53: The Mass-to-Volume Ratio

00:00SPEAKER- Hey Brad. - Yeah.

00:01SPEAKER- I got Windows 2004 last night.

00:04SPEAKER- You're in the future, you're living in 2004 now.

00:06SPEAKER- I haven't installed it yet,

00:07SPEAKERbut I have a full command line interface.

00:09SPEAKER- It's like the late night man used to say,

00:11SPEAKER♪ In the year 2004 ♪

00:13SPEAKER(laughing)

00:15SPEAKER- So, I have no idea what that is actually.

00:18SPEAKER- Wait, what?

00:18SPEAKER- No, I've never heard that one before.

00:19SPEAKER- That's the Andy Richter,

00:21SPEAKERthat's the previous iteration of Conan O'Brien,

00:24SPEAKERor was that two iterations ago?

00:25SPEAKERThe first-- - The one on NBC?

00:27SPEAKER- The first iteration of Conan O'Brien,

00:28SPEAKERwhich was I believe late night.

00:30SPEAKER- Late night, with Conan O'Brien.

00:32SPEAKER- Late night with Conan O'Brien,

00:33SPEAKERhis first real late night show.

00:34SPEAKERThat was, do you remember that?

00:37SPEAKERI don't remember the in the year 2000 bit.

00:40SPEAKER- Brad, I'm gonna go ahead and tell you something bad.

00:41SPEAKER♪ In the year 2000 ♪

00:43SPEAKER- I am super hoarse right now,

00:45SPEAKER'cause I'm slightly hungover.

00:47SPEAKER(laughing)

00:49SPEAKER- Look, that was 20 years ago, man.

00:51SPEAKER- Oh, a little more than that actually,

00:54SPEAKERlike 22, 23, 24 years ago, 25.

00:57SPEAKERThat was when I was in high school,

00:58SPEAKERso that was like 20, maybe 25 years ago.

01:01SPEAKER- That might've been the time period I wasn't watching,

01:02SPEAKER'cause if when you were in high school,

01:03SPEAKERI would have been in college

01:04SPEAKERand I was just, I was doing other things.

01:06SPEAKER- Yeah, what is it about college

01:08SPEAKERthat makes you not watch TV?

01:10SPEAKER- I didn't have a TV for a while.

01:11SPEAKER- Okay, that was me too, actually.

01:13SPEAKER- Yeah, and then I got a TV

01:15SPEAKERand it was mostly used to play NHL 96, it turns out.

01:18SPEAKER- Okay, yeah, it's a good collegiate activity.

01:20SPEAKER- Yeah, well, like it's the one time,

01:23SPEAKERat that time in your life, in my life at least,

01:25SPEAKERbeing able to play multiplayer games was a brand new thing.

01:28SPEAKERLike having constant access to multiplayer games was new.

01:31SPEAKER- Having constant access to people around to play them

01:34SPEAKERwas also, you didn't have to like drive to somebody's house,

01:37SPEAKERthey were just around.

01:38SPEAKER- Yeah, you could just be like,

01:39SPEAKER"Hey man, I got the 4Tap, you wanna play some NHL?"

01:42SPEAKERAnd you'd be like, "Yeah, I'll play some NHL."

01:44SPEAKER- Yeah, that's what living in the future is.

01:48SPEAKER- So I gotta say the first thing that Windows 2004 did

01:51SPEAKERis I have this task scheduler thing set up

01:53SPEAKERto run a Python script that makes my audio controller

01:57SPEAKERwork with my interface, right?

02:00SPEAKERAnd the first thing Windows 2004 did was break that.

02:03SPEAKER- Nice, great, yep, that's the dark secret

02:07SPEAKERof living the streamlined automation life in Windows

02:10SPEAKERis that it's just a giant house of cards.

02:12SPEAKER- Yeah, so task scheduler seems like it maybe

02:14SPEAKERdoesn't like Python scripts now or something.

02:16SPEAKERI don't think I can just like load the WSL stuff

02:19SPEAKERand set up a crontab and maybe never think about this again.

02:22SPEAKER- You spend hours and hours setting up task scheduler

02:25SPEAKERand group policy and shortcuts of all varieties.

02:31SPEAKER- Yeah.

02:31SPEAKER- And along comes a Windows update and just goes,

02:37SPEAKERit all collapses.

02:39SPEAKER- I have not spent as much time engaging

02:41SPEAKERwith task scheduler as you have.

02:43SPEAKER- Sucks, it's terrible.

02:44SPEAKER- It's bad, yeah, I was gonna say that.

02:47SPEAKERThe secret of it is it sucks ass.

02:48SPEAKER- It doesn't work.

02:50SPEAKER- No.

02:50SPEAKER- It doesn't work reliably, it doesn't schedule tasks,

02:53SPEAKERwhich is right there in the name.

02:55SPEAKER- Oh, it schedules them.

02:56SPEAKER- It does, but not reliably.

02:57SPEAKERIt's not always clear what user is running them.

02:59SPEAKERSo sometimes the thing you wanna do,

03:01SPEAKERyou don't actually have like privileges to do.

03:03SPEAKER- Well, and also sometimes like the one job

03:08SPEAKERthat task scheduler has is to look at some conditions

03:10SPEAKERand when those conditions are fulfilled to do a thing.

03:13SPEAKERSometimes it looks at the conditions,

03:15SPEAKERthe conditions are fulfilled.

03:16SPEAKER- And it just doesn't.

03:17SPEAKER- And it just doesn't do anything.

03:17SPEAKER- Yeah, totally, well, that's what I'm getting at sometimes

03:19SPEAKERand it doesn't tell you why, that's the thing.

03:21SPEAKER- No.

