TS]

  not everyone because this you know [TS]

  complicated division instructions and [TS]

  stuff like that but they would want they [TS]

  would not want to have an instruction [TS]

  like a one sis cpus where it's some big [TS]

  complicated instruction like the song [TS]

  machine it takes 17 cycles to execute [TS]

  you know adding two numbers here they [TS]

  wanted all their instructions to be so [TS]

  simple that every single one of them [TS]

  executed in a single cycle except for [TS]

  maybe a couple of outliers so again the [TS]

  entire instruction set the entire CPU [TS]

  architecture was designed to make the [TS]

  CPU go fast and not to make assembly [TS]

  programmers like you [TS]

  ye're and and or you know anything like [TS]

  that they've expected a compiler to spit [TS]

  out this stuff all right let me see what [TS]

  up man these are there's some good rants [TS]

  for me today yeah I'm really revisiting [TS]

  old territory all right let's do the [TS]

  second sponsor lynda.com online learning [TS]

  company these guys have more than 77,000 [TS]

  video tutorials i call them screencasts [TS]

  they call video tutorials they're gonna [TS]

  have to they're you know that's their [TS]

  thing they know what the call they teach [TS]

  you stuff they teach you everything from [TS]

  design photography business programming [TS]

  a you name it audio and video stuff even [TS]

  have 3d animation stuff in there they're [TS]

  adding new courses all the time all the [TS]

  time and the way this thing works is you [TS]

  get a membership and then you get access [TS]

  to everything that's in there so that [TS]

  you can teach yourself stop from experts [TS]

  they have experts who are working in the [TS]

  industry with real-world experience [TS]

  creating these things teaching these [TS]

  things and selling them there and it's [TS]

  really really cool I love learning with [TS]

  screencast and of course I mean I'm [TS]

  probably biased because I've made [TS]

  screencast before I used to make a [TS]

  living doing screencast but I love this [TS]

  stuff it's over on lynda.com that [TS]

  spelled ly n da lynda.com membership [TS]

  starts at 25 bucks a month but you get [TS]

  unlimited 24 by 7 access to the best [TS]

  videos out there they made a special URL [TS]

  you can visit linda.com / 5 by 5 and if [TS]

  you go there you'll get seven days for [TS]

  free access to everything seven days [TS]

  free and you'll support the show you'll [TS]

  make John Happ you'll make me happy [TS]

  Linda calm / 5 / 5 check them out so we [TS]

  got some numbers on actual risk and CPUs [TS]

  in terms of number of registers and [TS]

  stuff just to reinforce the point about [TS]

  the register whether they call the [TS]

  register file how many differenter says [TS]

  you have so the PowerPC cpu used to be [TS]

  in max that's also a risc cpu and like i [TS]

  said risks abused and have way more [TS]

  registers and they're all kind of you [TS]

  know no special purpose registers or few [TS]

  special purpose or mostly generate lots [TS]

  of general-purpose registers so power [TS]

  busy had 32 general purpose registers [TS]

  and 32 floating point registers and and [TS]

  compared to x86 intel chips back when [TS]

  they were 16 bit x86 chips [TS]

  had six registers which were kind of [TS]

  sort of dedicated special purpose not [TS]

  all them are really general sort of like [TS]

  I said you know when you did a divided [TS]

  instruction the result always had to go [TS]

  in a particular register and stuff like [TS]

  that plus a couple of other things for [TS]

  like stack pointers and stuff 32-bit x86 [TS]

  had six general purpose registers plus [TS]

  some extended versions of the you know [TS]

  stack pointer registers and stuff 64-bit [TS]

  they added a few more you got 14 general [TS]

  purpose registers plus the other ones [TS]

  and so you see like x86 has been trying [TS]

  to add more registers but PowerPC from [TS]

  the day was born had this you know huge [TS]

  number registered more registers than [TS]

  you would expect the human being to keep [TS]

  track of because if you're writing [TS]

  assembling you got to keep track of 32 [TS]

  registers like I will tell that i put in [TS]

  register 17 i want to add these two [TS]

  numbers together was that in register 32 [TS]

  or 57 it's just you can imagine writing [TS]

  so maybe you can remember adding [TS]

  assembly by hand and like it's hard to [TS]

  keep track of these things that they're [TS]

  all just called r 1 r 7 r 15 are three [TS]

  and they're all the same and how can you [TS]

  keep track of where anything is like [TS]

  it's totally made for a machine because [TS]

  the machine say okay I've got 32 [TS]

  registers to work with and I can just [TS]

  keep track of them and have sort of a [TS]

  register allocation algorithm I figure [TS]

  out how to use them when I compile this [TS]

  code it's much better for a machine a [TS]

  floating point is a good example to on [TS]

  x86 based have a separate floating-point [TS]

  unit called the x87 stack-based [TS]

  floating-point unit which was a a [TS]

  dedicated separate little thing itself [TS]

  would eventually moved on die but it was [TS]

  stack base where you put your number [TS]

  push another number and then push on a [TS]

  multiplication would multiply the top [TS]

  two numbers on the stack but you could [TS]

  also reach down lower into the stack and [TS]

  do stuff and I it's not really conducive [TS]

  to parallelization because when you have [TS]

  a stack it's like a stack of plates [TS]

  where you pushed push two plates on the [TS]

  top and perform an operation that [TS]

  combines those two plates into one plate [TS]

  and puts the result back on top of the [TS]

  stack and if you want have 17 people [TS]

  operating at the same time on a stack [TS]

  it's not really great because of the way [TS]

  a stack works you can have some guy in [TS]

  the middle yanking things in and out [TS]

  because what if the guy on the top was [TS]

  doing yet anyway stack-based floating [TS]

  point was also a problem for them [TS]

  because it was kind of designed in in a [TS]

  different error when the most important [TS]

  thing wasn't like parallel execution and [TS]

  simplicity of things now what all is [TS]

  article that is referred to at the very [TS]

  front of the article by John Stokes were [TS]

  in reference to like Cisco versus risk [TS]

  at the post ristic post risk era that [TS]

  was all [TS]

  I'll sort of rehashing this cisco stuff [TS]

  that I just talked about and explaining [TS]

  how although that was the origin of the [TS]

  risk movement it's much less relevant [TS]

  today because these days when you have [TS]

  this x86 quote-unquote CPU have these [TS]

  complex complex instructions the way [TS]

  they work internally is they take that [TS]

  complicated instruction that says move [TS]

  something from one memory address to [TS]

  another and RAM and they break it down [TS]

  into what they call micro operations so [TS]

  they say okay even though you said move [TS]

  this data word from address here at [TS]

  address there I'm going to break that [TS]

  single instruction into two smaller [TS]

  instructions one that says get that [TS]

  thing from memory and put it into this [TS]

  register and other that says take that [TS]

  register and stick and into memory all [TS]

  right so they basically taken a sisk [TS]

  instruction set and inside the CPU break [TS]

  it down into something more [TS]

  approximating a RISC instruction set and [TS]

  the actual CPU is built to execute these [TS]

  tiny little micro operations it's not [TS]

  executing you know it doesn't the CPU [TS]

  doesn't see this big complicated [TS]

  instruction that says to do 50 things it [TS]

  sees a stream of tiny little micro [TS]

  operations that were created from those [TS]

  single instructions so they build the [TS]

  entire CPU internally to execute these [TS]

  micro operations very efficiently and [TS]

  very quickly and they just have a [TS]

  decoder part on the front on the front [TS]

  end that takes your complicated [TS]

  instruction see and it decodes it into [TS]

  these simpler ones and that's the idea [TS]

  of getting at the post risk era where [TS]

  it's like all right so the instruction [TS]

  set itself how relevant is that if [TS]

  everybody inside is building basically [TS]

  risk CPUs and if you have a complex [TS]

  instructions that you just need a [TS]

  decoder now part of the CPU and an x86 [TS]

  chip is dedicated to all the stuff that [TS]

  has to say okay I've got this big [TS]

  complicated instruction coming in how [TS]

  many micro operations does that break [TS]

  down to can i have some piece of [TS]

  hardware that automatically breaks that [TS]

  down into the micro operations or is it [TS]

  a really complicated one but i have to [TS]

  have like a little software microcode [TS]

  program that tells me how to break this [TS]

  down you need to dedicate part of the [TS]

  chip just to dealing with this thing and [TS]

  by the way have to keep track of these [TS]

  little micro operations you have to sort [TS]

  of reassemble the result in to like make [TS]

  it look as if that complicated [TS]

  instruction really executed like you [TS]

  can't have things go out of order it has [TS]

  to look to the assembly programmer just [TS]

  like those old x86 chips from Intel that [TS]

  actually did implement this complex [TS]

  instruction set internally despite [TS]

  fact that inside is basically a risk [TS]

  like CPU executing all the stuff so you [TS]

  have all this the front-end hardware and [TS]

  the tracking hardware and the [TS]

  reassembling of things and making sure [TS]

  the exact timings and exact rules are [TS]

  maintained and they behave exactly the [TS]

  same as they used to uh and the argument [TS]

  back in the 90s was that all right so [TS]

  Intel may have figured out that this [TS]

  wrist thing is a good idea because they [TS]

  can make their CPUs run faster but they [TS]

  have to dedicate all this time and [TS]

  energy and parts of the chip and like [TS]

  they have to you know they basically [TS]

  have to have this whole whole bunch of [TS]

  gunk inside their CPUs to translate [TS]

  their crappy complicated instruction set [TS]

  into the simpler one and keep track of [TS]

  everything and that's bad because you [TS]

  know the PowerPC you guys don't need to [TS]

  do that they don't have the whole [TS]

  section of the chip doesn't have to [TS]

  exist to break down their instructions [TS]

  into micro operations and stuff like [TS]

  that is they already are micro and [TS]

  everything's fine so therefore PowerPC [TS]

  is going to win and risk is awesome and [TS]

  Intel drools and so on so forth and from [TS]

  a technical perspective I remember [TS]

  feeling that way that like Intel looks [TS]

  like the crusty crappy thing you're like [TS]

  well they may have clever engineers and [TS]

  they may have figured out to break up [TS]

  their instructions to smaller ones and [TS]

  stuff but they gotta deal with all those [TS]

  that you know making that stuff to break [TS]

  up the instructions and keeping track of [TS]

  them and reassembling the results and [TS]

  keeping the order or consistent that's [TS]

  thanks x86 thinks it's ugly it's gross [TS]

  right Intel also basically agreed that [TS]

  it was ugly in gross and they trot have [TS]

  been trying for years and years to get [TS]

  rid of the ugliest and grossest part so [TS]

  the stack based floating-point that's [TS]

  not really great for doing things [TS]

  quickly or in parallel they've been [TS]

  deprecating that they have to keep [TS]

  supporting it but they're like look [TS]

  don't if if you're writing a compiler [TS]

  for example I and someone's doing [TS]

  floating-point math don't generate x87 [TS]

  instructions for the stack place [TS]

  floating-point thing because that's slow [TS]

  and it's always been slow and it's not [TS]

  going to probably get much faster we [TS]

  still have to support it you just saw [TS]

  all your old programs are running stuff [TS]

  but instead what we have here is [TS]

  something that they called SSE streaming [TS]

  simdi extensions cingulate single [TS]

  instruction multiple data it's a way to [TS]

  process lots of data in parallel once [TS]

  like if you have you know two sets of 50 [TS]

  numbers and you want to add the first [TS]

  set of 50 numbers to the second set of [TS]

  50 numbers you're have instruction that [TS]

  does it all at once and that may not [TS]

  seem like it's the same thing as [TS]

  floating point [TS]

  but it also operates on floating point [TS]

  numbers and they said look compiler [TS]

  writers even though you may not think [TS]

  you're doing like you know doing a [TS]

  single instruction on 50 pieces of data [TS]

  the floating part hardware we have in [TS]

  this in these SSE parts of our chip is [TS]

  way faster in that stack based one we [TS]

  can do stuff in parallel don't generate [TS]

  instructions for a stack-based floating [TS]

  point if you see any floating point code [TS]

  in your you know high level you know no [TS]

  higher level bin assembly your C program [TS]

  or whatever just target the RSS see [TS]

  extensions and that will make your stuff [TS]

  go faster so they have they had four [TS]

  versions of SSC so far i think is for [TS]

  you know I just keep version them and [TS]

  making them more and more powerful and [TS]

  these days if you compile a program for [TS]

  an Intel chip you would hope that [TS]

  there's no stack-based floating-point [TS]

  stuff being sent to the chip despite the [TS]

  fact that it the chip is still able do [TS]

  that all the floating-point stuff is [TS]

  targeting this more modern more [TS]

  parallelizable non stack base flowing [TS]

  point Intel also went for 32-bit 64-bit [TS]

  and they took that while not Intel but [TS]

  AMD actually AMD came up with a x86 64 [TS]

  which is a 64-bit bunch of instructions [TS]

  that only work on 64-bit chips chips ba [TS]

  that are compatible with old x86 [TS]

  programs and they took that as an [TS]

  opportunity to make a new instruction [TS]

  set that is more amenable to modern ship [TS]

  designs than the old one because there [TS]

  were no 64-bit instructions at all [TS]

  before there was 64 bit chicks so we [TS]

  don't have to just make like we don't [TS]

  have to stick with the old ways this is [TS]

  an opportunity this is a discontinuity [TS]

  to come up with a better instruction set [TS]

  they'd added registers they you know [TS]

  made the rules about how the the 64-bit [TS]

  instructions execute make it easier to [TS]

  do encode and decode instruction [TS]

  tracking and parallelization a lot stuff [TS]

  Intel was off doing its own thing with [TS]

  the Itanium and another right I remember [TS]

  that day talk about a very long word [TS]

  instruction computing all stuff that's [TS]

  whole other topic but anyway so if I sit [TS]

  decide that Intel's thing didn't quite [TS]

  work out x86 64 caught on and now Intel [TS]

  makes chips with x86 64 but this wasn't [TS]

  another opportunity for them to get rid [TS]

  of old crusty stuff say if you're [TS]

  targeting if your compiler writer and [TS]

  you're targeting 64-bit CPUs use these [TS]

  cool new instructions and you have [TS]

  access to more registers and see isn't [TS]

  this thing getting [TS]

  sir isn't our cpu architecture getting [TS]

  to be more like those nice clean RISC [TS]

  architectures where you had just had a [TS]

  bunch of general purpose registers and [TS]

  bunch of general purpose registers and [TS]

  simple instruction set we're not there [TS]

  yet but we're slowly trying to abandon [TS]

  our craft right so that that is the path [TS]

  that that Intel the the leading CPU [TS]

  vendor with its complex instruction sets [TS]

  abused has been pursuing to try to reap [TS]

  the advantages of the risk revolution [TS]

  without actually having a RISC [TS]

  instruction set on it CPUs so hold this [TS]

  thought about risk versus cysts for a [TS]

  moment that hopefully you have some at [TS]

  least sort of feel for and we do a [TS]

  sidebar into one other topic that will [TS]

  come back and turn out to be relevant [TS]

  all right so all the stuff we've been [TS]

  talking about is you know for processor [TS]

  so what is a processor and how do we [TS]

  make it that turns out to be very [TS]

  important you all know the process a [TS]

  little black chip that we see in the [TS]

  diagram is the flat thing it's got a [TS]

  little metal contact somewhere on it and [TS]

  inside that little black package is a [TS]

  semiconductor it's like a piece of [TS]

  silicon it's very flat and the way they [TS]

  make stuff on it is they use lithography [TS]

  to etch things onto it and with ography [TS]

  is just basically shining light or some [TS]

  other light like thing through a mask [TS]

  that block some of the light and let [TS]

  some through and they use materials and [TS]

  techniques so that when the light hits a [TS]

  certain area it changes the material in [TS]

  one way and like doesn't hit the area it [TS]

  leaves the material the other way and [TS]

  they do this to make very small things [TS]

  basically by using different kinds of [TS]

  light and different ways to focus it and [TS]

  you know other techniques they can make [TS]

  very small features that no machine [TS]

  could ever edge like the little tiny [TS]

  things that are on these chips but like [TS]

  that they they can focus a light another [TS]