03:22SPEAKER- It might have tried to do the thing

03:23SPEAKERand was told by the system,

03:25SPEAKERyou don't have the rights to do this,

03:26SPEAKERbut it never tells you that.

03:27SPEAKER- But I can't like, I need to find like

03:30SPEAKERthe command line interface for task scheduler

03:32SPEAKERso I can run a task to see what's happening at the console.

03:35SPEAKER- It has that, we can talk later.

03:37SPEAKEROr I can just tell you right now what it is actually, hold up.

03:40SPEAKER- Yeah, tell me right now.

03:41SPEAKER- I've got a whole bunch of them set up right now.

03:43SPEAKERI've got, do you wanna know what,

03:45SPEAKERthis is gonna turn into the whole episode

03:46SPEAKERif we keep going like this.

03:47SPEAKER- Do I wanna know what love is?

03:48SPEAKERYes, show me, give me a moment.

03:50SPEAKER- Do you want me to tell you what command line stuff

03:53SPEAKERI have set up that I can just type a word in to make happen

03:56SPEAKERrather than navigating?

03:57SPEAKER- This feels strangely kinky, but okay.

04:01SPEAKER- For example, 'cause you know,

04:03SPEAKERlike most of the stuff in Windows,

04:04SPEAKERyou have to navigate like five layers of menus, right?

04:06SPEAKERTo find the button you want to click.

04:08SPEAKER- Yeah.

04:09SPEAKER- So like I have automated things like switching

04:10SPEAKERbetween the light and dark theme in Windows.

04:13SPEAKERSo.

04:14SPEAKER- Why would you ever switch

04:15SPEAKERbetween the light and dark theme in Windows?

04:16SPEAKER- 'Cause I like the light theme during the day

04:17SPEAKERand the dark theme at night.

04:19SPEAKER- Doesn't it do that automatically now?

04:20SPEAKER- No, it does not.

04:21SPEAKER- What?

04:22SPEAKER- We've talked about this, man.

04:24SPEAKERSo.

04:25SPEAKER- I thought it did that automatically now.

04:26SPEAKER- I have shortcuts called light and dark.

04:30SPEAKERAnd I can pop up the start menu and type light

04:33SPEAKERand it will automatically switch to the light theme.

04:36SPEAKERI have one called desk swap that will swap

04:39SPEAKERwhich one is my primary desktop.

04:42SPEAKER- Wow.

04:44SPEAKER- I've got a variety of them that's mute

04:47SPEAKERand unmute various input devices, sound input devices.

04:51SPEAKER- Oh yeah, that's a near soft jam right there, right?

04:54SPEAKER- Yes, no, I use task scheduler for that.

04:56SPEAKERBut hold up, let me tell you, I was just gonna tell you,

05:00SPEAKERschtasks.exe is the command line version of task scheduler.

05:07SPEAKERSo you can run schtasks/run and the name of the task.

05:13SPEAKERThat's for automating.

05:14SPEAKERThat's for your-

05:15SPEAKER- This is a fantastic, incredibly useful cold open.

05:16SPEAKER- That's for your automating purposes.

05:18SPEAKERWe'll talk more later.

05:20SPEAKER(upbeat music)

05:46SPEAKER(upbeat music)

05:51SPEAKER- Welcome to Brad and Will Made a Tech Pod.

05:53SPEAKERI'm Will.

05:54SPEAKER- I'm Brad, how you doing, Will?

05:56SPEAKER- I'm over here making a tech pod.

05:57SPEAKERHow about you, man?

05:58SPEAKER- That's right, man.

05:59SPEAKERIt's just here we are on a Saturday morning

06:00SPEAKERfeeling just lively and crisp, ready to talk.

06:03SPEAKER- Everything's great.

06:04SPEAKER- Yeah, nothing bad at all.

06:07SPEAKERWhatever, I'm sorry.

06:08SPEAKERThings are fine.

06:10SPEAKER- Look.

06:11SPEAKER- They're not fine, but they're fine.

06:13SPEAKER- Yeah, everything's fine.

06:15SPEAKER- I slept on the couch last night, not because

06:17SPEAKERof any infractions, I didn't do anything wrong.

06:19SPEAKER- Okay, you just couldn't be bothered to get up?

06:22SPEAKER- No, my daughter has been in our room

06:27SPEAKERthe last couple of nights and we are a queen size bed house.

06:30SPEAKERSo it's three people in a bed is too many people.

06:32SPEAKER- That's a lot.

06:33SPEAKER- So I, discretion was the better part of valor last night.

06:36SPEAKERI chose to just, you know, couch sleep.

06:39SPEAKER- She's enough of an actual little person now

06:41SPEAKERthat she occupies a full half of the bed.

06:43SPEAKERNot so much like a two people plus a tiny human.

06:48SPEAKERLike there's kind of.

06:49SPEAKER- If she just occupies half of the bed,

06:51SPEAKERshe's doing better than she ever has.

06:53SPEAKERShe's, you know, she's a full sprawl sleeper.

06:57SPEAKERIt's like arms, legs in all starfish positions.

07:01SPEAKER- Punching way above her weight class on the bed occupancy.

07:05SPEAKER- Her mass to volume ratio is off the charts.

07:08SPEAKER- Oh boy.

07:10SPEAKER- But yeah, so we're, we're a little,

07:14SPEAKERthings are, we're great.

07:16SPEAKERWe're doing great.

07:17SPEAKER- Fine, fine, I think it's fine.

07:18SPEAKER- What are we going to talk about today?

07:20SPEAKERWell, first off.

07:22SPEAKER- Yes.

07:23SPEAKER- I want to point out, we're still reading

07:23SPEAKERHitchhiker's Guide.

07:26SPEAKER- Yes.

07:27SPEAKER- Oh, Brad, you've got to get on this.

07:28SPEAKERIt's coming up.

07:29SPEAKER- I know, I need to get in.