  I don't know what you would call it but [TS]

  electromagnetic radiation into very [TS]

  small areas and so they can make the [TS]

  mask actually relatively big and then [TS]

  use lenses and stuff to focus it down [TS]

  and make different features they do this [TS]

  again and again until they wear away the [TS]

  material on certain things and what [TS]

  they're basically trying to make is a [TS]

  transistor think thing they can switch [TS]

  on and off and they do this by having [TS]

  doing layers of material that allow [TS]

  ektron electrons to flow through them or [TS]

  not depending on if some current is [TS]

  applied to some other layer of material [TS]

  but the point is that these things are [TS]

  very very small and the way they [TS]

  describe the size of the [TS]

  things that are etched on here with [TS]

  lithography is called the feature size [TS]

  or sometimes the process node I tried to [TS]

  look up a good definition to this we all [TS]

  hear these numbers bandied about we've [TS]

  talked about on this show when we say [TS]

  like oh the they're using ad I Shrunk [TS]

  version of the a5 in the ipad 2 and the [TS]

  new ipad 2 is it used to be about forty [TS]

  five nanometers and now it's 32 [TS]

  nanometers alright so nanometers is how [TS]

  they measure that's that's the scale [TS]

  things are measured these days it used [TS]

  to be was it a microns or micrometers [TS]

  and remember what it was before we went [TS]

  down to nanometers but they it nanometer [TS]

  is very small right and what they're [TS]

  measuring is the call the half pitch is [TS]

  the half that half the distance between [TS]

  identical features on the edge piece of [TS]

  silicon usually for a memory cell [TS]

  because memory cells are very very [TS]

  regular doesn't mean that every single [TS]

  thing on a 32 nanometer chip is 32 [TS]

  nanometers wide or 32 nanometers apart [TS]

  but it's like it's a they need some way [TS]

  to measure these things because larger [TS]

  features are chunkier and smaller [TS]

  simpler features can be skinnier but [TS]

  they've just come up with a single [TS]

  standard for how we talk about how small [TS]

  you can make stuff and that number is in [TS]

  marching down for all of our lives you [TS]

  used to be you could not make things [TS]

  that were you know very small at all [TS]

  compared to what we have now and then we [TS]

  just keep getting smaller and smaller [TS]

  and smaller and smaller every year and [TS]

  that means the entire thing shrinks [TS]

  because you can make every piece smaller [TS]

  than the whole thing gets smaller and [TS]

  smaller and smaller right so Ivy Bridge [TS]

  processors that's Intel's processes that [TS]

  in the macbook air is macbook pros [TS]

  Intel's current crop of desktop and [TS]

  laptop CPUs they are done at the 22 [TS]

  nanometer process note and they have all [TS]

  bunch of other features you know about [TS]

  their 3d transistors different ways to [TS]

  etch things onto the the silicon thing [TS]

  so too can jam more stuff in a closed [TS]

  the thing but bottom line is that there [TS]

  are 22 nanometers right uh and in [TS]

  general as you shrink this feature size [TS]

  as you make everything smaller and [TS]

  smaller almost everything improves the [TS]

  cost improves tremendously because your [TS]

  unit of cost is like how many of these [TS]

  little chips can I fit onto a silicon [TS]

  wafer so looking way for some big [TS]

  circular thing maybe the size of like a [TS]

  basketball or something you know that [TS]

  and if you can fit 10 chips on there and [TS]

  it costs you you know a hundred dollars [TS]

  to make one of those wafers then each [TS]

  chip cost you ten bucks to make but if [TS]

  you can get a thousand ships on that [TS]

  same way for suddenly the cost per chip [TS]

  goes way down so shrinking things is [TS]

  good because your costs are like I cost [TS]

  me a certain amount to etch a single [TS]

  piece of silicon and that costs like is [TS]

  dependent on a whole bunch other factors [TS]

  but the point is if I can jam more and [TS]

  more cpus onto that piece of silicon the [TS]

  cost of making it doesn't go go you know [TS]

  doesn't go up I would love to be able to [TS]

  put more things on here so the more [TS]

  things you can spit on to a single [TS]

  silicon wafer you know the better it is [TS]

  for you right the power required to run [TS]

  these dinky little things goes down [TS]

  because if you can make them smaller and [TS]

  smaller if you can imagine if you have [TS]

  to run it chip the size of a dinner [TS]

  plate a single chip it would take a lot [TS]

  of power to run that thing when you make [TS]

  them small you don't need as much power [TS]

  uh and the speed goes up for variety of [TS]

  reasons but even just like in it the at [TS]

  the very high end of the scale look if [TS]

  the thing was the size of a basketball [TS]

  it takes a certain amount of time for an [TS]

  electrical signal or propagate from the [TS]

  upper left corner to lower right corner [TS]

  whereas if the thing is the size of your [TS]

  thumbnail that distance shrinks and when [TS]

  you're going really really fast they may [TS]

  seem like oh I'll don't the electrons [TS]

  travel close to the speed of light and [TS]

  doesn't make that well it makes a [TS]

  difference when you're trying to you [TS]

  know go billions of cycles per second or [TS]

  whatever how r this one thing that that [TS]

  is a problem when you may think smaller [TS]

  you make them really really small [TS]

  suddenly it becomes really easy for [TS]

  electrons that you didn't want to go [TS]

  somewhere just sort of just slip through [TS]

  like you make things so small that just [TS]

  random electrons kind of rattling around [TS]

  end up sort of leaking through or you [TS]

  didn't want them to go just because [TS]

  everything is so small like when things [TS]

  are big and chunky you can put a big [TS]

  wall between uh no electrons can go [TS]

  through here I'm totally putting down [TS]

  this big wall but when things are really [TS]

  really really tiny that wall is so [TS]

  skinny that occasionally maybe a couple [TS]

  electrons start floating through there [TS]

  and that's called leakage current and [TS]

  there are various techniques to try to [TS]

  control that but that's when you start [TS]

  getting really small like down to the [TS]

  size of like you were you measuring [TS]

  feature sizes and you know how many [TS]

  atoms are there it between this area in [TS]

  that area then you have some problems [TS]

  but in general doing shrinks is better [TS]

  there's another article again by John [TS]

  Stokes also an older our goal talking [TS]

  about understanding Moore's [TS]

  aw Moore's Law is the thing we hear [TS]

  thrown around in the world of computing [TS]

  a lot that most light people i think [TS]

  understand as Moore's law means [TS]

  computers get faster and faster all the [TS]

  time which is not what Moore's law [TS]

  actually says it's like a third level [TS]

  consequence of it but Moore's law was [TS]

  set as saying as he was just noting that [TS]

  in the early days of semiconductors the [TS]

  number of transitions we can shove on to [TS]

  a particular area keeps doubling every [TS]

  12 months basically he was noting that [TS]

  they they keep being able to shrink [TS]

  their feature sizes so before on this [TS]

  one silicon wafer we could fit 10 CPUs [TS]

  then 12 months from now we can fit 20 [TS]

  and after that we can put 40 in it and [TS]