07:30SPEAKERSo I read, I read both intros or I read both

07:33SPEAKERlike the forward and the introduction.

07:35SPEAKER- The Neil Gaiman one and the Douglas Adams one.

07:37SPEAKER- The forward written by Neil Gaiman,

07:38SPEAKERwhich largely restated what was written

07:40SPEAKERin the introduction by Douglas Adams, but that's okay.

07:43SPEAKER- Yes.

07:44SPEAKER- I found, I will say this.

07:47SPEAKERI found the Douglas Adams authored introduction

07:51SPEAKERwildly more charming.

07:54SPEAKER- I, look.

07:55SPEAKER- Nothing against Neil Gaiman, but like there was something,

07:57SPEAKERthere's just a certain lilt to the way that Douglas Adams,

08:01SPEAKERlike his phrasing, if that's what I'm in for, I'm excited.

08:05SPEAKERThat's what I'm trying to say.

08:06SPEAKER- We'll get to this when we talk about the book,

08:08SPEAKERbut the thing that has struck me rereading this is how much

08:12SPEAKERof modern discourse is shaped by the weird quirks,

08:15SPEAKERespecially in our friend circle,

08:17SPEAKERby the weird quirks of the Douglas Adams' use of language.

08:21SPEAKER- Okay, great.

08:23SPEAKERI'm excited about that.

08:24SPEAKERLike, there's just something about like that.

08:27SPEAKERThere's a certain idiosyncrasy of like speech patterns

08:29SPEAKERthat I really go for.

08:31SPEAKERIt's like in the intro, like it stuck with me.

08:33SPEAKERThat's how memorable it is that like,

08:36SPEAKERwhen he was talking about being drunk in a field in Austria,

08:39SPEAKERwhen he had the idea for the book.

08:40SPEAKER- Outside Innsbruck.

08:41SPEAKER- Yes, like his, the way he described his drunkenness was,

08:46SPEAKERif I remember the phrasing,

08:48SPEAKERwe are speaking of the mild inability to stand up.

08:52SPEAKERIt's like, okay, you got the point across.

08:53SPEAKERThat's very, those are very tidy and economical

08:56SPEAKERand slightly charming way to get that idea across.

08:58SPEAKER- Yeah, so it's been fun rereading those.

09:03SPEAKERThis was our challenge for hitting a thousand patrons.

09:08SPEAKERWe're at almost 1100 now.

09:10SPEAKER- Wow, geez.

09:11SPEAKER- Yeah, a lot of people signed up this month.

09:12SPEAKERA lot of people are stoked about Hitchhiker's Guide.

09:14SPEAKER- Everybody wants to read Hitchhiker's Guide.

09:16SPEAKERThank you, that's, wow, that's flattering.

09:18SPEAKERThe last thing I'll say about it real quick is I,

09:21SPEAKERit doesn't look that long.

09:24SPEAKER- It's not.

09:24SPEAKER- Maybe it's the omnibus edition that I've got,

09:26SPEAKERbut like I thought somebody had said,

09:28SPEAKERit was like your kind of standard 300 page novel,

09:30SPEAKERbut like in the omnibus, it's like 180 pages or something.

09:35SPEAKER- It's like, depending on how fast you read,

09:38SPEAKERI think Mike Kendall told me it was going to be five hours.

09:41SPEAKER- Oh, okay, yeah.

09:42SPEAKER- Yeah, at the outside.

09:44SPEAKER- Not a problem, I got plenty of time.

09:46SPEAKER- It's a good Sunday afternoon.

09:47SPEAKER- Yeah, so I'm getting ready to get into it

09:48SPEAKERand I am looking forward to it.

09:49SPEAKERI know somebody, at least one person I've seen

09:51SPEAKERon the Discord said they just finished reading it

09:53SPEAKERfor the first time and we're just in love with it, so.

09:56SPEAKER- It is, one of the things that we'll talk,

09:59SPEAKERthat I want to talk about is it is striking to me

10:01SPEAKERhow prescient it is for a thing that is written

10:04SPEAKERin the late 70s about the modern technology of today.

10:09SPEAKER- I don't know if that's a good thing or not.

10:12SPEAKER- Pros and cons.

10:12SPEAKER- Okay.

10:13SPEAKER- Anyway, today we didn't mean to make this Nvidia month,

10:17SPEAKERbut Nvidia, yeah, I got announced.

10:19SPEAKER- We went into this podcast with this topic in mind

10:22SPEAKERand it occurred to me like we probably need to say

10:23SPEAKERlike this is not sponsored by Nvidia.

10:26SPEAKER- Yeah, no, we, but--

10:28SPEAKER- We are doing this show purely

10:30SPEAKERbecause this is a very newsworthy topic.

10:32SPEAKER- Well, I mean, the new GPUs came out

10:33SPEAKERand they were also discussing.

10:35SPEAKER- Yes, it's just like Nvidia has no chill at the moment.

10:38SPEAKERThey just keep dominating the news.

10:40SPEAKER- Well, yeah, and we had the game streaming thing last week,

10:42SPEAKERwhich is not entirely an Nvidia thing,

10:44SPEAKERbut like the thing that I like best

10:47SPEAKERand I think is maybe not the easiest to use,

10:49SPEAKERbut the best results is the Nvidia solution with Moonlight.

10:53SPEAKER- Can I ask real fast not to derail this?

10:55SPEAKERHow is Moonlight?

10:56SPEAKERI don't know if we covered this.

10:59SPEAKERHow is Moonlight on the physical Steam Link,

11:01SPEAKERthe actual box that Valve sold?

11:02SPEAKERLike, is it adequate?

11:04SPEAKER- So that's where I run it.