  12 months after that we can fit ad and [TS]

  it just keeps going up and up and he did [TS]

  the graph means like wow soon we're [TS]

  going to be able to shove huge amount of [TS]

  transistors on to yeah right now we can [TS]

  only do chips with a thousand [TS]

  transistors of the size of your [TS]

  thumbnail soon we'll be able to do chips [TS]

  with you know ten times that and [TS]

  especially in the early days of [TS]

  computing if you gave a chip designer [TS]

  more transistors to work with they could [TS]

  make a more powerful CPU beach like all [TS]

  before I could I can only add two [TS]

  numbers together at once but now I have [TS]

  double the number of transistors I can [TS]

  just add another a door right nice that [TS]

  and that they can both go at the same [TS]

  time and i can add four numbers together [TS]

  at the same time isn't that great and i [TS]

  can add more memory and more cache [TS]

  memory and also the things that get [TS]

  better with shrinking things so being [TS]

  able to have more transistors at your [TS]

  disposal has led to you know for the [TS]

  first couple decades of cpu development [TS]

  every year computers would get faster [TS]

  and faster because those guys had more [TS]

  cpus that they could deal with and you [TS]

  know they could either make your your [TS]

  thing your existing cpu cost half as [TS]

  much or they can make a new cpu that [TS]

  cost the same amount that goes twice as [TS]

  fast and that was a great run for how [TS]

  long it lasted but not in general the [TS]

  number of transistors doesn't exactly [TS]

  equal speed but in the early days it did [TS]

  now this was during the the risk versus [TS]

  sisk wars we're like oh we've got this [TS]

  increasing number of transistors but [TS]

  your instruction set is stupid and we [TS]

  have a better one and you've got a [TS]

  parole this decode hardware on there [TS]

  everything and risk was technically [TS]

  superior because it didn't you know they [TS]

  said we found a way to make you know [TS]

  you're doing a sewing machine and we've [TS]

  got an electric motor and we can make [TS]

  this go so fast look at this we've got a [TS]

  1 gigahertz alpha processor you can't [TS]

  get the 1 gigahertz because you can't [TS]

  run your sewing machine that fast and so [TS]

  tell guys had to go back to the drawing [TS]

  board how how how the hell can we make [TS]

  this thing go faster we've got to break [TS]

  things things down into small pieces and [TS]

  then do our little internal core that [TS]

  runs fast then have the external part [TS]

  track everything and like now the risk [TS]

  guys were laughing at him right you know [TS]

  you guys suck you made a bad choice you [TS]

  know you were there early we come in [TS]

  later we have a better solution we're [TS]

  going to crush you the dec alpha the [TS]

  PowerPC all sorts of killer chips from [TS]

  these risc cpu vendors in like the the [TS]

  90s we're going to put Intel out of [TS]

  business now Intel smart and they do a [TS]

  lot stuff with decoding their [TS]

  instructions and stuff like that but the [TS]

  real thing that saved Intel's bacon is [TS]

  the aforementioned Moore's law that [TS]

  every year you could fit more and more [TS]

  transistors on to the same size chip ah [TS]

  and that is relevant because as the [TS]

  number of transistors per unit area [TS]

  increases and you know as the number of [TS]

  transistors available to given cost [TS]

  increases the relative percentage of the [TS]

  total transistor count used by all that [TS]

  ugly x86 decoding and tracking hardware [TS]

  becomes a smaller and smaller percentage [TS]

  of the total alright so you can do a [TS]

  complete x86 implementation in like [TS]

  29,000 transistors like the 8086 at [TS]

  29,000 transistors right obviously [TS]

  modern CPUs need you know they're doing [TS]

  more complicated stuff than the 8086 but [TS]

  it goes goes to show the like how many [TS]

  transistors do you need to take x86 [TS]

  instructions crack them into micro [TS]

  operations track them as they go through [TS]

  the machinery and reassemble them in the [TS]

  end it's more than 29,000 ah you know [TS]

  and of course it's not it's not as [TS]

  simple as that was but I've you bid [TS]

  transistors the current like I've your [TS]

  bridge line of CPUs have 1.4 billion [TS]

  transistors they have billion [TS]

  transistors in them that's a lot right [TS]

  so what percentage of that 1.4 billion [TS]

  transistors is dedicated to dealing with [TS]

  x86 decode tracking and reassembly stuff [TS]

  the the press what eventually happened [TS]

  was yeah x86 was uglier and the other [TS]

  market forces involved or whatever but [TS]

  the the penalty for having a crappier [TS]

  instruction set just started to shrink [TS]

  and shrink and Rigby's transistors were [TS]

  just falling out of the sky [TS]

  it's like oh we gotta dedicate half-hour [TS]

  chip to do deal with x86 okay now we [TS]

  have to get a fourth ok now we're [TS]

  dedicating eighth now dedicating a [TS]

  sixteenth eventually it's like it [TS]

  doesn't matter you know you know how [TS]

  many CPUs it takes for us to deal with [TS]

  x86 it is an insignificant percentage of [TS]

  the total number of CPUs of transistors [TS]

  in the CPU in fact most of the CPU like [TS]

  half of it is cash at this point you [TS]

  know but it's it doesn't matter that we [TS]

  have an uglier instruction set because [TS]

  of Moore's law and many other factors [TS]

  but in a large part because of Moore's [TS]

  law the x86 penalty for having to do all [TS]

  that ugly stuff went away and the risks [TS]

  iska is theoretical cool advantage of [TS]

  you know we have a simpler instruction [TS]

  set we don't have to deal ugly decoding [TS]

  and coding was no longer relevant and [TS]

  along the way in that process since it [TS]

  was so important for Intel to be able to [TS]

  have as many transistors as is at its [TS]

  disposal as a as it wanted and also [TS]

  because you know doing die shrinks have [TS]

  many other advantages Intel got really [TS]

  good at doing those shrinks like say [TS]

  okay this year we're going to be a 65 [TS]

  nanometer and next year we're going to [TS]

  be at 45 and then you know for that [TS]

  we're gonna be a 32 and they're just [TS]

  like a shrink shrink shrink because we [TS]

  want more and more transistors on our [TS]

  chips because it's you know it's it [TS]

  gives us an advantage we want to have as [TS]

  many as we want it helped them out when [TS]

  they were trying to decrease the burden [TS]

  of the x86 burden because like so what [TS]

  we got to do this decode stuff whatever [TS]

  we just need more see more princess just [TS]

  throw them on there right put on you [TS]

  know increase caches on our chips to [TS]

  make them faster all stuff like that [TS]

  yeah so so while the x86 burden is [TS]

  technically gross and upsetting to [TS]

  techno purists like me it at this point [TS]

  it has very little practical effect or [TS]

  rather its practical effects are able to [TS]

  be managed by Intel because then tells [TS]

  really smart and they have a lot of [TS]

  transistors and as mentioned earlier [TS]

  Intel's currently shipping 22 nanometer [TS]

  chips us there they're a feature size [TS]

  all right now Apple's a six they're cool [TS]

  fancy CPU that's in the iPhone 5 and the [TS]

  iPad for that is a 32 nanometer chip and [TS]

  32 is bigger than 22 and this highlights [TS]

  something [TS]

  for many years now but the past several [TS]

  years intel has consistently been one [TS]

  generation at least ahead of everybody [TS]

  else on feature size so intel has been [TS]

  shipping 22 nanometers ships for a long [TS]

  time and apples a six which just came [TS]

  out in the iphone five fairly recently [TS]

  is 32 nanometer Intel is ahead and then [TS]

  by the way it helps fit shipping [TS]

  full-fledged CPUs at 22 nanometers with [TS]

  1.4 billion transistors it's not even [TS]

  close to what the a6 is right most chip [TS]

  makers when they start off on a new [TS]

  process node the Merse they make memory [TS]

  because memory is very regular very [TS]

  simple not complicated at all they just [TS]

  make memory chips and after they've [TS]

  sorted that out okay we can make 32 [TS]

  nanometer memory chips let's try making [TS]

  some more complicated things and [TS]

  eventually they work their way up to big [TS]

  complicated cpu intel is shipping GPUs [TS]

  with billions of transistors at 22 [TS]

  nanometers Apple for all its amazing [TS]

  power and money is shipping 32 nanometer [TS]

  chips and this is in a context where [TS]

  like boy you know incredibly power [TS]

  sensitive you want it to be small you [TS]

  want to use not a lot of power wouldn't [TS]

  it be great if apple cup of 22 nanometer [TS]

  a sex why don't they because intel has [TS]

  been ahead of everyone else because they [TS]

  invest tons of money in it they have the [TS]

  most experience and it's been a pretty [TS]

  sustained advantage it's not like Oh a [TS]

  one-year Intel's ahead then another year [TS]

  AMD is ahead no intel has been the big [TS]

  dog here they are making ships a [TS]

  generation of head where everyone else [TS]

  has been doing in terms of feature size [TS]

  all which finally brings us back to [TS]

  Apple switching to arm and this may be a [TS]

  good time to do the final sponsor before [TS]

  I before I bring this home happy to [TS]

  oblige at squarespace.com everything you [TS]

  need to make an amazing website we use [TS]

  them for a lot of the blogging stuff [TS]

  that's kind of what they're known for [TS]

  that's kind of they got their start but [TS]

  you can build entire websites you can [TS]

  have multiple blogs on the web sites you [TS]

  can have galleries they're basically [TS]

  completely managed environment leidy [TS]

  create and maintain maintain is the key [TS]

  a website a blog portfolio what's cool [TS]

  about Squarespace is you don't have to [TS]

  know anything maybe your maybe all you [TS]

  do is worry about risk and sisk [TS]

  processors and you don't want to [TS]

  understand how to build a site you can [TS]

  use this drag and drop stuff but if you [TS]

  are an uber geek and you'd like things [TS]

  like writing code in HTML and [TS]

  controlling JavaScript and you want to [TS]

  control every single aspect of the site [TS]

  you can do that with their templates you [TS]

  just switch to developer mode and you [TS]

  can use SFTP or you can use get to [TS]

  publish your site it's super super cool [TS]

  and you can go to developers not [TS]

  squarespace com to learn more about that [TS]

  all of their templates at least until [TS]

  you bust a mob start as completely [TS]

  responsive things they were structure [TS]

  automatically when you look at them on [TS]

  an iOS device what this means is you're [TS]

  not gonna have to deal with serving up a [TS]

  crummy mobile version of your site to [TS]

  someone you send so you build it once [TS]

  and it looks good and everywhere or just [TS]

  use their templates are already built to [TS]

  do that so here's the way this works [TS]

  this is not free they give you 24 7 [TS]

  customer support you got to pay for that [TS]

  it's not something that just happens [TS]

  they don't have a staff of people there [TS]

  that you know for nothing but it's 10 [TS]

  bucks a month actually it's less than [TS]

  that if you use the coupon code which [TS]

  I'll tell you about in a second you want [TS]

  the unlimited plan its twenty bucks a [TS]

  month you send up for a year you get [TS]

  twenty percent off you sign up for two [TS]

  years to get twenty-five percent off [TS]

  some people just might be month month [TS]

  that's fine too you don't have to even [TS]

  give them a credit card to try it out [TS]

  and that's what I would like for you to [TS]

  do trying it out will support the show [TS]

  go to squarespace com / 5 by 5 you don't [TS]

  have to give them anything you just come [TS]

  up with a little name for your site and [TS]

  get going the code that you want to use [TS]

  is Dan sent me 11 well a couple people [TS]

  were saying oh dance at me 11 code [TS]

  wasn't working it to working now dan [TS]

  sent me 11 this is going to get you ten [TS]

  percent off whether you do the month to [TS]

  month or the year it's all cumulative so [TS]

  go check them out support this show [TS]

  squarespace com / 5 by 5 code is dance [TS]

  at me 11 more details in the show notes [TS]

  all right Apple switching to arm meaning [TS]

  Apple stops putting Intel CPUs in their [TS]

  macs and replaces would this be with [TS]

  this is the arm CP would this be a [TS]

  bigger move John then switch to Intel [TS]

  back in what was it Oh 306 yes it would [TS]

  be no I explain why so what are the [TS]

  challenges facing Apple if they decide [TS]

  to do this first thing they have to do [TS]

  is they have to match the performance of [TS]

  Intel CPUs maybe you don't have to [TS]

  exceed them but you at least have to [TS]

  match it because if you have into [TS]

  cpus now and you're saying we don't want [TS]

  them anywhere we want arm CPUs well you [TS]

  don't want to make everything slower [TS]

  right and you don't have to match the [TS]

  current Intel CPUs you have to match the [TS]

  upcoming Intel CPUs like you know you [TS]

  have to look at intel's roadmap and say [TS]

  well they say in this year they're going [TS]

  all with the cpu can we make something [TS]

  that's as has that kind of performance [TS]

  Intel is really good at making CPUs [TS]

  right Apple's asics is by all accounts [TS]

  very good and nicely designed everything [TS]

  but Apple didn't have to design the [TS]

  instruction set itself it sort of [TS]

  licensed that from arm and the stuff [TS]

  that is doing at CPU it's a smaller CPU [TS]

  it looks like it was hand laid out and [TS]

  stuff like that but it is not as [TS]

  ambitious as the crazy stuff that [TS]

  entailed does routinely to make its x86 [TS]

  chips fast because x86 is still ugly and [TS]

  disgusting and they manage to make that [TS]

  go fast they've got serious skills there [TS]

  look at AMD the only other real viable [TS]

  competitor to Intel back in the old [TS]

  desktop server CPU wars they am d [TS]

  invented the x86 64 instructions that [TS]

  ball Intel was off screwing things up [TS]

  with the Itanium AMD has lots of you [TS]

  know has some skilled there for a while [TS]

  AMD was faster and better than intel [TS]

  back when Intel was in the net burst [TS]

  pentium 4 days where they just [TS]

  concentrate entirely on clock speed to [TS]

  the detriment of everything else that [TS]

  turned out to not be a good idea so AMD [TS]

  was ahead for a while uh but these days [TS]

  intel has come roaring back and got [TS]

  their house in order and they've just [TS]

  been trouncing AMD in terms of [TS]

  performance for many years now uh and if [TS]

  you look at like the Geekbench scores of [TS]

  like you know people say what about [TS]

  we'll just use the a6 in a macbook air [TS]

  like I'm picking geekbench just because [TS]

  it's a site that I could find you know [TS]

  there's very difficult to benchmark CPUs [TS]

  and say what makes one fast or whatever [TS]

  but no matter how bad and synthetic [TS]

  these benchmarks are at a certain point [TS]

  it makes a difference so like the the [TS]

  iphone 5 from the geekbench score is it [TS]

  was like 1600 or something like that [TS]

  under there their scores confuse the [TS]

  ipad 3 someone said an ipad 3 with an a6 [TS]

  but the ipad 3 doesn't have an asic so [TS]

  that's also computing but anyway the [TS]

  ipad the fastest ipad score i could find [TS]

  was like 5,000 the macbook air scores [TS]

  that I could find we're like 8,000 right [TS]

  and then the Mac Pro scores for like [TS]

  thirty six thousand so [TS]

  the mac pro at the top of 36,000 and the [TS]

  iphone 5 with an a6 at 1000 that's like [TS]

  an order of magnitude difference they're [TS]

  in range between what what these little [TS]

  arm CPUs can do no one's saying you're [TS]

  going to put an a6 into a mac but it [TS]

  just goes to show that Apple has yet to [TS]

  build an arm CPU that is performance [TS]

  competitive with it so it's not like [TS]

  they have something ready to go even [TS]

  like the a-15 which is a the reference [TS]

  design from RM CPUs or whatever there's [TS]

  a long gap between anything any actual [TS]

  shipping arm CPU and the fastest [TS]

  shipping Intel CPUs so somehow Apple [TS]

  would be it's their responsibility to [TS]

  now bridge that gap because again you do [TS]

  not want to ship the next generation of [TS]

  MacBook Airs that are slower than the [TS]

  current generation so you have to match [TS]

  them that's hard to do right second and [TS]

  this is the real killer you want to [TS]

  match their power consumption because [TS]

  the whole reason you're you know [TS]

  supposedly going to arm is like oh we [TS]

  use arm CPUs and our phones and they sit [TS]

  battery and then we keep using these big [TS]

  hot power-hungry Intel CPUs and our [TS]