11:05SPEAKERIt seems like the hardware in the, so the--

11:09SPEAKER- Because that hardware cannot do H.265

11:12SPEAKERis the reason that I ask, so I wonder--

11:13SPEAKER- That's it, yeah.

11:14SPEAKER- So I wonder if H.264 is up to the task.

11:16SPEAKER- It's fine for 1080p.

11:18SPEAKERI have not tried, you can't do 4K on the Steam Link.

11:21SPEAKER- Yeah, okay.

11:22SPEAKER- The setup is a little fiddly.

11:26SPEAKERLike, there's not an auto updater for Moonlight,

11:28SPEAKERand I'm not sure if that's because Moonlight is still early

11:31SPEAKERor if it's that there's some limitation of the Steam Link

11:34SPEAKERthat won't let applications update themselves.

11:36SPEAKER- But just performance latency-wise, it's fine?

11:38SPEAKER- It's great.

11:39SPEAKERI played a fair amount of Death Stranding,

11:43SPEAKERor Watch Dogs 2 last night for, yeah, using it.

11:46SPEAKER- I might actually get that Steam Link out of the box

11:49SPEAKERand put Moonlight on it, see how it is.

11:50SPEAKER- If you haven't, so I use one Steam Link that is WiFi,

11:54SPEAKERone that is ethernet, they both work fine.

11:58SPEAKERThe WiFi one is a little bit,

12:00SPEAKERoccasionally you get graphics artifacts.

12:02SPEAKERThe wired one, I never see that.

12:04SPEAKER- Are there kind of your commercial video streaming

12:08SPEAKERentertainment apps available for the Steam Link?

12:11SPEAKERProbably not.

12:12SPEAKER- Not really.

12:12SPEAKERYou can't get Netflix for it.

12:13SPEAKER- Right, okay, that's what I meant, okay.

12:14SPEAKERSo it could not be a good Apple TV replacement necessarily?

12:18SPEAKER- I think you're better off with a dedicated piece

12:20SPEAKERof hardware for that that is not Steam Link.

12:23SPEAKER- So the flip side of that is I think probably,

12:27SPEAKERI just got iOS 14 on, or TVOS 14 on my Apple TV yesterday.

12:32SPEAKERSo I haven't tried the new support for controllers,

12:35SPEAKERbut theoretically they fixed controller support

12:37SPEAKERwith this release of iOS or TVOS.

12:40SPEAKER- Wait, was it broken before?

12:42SPEAKER- Well, so it supported all the sticks and buttons

12:44SPEAKERexcept for the clicky, the sticks on the Apple TV

12:48SPEAKERand the iOS devices were the M5 standard.

12:52SPEAKERSo they would map to that,

12:54SPEAKERwhich meant there was no clicky sticks,

12:57SPEAKERwhich is fine if you're playing games built for iOS,

12:58SPEAKERbut if you're trying to play stuff that's built for the PC

13:00SPEAKERthat expects to have clicky sticks,

13:02SPEAKERyou had situations where there were weird cords

13:04SPEAKERwhere you had to press the Xbox button

13:06SPEAKERon your Xbox controller and it was not great.

13:10SPEAKER- No.

13:12SPEAKER- But today we're talking about why this ARM Nvidia deal,

13:15SPEAKERand it's an interesting, the reason we're talking about it

13:19SPEAKERis 'cause it's one of the biggest semiconductor deals

13:22SPEAKERof all time.

13:22SPEAKER- Yeah, for sure.

13:24SPEAKERNvidia has acquired ARM for $40 billion in cash and stock.

13:28SPEAKER- Yes, Nvidia is the biggest semiconductor company

13:32SPEAKERright now, they're the biggest chip designer.

13:36SPEAKERAnd it's important because the two parts of,

13:41SPEAKERso Nvidia, we all know Nvidia 'cause they make video cards

13:43SPEAKERand they made the switch CPU, GPU, SOC,

13:47SPEAKERyou know, all of this other stuff.

13:50SPEAKER- Games, mostly games kind of applications

13:52SPEAKERis where they have been known forever.

13:55SPEAKER- Right, but the games are kind of the secondary part

13:59SPEAKERof Nvidia's business these days.

14:02SPEAKER'Cause it seems like they're trying to be the hardware

14:05SPEAKERthat runs the machine learning infrastructure of our future,

14:08SPEAKERof future compute.

14:10SPEAKER- Like I kind of had an inkling of this before,

14:12SPEAKERbut like this deal has made it very obvious

14:14SPEAKERthat Nvidia is kind of leading this whole other life,

14:17SPEAKERI feel like.

14:18SPEAKERLike they kind of have a double life now, right?

14:20SPEAKEROf like everybody knows them for their gaming applications,

14:22SPEAKERbut like they're getting into data centers

14:25SPEAKERand machine learning and potentially some other stuff

14:28SPEAKERoff of this deal that has nothing to do with games.

14:31SPEAKER- Well, they've been in that for like,

14:33SPEAKERthere's a very good reason that they released Ampere first

14:36SPEAKERas a data center product, right?

14:38SPEAKERNot as a compute, a machine learning compute product,

14:41SPEAKERnot a gaming product.

14:44SPEAKER- So Arm, if you don't know what Arm is,

14:48SPEAKERArm is a chip designer design company out of the UK

14:55SPEAKERthat basically spun out of Apple and Acorn, I think.

14:59SPEAKER- Did, did it really? - Or Apple and Acorn

15:01SPEAKERstarted it.

15:02SPEAKER- I didn't know that.

15:04SPEAKER- Hold on, I want to make sure I got that right.

15:06SPEAKER- I had no idea.

15:08SPEAKERArm Holdings or Arm Limited now is the...

15:12SPEAKERYeah, so Acorn computers developed

15:14SPEAKERthe Acorn risk machine architecture in the '80s

15:18SPEAKERand then Apple puts money into them.