  MacBook Airs uh well you can't just take [TS]

  the current crop of arm CPUs and stick [TS]

  them to a Mac because they're way too [TS]

  slow and if you can make them as fast [TS]

  now you have to make them as fast and [TS]

  make them you know and not make them [TS]

  consume as much power as Intel's [TS]

  upcoming line of stuff uh and one of the [TS]

  biggest factors not the biggest factor [TS]

  and how much power your CPU takes is the [TS]

  process size and as previously mentioned [TS]

  Intel is an entire generation ahead on [TS]

  process while you are making 32 [TS]

  nanometer chips Intel's shipping 22 [TS]

  nanometer chips and that's a huge [TS]

  advantage when it comes to to power [TS]

  right so for desktops and laptops is it [TS]

  possible to match or exceed Intel's [TS]

  performance and its performance per watt [TS]

  at a bigger process node like can you [TS]

  make can you make a cpu at 30 nanometers [TS]

  that has better performance and better [TS]

  performance per watt than intel when [TS]

  they get to use 22 nanometers probably [TS]

  not that's probably pretty much a [TS]

  physical impossibility no matter how [TS]

  awesome you are and how in company intel [TS]

  is an intel is not incompetent right and [TS]

  and even if you could what would your [TS]

  advantage be say okay i think that at 32 [TS]

  nanometers i can make a chip that it's [TS]

  better price performance than intel even [TS]

  though they get to make all the chips of [TS]

  22 nanometers right ah say you could do [TS]

  that can you do it [TS]

  lower costs you know because they can [TS]

  fit way more chips onto a wafer and [TS]

  they've been doing it for longer and by [TS]

  the way you have to find someone to [TS]

  manufacture this chip for you and with [TS]

  the articles mentions I could use a TSMC [TS]

  taiwan semiconductor manufacturing [TS]

  company like you know dedicated [TS]

  semiconductor foundry you give them a [TS]

  design they manufacture it they're not [TS]

  in the process they're not in the [TS]

  business of making ships you just say [TS]

  okay you've got a design fab it for me [TS]

  make these chips again they're they're [TS]

  consistently a generation behind what [TS]

  Intel can produce an Intel does not [TS]

  really rent out its boundaries to its [TS]

  competitors oh yeah until it's not going [TS]

  to say hey you want to make a chip [TS]

  Sierra will make your arm chips for you [TS]

  20 nanometers just give us a design and [TS]

  we'll turn them out no that's not really [TS]

  what they're interested in for obvious [TS]

  reasons right when Apple switched from [TS]

  PowerPC to Intel i wrote an article [TS]

  about this i linked in the show notes in [TS]

  general it was a relief because finally [TS]

  apple fans and Apple itself could stop [TS]

  worrying about oh god Intel's got some [TS]

  new chips coming out and they're really [TS]

  awesome and their generation ahead and [TS]

  process node and they're there really [TS]

  fast and we need power pc chips that can [TS]

  compete with those and we have to go beg [TS]

  IBM to make them for us or big motor [TS]

  role to do something that's just finally [TS]

  it was really fly right we're off that [TS]

  bandwagon whatever Intel comes out with [TS]

  we get to reap those benefits now [TS]

  suddenly Apple users for the past I [TS]

  don't know how many years it's been for [TS]

  a long time now we've not had to worry [TS]

  about sea views we've not had to worry [TS]

  about Max or slow because they use [TS]

  crappy CPUs they use the same ones that [TS]

  everyone else which also happened to be [TS]

  the best ones in terms of price and [TS]

  performance which are from Intel going [TS]

  back to arm would mean that you are back [TS]

  on the hook of like oh god oh god [TS]

  Intel's coming out with new chips we [TS]

  need to we need to make something that's [TS]

  as good as these new Intel chips and [TS]

  Apple hurry up and designed that arm [TS]

  chip that's gonna you know compete with [TS]

  Intel chips and that horse race will be [TS]

  back which would be exciting but also [TS]

  depressing if Intel starts kicking our [TS]

  butts uh one obvious question here is [TS]

  like okay so if Intel's always ahead and [TS]

  process note and they're making 22 [TS]

  nanometer chips why don't all the [TS]

  iphones and ipads use intel chips ah [TS]

  that's a good question for the senior [TS]

  leadership at intel because part of the [TS]

  problem has been that thus far intel has [TS]

  not well first of all when you get down [TS]

  to small sizes suddenly that x86 burden [TS]

  becomes more relevant again because when [TS]

  you're making a 1.4 billion transistor [TS]

  chip yeah the [TS]

  number of transistors you need to deal [TS]

  with x86 crap is a small percentage but [TS]

  if you make a much much much much [TS]

  smaller chip with far fewer transistors [TS]

  like the a6 suddenly the excess stuff [TS]

  there starts to count you're like oh [TS]

  geez I would really like if I didn't [TS]

  have to deal with the x86 stuff and the [TS]

  arm chips don't have to deal with their [TS]

  very simple very straightforward risk [TS]

  machines that's why a 32 nanometer a6 [TS]

  chip can even compete in the mobile [TS]

  space because the x86 burden is relevant [TS]

  they're like the best of both worlds [TS]

  would be take apples a6 design and have [TS]

  Intel phablet at 22 nanometers but that [TS]

  doesn't exist because until it's not [TS]

  going to do that for you and Intel is [TS]

  trying to make wimpier and wimpy or CPUs [TS]

  it's got its Adam line of processors and [TS]

  they just keep going you know Intel's [TS]

  going down down down scale and as it [TS]

  goes down scale smaller chips fewer [TS]

  transistors the x86 stuff starts to [TS]

  become relevant again which must stick [TS]

  and Intel's crawl but what else can you [TS]

  do because that's what they've got until [TS]

  used to have holdings in arm and that [TS]

  divested itself from arm and dedicated [TS]

  itself to x86 and as you know that's all [TS]

  they show to about whether they're doing [TS]

  the right move over there but the bottom [TS]

  line is that you've got arm way down at [TS]

  the low end with a very simple low power [TS]

  chip design but a generation behind in [TS]

  manufacturing not that our manufacturers [TS]

  stuff but like whoever they're fabbing [TS]

  them it's not Intel I mean you and Intel [TS]

  trying to bring it's it's incredibly [TS]

  successful desktop products down market [TS]

  and mostly failing to do so and that's [TS]

  that's the pensioner that's the gap we [TS]

  have right now so the obvious [TS]

  explanation that I see it for this [TS]

  floated rumor about Apple going to Intel [TS]

  and this is you know what everyone [TS]

  thinks as soon as they see it it's like [TS]

  apples not ditching Intel there's [TS]

  nothing viable they can move to this is [TS]

  just a negotiating ploy so that Apple [TS]

  can get good prices out of Intel for its [TS]

  next generation of chips or so that [TS]

  Apple can influence Intel and say until [TS]

  we really want you to do XY and Z and [TS]

  make him tell us into them because you [TS]

  know they all ride we heard you might go [TS]

  to arm don't do that let's kiss and make [TS]

  up right ah that's kind of like when [TS]

  Apple was talking to Intel for years and [TS]

  years when he was shipping powerpc macs [TS]

  part of us like oh they're just talking [TS]

  to Intel because they're trying to force [TS]

  IBM or motorola or whoever was at that [TS]

  time to make them the chip that they [TS]

  wanted they're not actually going to go [TS]

  to Intel but they have to keep talking [TS]

  with Intel because otherwise you know [TS]

  IBM won't make the g5 for Apple right [TS]

  it's all a negotiating ploy [TS]

  but the thing is eventually that [TS]

  negotiating ploy apparently stopped [TS]

  working uh and I've actually did go to [TS]