15:20SPEAKER- That's where Arm comes from.

15:22SPEAKER- Yeah, but now Arm stands for advanced risk machine

15:26SPEAKERrather than Acorn risk machine,

15:28SPEAKERwhich is what they originally did.

15:29SPEAKERAnyway, the point is they designed CPUs

15:32SPEAKERand then licensed that technology out to everybody.

15:34SPEAKERAnd they, unlike Intel, which say designs a CPU

15:39SPEAKERand then makes that CPU and then sells it to you.

15:42SPEAKER- And that's the only way you can get it.

15:43SPEAKER- Yeah, that's it.

15:45SPEAKERArm says, "Hey, here's all the bits that go into our CPUs."

15:48SPEAKERThese systems on a chip, SOCs.

15:52SPEAKERYou can get whatever you want.

15:54SPEAKERYou can take the pieces you want.

15:55SPEAKERYou can, like, if you want this, here's a package,

15:58SPEAKERhere's the overall box.

16:00SPEAKERAnd then you can kind of Chinese menu it out

16:02SPEAKERwhere you say, "I want one of this math coprocessor

16:06SPEAKERand six of these linear math accelerators."

16:09SPEAKERAnd you can take the bits that they give you and assemble

16:12SPEAKERthem however you want.

16:13SPEAKER- Right, and like if you're Apple, for example,

16:14SPEAKERyou can be like, "Ah, you can keep your GPU.

16:16SPEAKERWe have our own now, and we're gonna put our own GPU

16:20SPEAKERon the die with your Arm stuff."

16:22SPEAKER- Exactly, so you're licensing an instruction set

16:26SPEAKERand a chip design is my understanding with Arm.

16:29SPEAKER- Yes, I wrote, I started writing the notes.

16:31SPEAKERLike, I don't know that either of us are equipped

16:32SPEAKERto get into this, or if we even need to,

16:34SPEAKERthe whole kind of like CISC versus risk dichotomy.

16:37SPEAKER- We can talk about it a little bit.

16:38SPEAKERBut like I've been interested in what it means for 25 years,

16:43SPEAKERbut I've never a hundred percent gotten my head around it.

16:45SPEAKER- Yeah, so.

16:48SPEAKER- Or should we even like talk about what an instruction set

16:50SPEAKERis for people that don't know?

16:52SPEAKERLike it's essentially the very base level operations

16:56SPEAKERthat a CPU is carrying out inside itself, right?

16:59SPEAKER- Pretty much, and so at one level there's like the Turing,

17:04SPEAKERthe set of Turing, the things you need to be Turing complete

17:07SPEAKERin a computer, which is like ands and ors and nors.

17:10SPEAKER- You're talking Alan Turing, not the Turing GPU.

17:14SPEAKER- Yeah, I'm talking about Alan Turing,

17:15SPEAKERwho the Turing GPU is named after, yes.

17:17SPEAKER- Just to be clear.

17:19SPEAKER- So at like the low level,

17:21SPEAKERall of the math that happens in the computers

17:23SPEAKERis like four or five different kinds

17:24SPEAKERof mathematical operations.

17:25SPEAKERBut when you go up a level,

17:28SPEAKERthen you start to build specific functions on the chip

17:33SPEAKERthat are hardware accelerated

17:34SPEAKERthat makes certain types of math faster.

17:36SPEAKERSo like the difference between integer math

17:38SPEAKERand floating point math is a good example of that.

17:40SPEAKERIf you think back to the old 386, 387 days,

17:45SPEAKERIntel said, okay, these are mostly integer processors.

17:48SPEAKERAnd if you need floating point math,

17:50SPEAKERthen you can buy this extra 387 chip that sockets in.

17:55SPEAKER- A math co-processor that you can stick on your board.

17:57SPEAKER- Yeah.

17:58SPEAKER- And then eventually that logic started being included

18:00SPEAKERjust inside the CPU 'cause it made sense.

18:03SPEAKER- Well, because it turns out

18:04SPEAKERthat we're a buttload of applications

18:06SPEAKERfor massive amounts of floating point math, like games.

18:11SPEAKER- I don't know if this is illustrative or not

18:15SPEAKERfor people who aren't computer scientists,

18:16SPEAKERwhich I certainly am not.

18:17SPEAKERBut for example, you can go on Wikipedia

18:20SPEAKERand just pull up a list of every instruction

18:22SPEAKERin the x86 instruction set, for example.

18:25SPEAKER- For example.

18:25SPEAKER- And you can scroll down and it's literally just like,

18:28SPEAKERlike there's an operation called add, which adds.

18:31SPEAKERThere's like add with carry.

18:32SPEAKERThere is convert byte to word, compare operands,

18:36SPEAKERconvert word to double word, decrement by one,

18:40SPEAKERsigned multiply, you know what I mean?

18:42SPEAKERLike those are just, and to clarify for my own benefit

18:45SPEAKERas well as anybody listening, like those are fixed functions

18:49SPEAKERthat are implemented in the hardware of the chip, right?

18:51SPEAKER- When you are talking about somebody writing machine code,

18:54SPEAKERthey are writing code that calls

18:55SPEAKERthese specific instructions, right?

18:57SPEAKER- Right.

18:58SPEAKER- So, whereas if you write a C program,

19:01SPEAKERyou write a program that has some basic human readable logic

19:05SPEAKERthat isn't just a series of math instructions

19:07SPEAKERthat then becomes translated into that by a compiler.

19:11SPEAKER- Right, so yeah, so like, so everything from,

19:14SPEAKERat the end of the day, like everything from the highest level

19:16SPEAKERlike Python or JavaScript or whatever down to C,

19:19SPEAKERto any compiled language, like at the end of the day,

19:22SPEAKERall that's actually happening on the CPU

19:23SPEAKERis these instructions being executed.

19:25SPEAKER- It's interpreted down into this stuff, yes.

19:26SPEAKER- It's just that the compiler is what handles

19:27SPEAKERturning the human code into machine operations and okay.

19:33SPEAKER- So in the 90s or late 80s, I guess,

19:38SPEAKERthe CPU architectures had gotten large

19:41SPEAKERand the instruction sets were increasingly complicated.

19:45SPEAKER- Complex you might say.

19:46SPEAKER- Well, that's a retronym, you know.

19:48SPEAKER- Is it?

19:49SPEAKER- The complex was a response

19:50SPEAKERto reduced instruction set computing.

19:53SPEAKER- Oh, really?

19:54SPEAKEROkay.

19:54SPEAKER- Yeah, so these guys, I think the MIPS people

19:57SPEAKERcame up with this first,

19:58SPEAKERbut it seems like it was maybe developed

19:59SPEAKERin parallel multiple places.

20:00SPEAKER- What, like the concept of risk?

20:02SPEAKER- Yeah.

20:03SPEAKER- I think I read this morning, it started in Berkeley.

20:05SPEAKERI think it started in academia.

20:09SPEAKER- Well, anyway, but the point is,

20:11SPEAKERrather than have a large complicated set of instructions

20:14SPEAKERthat did a bunch of accelerated things,

20:16SPEAKERthey were like, why don't we just make

20:17SPEAKERall the processes run faster and keep them simpler?

20:22SPEAKERAnd I think specifically a fixed length.

20:25SPEAKERSo where complicated instructions could use multiple cycles

20:28SPEAKERper compute, these are the same size instructions

20:33SPEAKERthat run at the same speed so you can do things more.

20:36SPEAKER- Okay, okay.

20:37SPEAKER- We're getting into computer science stuff

20:38SPEAKERthat doesn't matter here.

20:39SPEAKER- Yeah, yeah, yeah.

20:39SPEAKERAnd that's pretty far out of our depth.

20:42SPEAKERSo you're saying like the instruction has to like execute

20:44SPEAKERand complete within the same cycle kind of situation?

20:46SPEAKER- Or within a specified number of cycles.

20:49SPEAKER- I see, I see.

20:50SPEAKER- The other thing that is important to know

20:52SPEAKERabout all this stuff,

20:53SPEAKERis this was like in the '90s when Mac and PC people

20:56SPEAKERwere crazier and would yell at each other

20:58SPEAKERabout which computer was better.

20:59SPEAKER- Yes.

21:00SPEAKER- You'd hear all this RISC versus CISC stuff.

21:02SPEAKERBack then, there was an appreciable difference.

21:04SPEAKER- Well, a lot of power PC,

21:06SPEAKERlike people would wear power PC t-shirts and stuff like that.

21:08SPEAKER- People wore all sorts of bad t-shirts.

21:10SPEAKERLike I had one of those color shifting t-shirts.

21:12SPEAKERWe don't need to talk about sartorial choices.

21:15SPEAKERThe point is, in the modern context,

21:18SPEAKERthere isn't really a CISC versus RISC anymore.

21:22SPEAKERThere are RISC CPUs, RISC instruction sets

21:25SPEAKERwith complex elements,

21:27SPEAKERand there are complex instruction sets

21:29SPEAKERwith the RISC-like elements.

21:31SPEAKERThey're all, everything's mishmashed together now.

21:35SPEAKER- Yes, multiple people who absolutely know

21:37SPEAKERwhat they're talking about have told me basically

21:39SPEAKERthat like, yeah, even a modern x86 CPU

21:41SPEAKERhas RISC-like elements inside it.

21:44SPEAKER- Well, I mean, the point, the big thing,

21:47SPEAKERwe should do an episode about this

21:48SPEAKER'cause it's really interesting,

21:49SPEAKERbut the Pentium 4 was a turning point for all of this stuff.

21:52SPEAKER'Cause the Pentium 4, they were like,

21:53SPEAKERwe can make these really crazy complex, super deep pipelines

21:56SPEAKERand we can predict what's gonna happen fast enough

21:58SPEAKERthat we can run it at like five, six gigahertz.

22:00SPEAKERIt's gonna be so fucking fast.

22:02SPEAKERIt's gonna blow your ears off.

22:04SPEAKERIt's gonna be the most amazing thing ever.

22:05SPEAKERAnd then it didn't work.

22:06SPEAKERAnd that was the moment that like the CISC wave broke

22:09SPEAKERand they were like, well, maybe we should simplify

22:12SPEAKERthis whole thing and make it just parallel, real parallel.

22:15SPEAKERSo anyway.

22:16SPEAKER- So, I mean, I guess back to the actual topic at hand,

22:18SPEAKERmoving back to the ARM stuff,

22:21SPEAKERmy understanding of why ARM is so big in the mobile space

22:23SPEAKERis that primarily it's because it's so power efficient,

22:26SPEAKERthe performance per watt is very good.

22:28SPEAKERIs that a direct result

22:30SPEAKERof the way that the RISC stuff processes instructions?

22:33SPEAKER- I don't know if it's a direct result

22:34SPEAKERof the way the RISC stuff processes instructions

22:37SPEAKERor a direct result of the design goals

22:40SPEAKERof the ARM processors over the years.

22:43SPEAKER- I see, I see.

22:44SPEAKER- Like you can make a really power hungry RISC chip,

22:47SPEAKERjust like you can make a really efficient CISC chip,

22:51SPEAKERyou know, like there are some 0.3 watt TDP iX86 processors.

22:58SPEAKER- Okay.

23:00SPEAKER- But just--

23:01SPEAKER- You're just not getting a lot of performance

23:02SPEAKERout of them necessarily.