  Intel and so that that type of [TS]

  phenomenon over the course of many many [TS]

  years could happen here where these arm [TS]

  you know bringing the next arm rumor [TS]

  keeps getting floated and floated and [TS]

  they keep using it is negotiating [TS]

  leverage at a certain point until calls [TS]

  they're bluffing says you know what [TS]

  we've got to take care of our own [TS]

  business oh you know we've got to do [TS]

  something and yeah the chips were going [TS]

  to make are perfectly suited to you but [TS]

  you're just gonna have to deal with it [TS]

  maybe until it does leave them and go to [TS]

  arm and starts up that race again uh I [TS]

  don't know final couple closing thoughts [TS]

  here yes Intel's market cap is a as of [TS]

  the time of a these show notes 1.4 [TS]

  billion Apple has more cash than that [TS]

  like in its bank account at some of its [TS]

  long-term you know not not immediately [TS]

  liquid or whatever but could Apple by [TS]

  Intel probably ever really wanted to if [TS]

  you wanted to spend all it's like [TS]

  spending you know you get you you save [TS]

  money for years and years and years and [TS]

  you spend it all in one spot yeah Apple [TS]

  could probably try to buy Intel [TS]

  antitrust would be a problem there or [TS]

  whatever but I don't know of a parolee [TS]

  wants to be in that business apple [TS]

  doesn't buy like the people who make its [TS]

  products it you know contracts it out [TS]

  and maybe buys hardware for them and [TS]

  funds the development of their thing but [TS]

  they don't want to be in the business [TS]

  they want someone else to deal with it [TS]

  the ugly low-margin part of the business [TS]

  they just want the high Martin so I [TS]

  don't think apples going to buy Intel ah [TS]

  apple's market cap is currently five [TS]

  hundred billion and rapidly dropping uh [TS]

  because of their stock being slam [TS]

  recently but there's this I mean I'm [TS]

  yeah I know you're not a stock analysis [TS]

  analyst then you won't do an analysis on [TS]

  the show but uh is this a good time to [TS]

  buy Apple stops I don't know I'm not [TS]

  gonna you won't even go into it you [TS]

  won't do it I I don't know if it's a [TS]

  good time the stock market I don't [TS]

  underst no one understands the stock [TS]

  market I don't know sorry certainly it's [TS]

  better to buy now than it was the by [TS]

  Mary was 700 dollars but I don't know I [TS]

  don't know I can I don't know why the [TS]

  stock is going down I don't know why it [TS]

  keeps going down so if you don't know [TS]

  that's probably a bad time to buy [TS]

  anything if you don't understand what's [TS]

  happened I don't understand what's [TS]

  happening so I give you know advice [TS]

  about whether you should buy it or not [TS]

  but so I'll take that as a strong [TS]

  recommendation to buy [TS]

  yeah but anyway that's okay that's the [TS]

  situation Apple is in I don't think it [TS]

  can feasibly go to arm CPUs in its max [TS]

  unless it dedicates a huge amount of [TS]

  money and time over the course of the [TS]

  next several years to coming up with [TS]

  something that can compete with intel [TS]

  intel has its own problems because many [TS]

  many more mobile cpus are being bought [TS]

  then desktop CPUs and that ratio keeps [TS]

  getting worse and worse for them so [TS]

  intel has to get into the mobile space [TS]

  somehow it seems to me though that these [TS]

  two forces should lead to an arrangement [TS]

  whereby Apple and Intel's interests [TS]

  become aligned at some point until once [TS]

  in the mobile space Apple wants [TS]

  processors a certain way Apple talks to [TS]

  Intel eventually they come to some sort [TS]

  of agreement and this arm stuff is just [TS]

  noise right but on the other hand like [TS]

  you know Bob Mansfield coming back and [TS]

  have big plans for the semiconductor [TS]

  division or whatever it could just be [TS]

  good business just like it was good [TS]

  business to constantly be talking with [TS]

  Intel and constantly be having that [TS]

  version of you know Mac OS 10 for Intel [TS]

  in the labs or whatever it's probably [TS]

  good business for Apple to at least [TS]

  investigate what would it take to get an [TS]

  arm CPU design they'll be competitive [TS]

  with Intel and how many years would it [TS]

  take how much money would it take and [TS]

  let's let's have that project and that's [TS]

  also consider how much would it take if [TS]

  we just paid Intel to make our chips for [TS]

  us and can we convince in tell toofab [TS]

  our stuff at 22 nanometers can Apple [TS]

  itself invest enough money to suddenly [TS]

  match Intel's ability to fab chips you [TS]

  know and their expertise probably not [TS]

  like many options are on the table but i [TS]

  think this this entire story at this [TS]

  point is like this is something to watch [TS]

  in the same way that you should have [TS]

  been watching those silly rumors about [TS]

  Apple talking to Intel or talking to AMD [TS]

  for years and years and years before [TS]

  they actually did the switch it's like [TS]

  it's easy to dismiss that stuff but i [TS]

  think it's it's good idea just keep an [TS]

  eye on that because Apple has many [TS]

  options here and none of them are really [TS]

  that good and same with Intel they have [TS]

  many options and not really that good [TS]

  either but I think what I think is that [TS]

  eventually if Apple and Intel keep going [TS]

  along the current path their interest [TS]

  will be sufficiently aligned such that [TS]

  they will become be able to come to some [TS]

  mutually beneficial agreement that will [TS]

  not involve Apple 100% jilting them and [TS]

  making its own arm processors ah because [TS]

  I think that is a really really tall [TS]

  order and the biggest thing against it [TS]

  is because they [TS]

  tells still ahead of everyone by a [TS]

  generation and processed note if that [TS]

  changes if suddenly everyone else [TS]

  catches up the Intel and Apple can get [TS]

  its chips fabbed at the same size as [TS]

  Intel can then that's a whole new [TS]

  ballgame but like I said for years and [TS]

  years intel has been ahead and it seems [TS]

  to be a sustainable advantage on their [TS]

  on their part and they invest tremendous [TS]

  amount of money billions billions of [TS]

  dollars into maintaining that advantage [TS]

  very few other companies have the kind [TS]

  of money to invest in that or the [TS]

  expertise or the people Apple probably [TS]

  has the money but probably doesn't have [TS]

  the expertise and I don't think they can [TS]

  hire away all of them tells people so [TS]

  I'll let us say keep watching this space [TS]

  but for now don't worry next year's crop [TS]

  of max will not have ARM processors in [TS]

  them or at least all the moon anyway [TS]

  maybe one of them will who knows but [TS]

  yeah don't hold your breath all right [TS]

  even if you say so I say so pretty good [TS]

  argument it's not much of an argument [TS]

  there little background and then an area [TS]

  to watch do you expect on arm max next [TS]

  year not next year ner max a single [TS]

  pilot program arm Mac and experimental [TS]

  one just one nothing behind the scenes [TS]

  behind closed doors I'm sure but you [TS]

  know here if you think about what [TS]

  everything that Apple has done as far as [TS]

  wanting to control their own destiny the [TS]

  move to Intel is in large part so that [TS]

  they could have more control that was [TS]

  the whole issue that they had with power [TS]

  pc chips is they could never get [TS]

  anything fast they switched until they [TS]

  get the day get the fast stuff right and [TS]

  not so much because they controlled [TS]