23:03SPEAKER- Well, not even that, people don't use them

23:06SPEAKERfor the types of things that you use ARM processors for

23:08SPEAKERbecause the market has gone in the ARM direction.

23:10SPEAKER- Right, right, that's what I mean.

23:10SPEAKERLike that's why there aren't phones

23:12SPEAKERwith X86 mobile chips in them.

23:14SPEAKER- I mean, there are, you just don't know

23:15SPEAKERwhat even there's what they are.

23:16SPEAKERYeah. - What?

23:17SPEAKER- Yeah, you can buy X86, X86 Android phones

23:20SPEAKERor you could for a while, I don't know if you still can.

23:22SPEAKER- That's bizarre.

23:24SPEAKER- So, okay, so SoftBank bought ARM a few years ago.

23:28SPEAKERSoftBank is a investment holding company from Japan

23:33SPEAKERwho invested in such companies as WeWork, Theranos and Uber.

23:37SPEAKER- Oh no, oh no.

23:39SPEAKER- They--

23:40SPEAKER- It's a questionable portfolio.

23:42SPEAKER- Look, they've done some stuff that worked out for them,

23:44SPEAKERbut like it's not, they've not had great times

23:46SPEAKERthe last few years it seems like.

23:48SPEAKERAnd they sold their stake in ARM

23:55SPEAKERfor $40 billion in stock and cash to Nvidia

23:58SPEAKERor they want to, they've announced that they want to.

24:01SPEAKERThis is a huge deal.

24:04SPEAKERIt is interesting because they didn't talk about it

24:07SPEAKERwith investors or regulators

24:09SPEAKERbefore they announced that they wanna do this.

24:12SPEAKERAnd there's a pretty heavy regulatory burden here

24:14SPEAKER'cause the UK, the U.S., China, and maybe even Japan

24:18SPEAKERare gonna care about this.

24:19SPEAKER- Okay, and the EU, right?

24:22SPEAKER- Well, no, 'cause it's UK.

24:25SPEAKERSo probably not the EU by the time the deal

24:28SPEAKERhas to pass regulatory bounds.

24:29SPEAKER- Maybe not.

24:31SPEAKER- But yeah, it's gonna be a while before this is active.

24:37SPEAKERThis is an announcement of intent,

24:39SPEAKERnot an announcement of success.

24:43SPEAKER- It's interesting and it's important

24:46SPEAKERbecause there are literally billions of ARM chips

24:48SPEAKERshipped every year made by Apple, Qualcomm, Samsung, Nvidia,

24:54SPEAKERand a bunch of other smaller companies

24:55SPEAKERthat you probably don't know about.

24:56SPEAKER- Yeah, so should we just rattle off

24:58SPEAKERkind of the most of the typical devices

25:00SPEAKERyou would find ARM chips in?

25:01SPEAKERIt's like every, it's every iPhone on the market,

25:03SPEAKERjust about every Android phone in the market.

25:05SPEAKER- Almost every Android phone, yeah.

25:06SPEAKER- Just damn near every tablet on the market of any kind.

25:09SPEAKERThe switch.

25:10SPEAKER- Well, no, no, 'cause like Surface and stuff like that

25:12SPEAKERare all x86, like anything.

25:14SPEAKER- Yeah, but that's, I feel like they're sort of stretching

25:16SPEAKERthe definition of tablet there.

25:17SPEAKERLike that's moving, that's almost moving

25:19SPEAKERin the computer laptop-y kind of space.

25:21SPEAKER- Well, I mean, it is a laptop.

25:22SPEAKERIt's just a laptop without a keyboard attached.

25:24SPEAKER- Yeah, like the Surfaces are weird hybrid devices

25:26SPEAKERthat are, but the iPad, all the Amazon Fire stuff.

25:31SPEAKER- Oculus Quest.

25:32SPEAKER- Yes, yes, the Switch, the Nintendo Switch is ARM-based.

25:37SPEAKER- Some routers, which is shocking.

25:40SPEAKER- Oh, they're the Raspberry Pi, like all, practically,

25:42SPEAKERpractically all of the, the entire-

25:44SPEAKER- The MiSTer.

25:44SPEAKER- Well, yeah, the MiSTer has an ARM core on it

25:46SPEAKERin addition to the FPGA.

25:48SPEAKERPretty much the entire single board computer scene,

25:50SPEAKERlike the Raspberry Pi and the,

25:52SPEAKERall that stuff is ARM-based.

25:54SPEAKER- I haven't torn one apart to see or looked at a tear down,

25:57SPEAKERbut my guess is like stuff like Alexa's and Google Homes

26:00SPEAKERand all of those like Nest screen things,

26:03SPEAKERyour video doorbells, a lot of your web,

26:05SPEAKERyour like internet connected webcams.

26:09SPEAKER- Any kind of internet of home smart device

26:12SPEAKERprobably has an ARM chip in it?

26:14SPEAKER- Yeah.

26:15SPEAKER- And it's an internet of things smart device.

26:17SPEAKER- Like suffice to say, if it is not a,

26:20SPEAKERif it's not an Intel or AMD made x86 based device,

26:23SPEAKERlike the odds are it's ARM.

26:24SPEAKER- Well, no, there's also MIPS.

26:26SPEAKER- Well, yeah, so like my router,

26:28SPEAKERwell, I didn't want to get into that.

26:29SPEAKERThat's kind of an edge case.

26:30SPEAKERLike my router has a MIPS chip in it, but-

26:32SPEAKER- So a lot of networking stuff

26:33SPEAKERhas MIPS chips because they're cheaper.

26:35SPEAKER- In fact, we'll get to that.

26:35SPEAKERThat is a topic to talk about later,

26:37SPEAKERbut ARM is extraordinarily widespread is the point.