  Intel but just because Intel's interest [TS]

  was to make the best jede falsche stuff [TS]

  yeah right and so now they were never [TS]

  second leg Intel Intel will you please [TS]

  make us good desktop making good desktop [TS]

  processors right we can't let the big [TS]

  them right but now it's getting to the [TS]

  point where Apple's like Intel please [TS]

  like I know you want to make these [TS]

  processors but we need very very [TS]

  specific things we want to retina [TS]

  display can you build in the GPU can [TS]

  drive a Retina display all this noise [TS]

  i'm on the intel HD graphics 3000 just [TS]

  barely or 4000 we're just barely can [TS]

  drive a Retina display but we totally [TS]

  wanted to [TS]

  reten everywhere can you please make [TS]

  your in ship GPU better and was like I [TS]

  know that's not really all not a [TS]

  priority for us right now yeah so [TS]

  suddenly Apple's having to beg Intel a [TS]

  little bit but yeah going away from [TS]

  power pc is like finally we don't the [TS]

  beggining one for a CBO they they still [TS]

  by the way i will use the opportunity to [TS]

  get as much leverage as I possibly could [TS]

  an Intel like you know until is gonna [TS]

  make chips just for us and we have a [TS]

  site member they made that crazy shrink [TS]

  version for the original macbook air [TS]

  like they got Intel to make like a [TS]

  custom chip just for them it's like a [TS]

  jam a cpu I think it was the original [TS]

  macbook air like Apple has been using [TS]

  its leverage like you know at first it [TS]

  was like don't you want us we're like a [TS]

  trophy brand you can say your stuff is [TS]

  in Apple CPUs and then it was like oh we [TS]

  sell a lot of intel cpus and your other [TS]

  you know and now it's like apples like [TS]

  you know how you have any idea how many [TS]

  mobile cpus we sell if you want any part [TS]

  of that business ever you better listen [TS]

  to us I know you're not getting any of [TS]

  it now but someday you know talk to us [TS]

  maybe you can convince us to change all [TS]

  of our iphones and ipads and iOS devices [TS]

  to x86 if you've ever maked these an x86 [TS]

  chip you know like the negotiations are [TS]

  getting more strained between those two [TS]

  companies but i have to think that [TS]

  eventually they will their heads will [TS]

  come together and we'll get something [TS]

  good out of this well let's say you know [TS]

  what I mean I think Apple like you said [TS]

  Apple would be thrilled if Intel would [TS]

  just do that kind of thing but then each [TS]

  if you think about it each of the things [TS]

  that Apple relies on for from other [TS]

  vendors that it doesn't control that it [TS]

  feels that it's like there's this rumor [TS]

  that just came out that the new imacs [TS]

  are going to be delayed because of a [TS]

  welding issue this is something that [TS]

  Apple pretty much can can control so [TS]

  that's their fault but if they can't get [TS]

  enough screens you know that that that's [TS]

  got a bug now we can't get enough [TS]

  screens if they can't get a cpu that's [TS]

  gonna do what they want if they can't [TS]

  offload graphics to the cpu the way that [TS]

  they want these are things that you [TS]

  would think apple's gonna say well we've [TS]

  got some money let's let's just do this [TS]

  ourselves I don't like that they want to [TS]

  mean Samsung as we haven't you mentioned [TS]

  like they don't want to pay Samsung now [TS]

  to make all the ASIC cpus and they're [TS]

  like that's why they've been pulling [TS]

  this stuff in-house like we don't know [TS]

  I'm Jack may be five we want to make a [TS]

  custom design ship and we want to be [TS]

  able to pay anyone to fab [TS]

  think the ax is not just being fat by [TS]

  sams I may be wrong about this but yeah [TS]

  but paying Samsung to make your cpu's [TS]

  that's bad Samsung is the only other [TS]

  person making money in the smartphone [TS]

  market you don't want to be paying them [TS]

  any money so yeah I can imagine Apple [TS]

  wanting to go elsewhere for that [TS]

  expertise but like they can't just they [TS]

  can't just leave Samsung immediately [TS]

  there's a certain you know can you get [TS]

  what Samsung gives you from someone else [TS]

  at the same price and if the answer is [TS]

  no you got to grit your teeth and keep [TS]

  going to Samsung all the while Sam we [TS]

  got to have a plan to get away from [TS]

  Samsung like wooden tap beloved oocyte [TS]

  pleat like if Apple can get in tell [TS]

  toofab the a7 in its best process node [TS]

  at that time that would be a humongous [TS]

  victory but it is not an Intel strategic [TS]

  interest to become like a fab four other [TS]

  people ship designs right so the you [TS]

  know most in need to come together on [TS]

  something we want we want your [TS]

  technology we don't want your chips then [TS]

  tells like we want to sell our chips [TS]

  because becoming a fab is a low-margin [TS]

  business and we get much better margins [TS]

  by making our own you know so I don't [TS]

  know this I would love to beat see these [TS]

  negotiations because they're happening [TS]

  now like what is the a7 gonna be who's [TS]

  gonna make it and where is it going to [TS]

  be shipped like maybe that maybe it's [TS]

  too late for those and those discussions [TS]

  have already taken place right but for [TS]

  the one after that for the a8 and for [TS]

  any possible like arm and Intel arm and [TS]

  Max type negotiations those have to be [TS]

  happening now like we need a plan we [TS]

  need to talk to people we know what we [TS]

  don't want to do we don't want to give [TS]

  Samsung any more money and we also don't [TS]

  want to use any of the chips that [TS]

  Intel's currently offering us because [TS]

  they're way too power hungry we want ra8 [TS]

  design manufactured in the best [TS]

  manufacturing process in the entire [TS]

  world uh at a really cheap price and [TS]

  Intel doesn't want that they want [TS]

  something different so yeah those they [TS]

  need to come together on something but I [TS]

  you know the only the only way i can [TS]

  imagine apple going full split off is if [TS]

  they have like a concrete plan where [TS]

  they really believe that they can make a [TS]

  line of max that is not embarrassingly [TS]

  slower or crappier than what they would [TS]

  have been if they had intel cpus and [TS]

  intel is going to do everything in its [TS]

  power to make sure that apple doesn't [TS]

  never even thinks that's possible like [TS]

  they just need to make apple like so [TS]

  despair for their ability to match intel [TS]

  cpu so they won't even attempt it [TS]

  because I until will show them their [TS]

  roadmap and say you're never going to [TS]

  have [TS]

  it's gonna compete with this look at [TS]

  this do you this is art this is our road [TS]

  map that's what they did to get apples [TS]

  business they said yeah current line of [TS]

  CPUs suck but look at this upcoming lot [TS]

  of core processors they are awesome they [TS]

  are really fast they're low power and [TS]

  that was all true and Intel delivered on [TS]

  that commitment like they you know back [TS]

  then was like why doesn't happen with [TS]

  AMD don't they make better CPUs and [TS]

  stuff but Intel shows in their roadmap [TS]

  Apple said fine you're going to make [TS]

  that then go make that and Intel did go [TS]

  make that and until makes amazing CPUs [TS]

  that powered you know you know decades [TS]

  worth of great intel macs or ever many [TS]

  years has been but now we're coming to [TS]

  that point again and so intel does have [TS]

  credibility here they come to apple and [TS]

  say look yeah don't even think about [TS]

  trying to make your arm cpus can you [TS]

  compete with this you can't match us [TS]

  just you know we need to come to an [TS]

  agreement here don't even try you're [TS]

  gonna you're gonna spend billions and [TS]

  billions of dollars and it's not gonna [TS]

  work out for you just look out what [TS]

  happened to AMD we crush them don't even [TS]

  attempt it but we're friends let's be [TS]

  friends business it's like war but with [TS]

  money keep your friends close enemies [TS]

  closer yes and your frenemies i don't [TS]

  know where you keep them but like I [TS]

  guess somewhere in the middle all right [TS]

  even wrap this up think we're done all [TS]

  right so you can go to five by five TV [TS]

  such hypercritical sash 94 and you will [TS]

  get they will get all of the notes and [TS]

  links that jon is put there you can [TS]

  learn about sisk and risk on other [TS]

  things you can follow John on Twitter he [TS]

  is siracusa s I RAC USA USA and I'm damn [TS]

  Benjamin on Twitter he is also on Alpha [TS]

  app.net at Siracusa I'm Dan over there [TS]

  and I guess I said you don't wanted me [TS]

  to talk about the other one cuz you're [TS]

  not using I used tent that is 10 thought [TS]

  I oh whatever I do all right so he's [TS]

  Syracuse over there too we appreciate [TS]

  you listening if you would like you can [TS]

  rate the show in iTunes you can [TS]

  visit the sponsors it's a wonderful way [TS]

  to help support the show in the hood [TS]

  network we thank you very much for [TS]

  listening and John will see next week [TS]

  about [TS]

  [Music] [TS]