26:40SPEAKER- Also, do you know what else has ARM?

26:42SPEAKER- What's that?

26:43SPEAKER- The world's fastest supercomputer.

26:44SPEAKER- Oh, really?

26:46SPEAKER- Yeah, so supercomputers and data centers

26:49SPEAKERhave butt loads of ARM chips now.

26:51SPEAKER- Okay, so that is a relatively recent development, right?

26:54SPEAKER- The last five, 10 years probably, yeah.

26:56SPEAKER- Yeah, yeah.

26:57SPEAKER- And the reason that happens is because in a data center,

27:01SPEAKERin a supercomputer context, well, okay,

27:04SPEAKERdata, two different use cases.

27:06SPEAKERData centers care a lot about performance per watt

27:09SPEAKERand performance both in terms of heat generated

27:11SPEAKERand power consumed.

27:12SPEAKER- Yeah, I feel like you hear a lot about the cooling needs

27:15SPEAKERof a data center, right?

27:16SPEAKERLike to the extent that that sometimes helps dictate

27:20SPEAKERwhere they're built, stuff like that.

27:21SPEAKER- There's a reason that data centers get built

27:23SPEAKERon the Columbia River in Oregon,

27:25SPEAKERand it's because there's a lot of hydroelectric power

27:27SPEAKERand a lot of cool water.

27:28SPEAKER- Yeah.

27:30SPEAKER- So yeah, you see, basically that's where we're at.

27:36SPEAKERIs like, there are some places where you want a big giant

27:40SPEAKERX86 or X64 CPU with hundreds of cores and the whole thing.

27:46SPEAKERBut also there are places that there that you're gonna want

27:48SPEAKERsomething that's lower power and less,

27:50SPEAKERa little bit less capable maybe,

27:51SPEAKERbut also does exactly what you need it to in the place.

27:54SPEAKER- Do you ever want to just hang out in a data center?

27:59SPEAKER- They're really loud.

28:00SPEAKER- Well, I'm sure you'd probably want some noise canceling.

28:01SPEAKERI bet noise canceling headphones would work extremely well

28:04SPEAKERin that white noise kind of environment.

28:06SPEAKERBut like, are you, 'cause that's the ideal use case

28:10SPEAKERfor noise canceling headphones.

28:12SPEAKERBut are you as big a fan of air conditioning as I am?

28:15SPEAKER- I mean, right now, yes.

28:16SPEAKERI am 100% a fan of air conditioning.

28:18SPEAKERI would love to have air conditioning.

28:19SPEAKERIt seems like it'd be really nice.

28:20SPEAKER- How cool do you think it is in your average data center?

28:22SPEAKERDon't you want to just go in there

28:24SPEAKERand like lie down on the floor for a while?

28:26SPEAKER- So I've never been in a data center.

28:27SPEAKERI've been in a supercomputer before and it was not

28:32SPEAKERa comfortable environment for humans.

28:33SPEAKER- Oh, is it too cold?

28:35SPEAKER- It was very cold.

28:36SPEAKER- Really?

28:37SPEAKERAre we talking like convenience store

28:39SPEAKERwalk-in freezer type cold or what?

28:42SPEAKERSurely not that cold.

28:43SPEAKER- It was not, no, it was not 40 degrees.

28:45SPEAKERIt was probably 65.

28:47SPEAKER- Oh, that sounds pretty good.

28:48SPEAKER- It was chilly.

28:49SPEAKER- That sounds pretty good.

28:50SPEAKER- Look, I was in Texas and it was great

28:53SPEAKERfor the first minute or two.

28:54SPEAKERAnd then you're like, "Oh man, it's kind of cold in here."

28:56SPEAKERYou know when you're walking through the grocery store

28:59SPEAKERin the summertime and you get into the freezer aisle

29:02SPEAKERor the open, the butter aisle where the refrigerators

29:05SPEAKERare all open, wasting a ton of energy every day.

29:07SPEAKERIt's like that one where you're like,

29:09SPEAKER"Oh man, I wish I wasn't wearing shorts right now,

29:10SPEAKERbut this feels okay."

29:11SPEAKER- That sounds great.

29:13SPEAKER- Yeah.

29:15SPEAKER- If anybody can get us into a data center, let me know.

29:18SPEAKER- The supercomputer angle is interesting too,

29:20SPEAKERbecause in supercomputers, they use ARM processors

29:23SPEAKERto reduce latency and like kind of traffic cop

29:28SPEAKERa lot of the interactions between the other processors

29:30SPEAKERin the setup and different supercomputers

29:36SPEAKERhave different requirements.

29:37SPEAKERLike I spoke to somebody who works on supercomputers

29:41SPEAKERthat do fluid dynamics calculations where like

29:44SPEAKERthey're trying to figure out how wind interacts

29:45SPEAKERwith skyscrapers to affect weather patterns in big cities.

29:50SPEAKERAnd in order to do that, like every particle

29:54SPEAKERthat simulates a bit of wind has to be able to talk

29:57SPEAKERto every other particle 'cause they all interact

29:59SPEAKERand you end up with these massively complex situations

30:02SPEAKERwhere latency really matters.

30:04SPEAKERAnd if one part of the simulation gets out of sync

30:07SPEAKERwith the others, then the whole thing has to be scrapped

30:10SPEAKERand started over.

30:11SPEAKERSo like they build these clusters

30:13SPEAKERso that there's a negligible latency

30:16SPEAKERfrom one side to the other, and you need a bunch of CPUs

30:19SPEAKERlike evenly distributed through the network to do that.

30:21SPEAKER- To manage that.

30:22SPEAKER- Am I right in assuming that maybe the ARM cores

30:24SPEAKERin that scenario are not handling the heavy math

30:26SPEAKERof the simulation or they're just doing the,

30:28SPEAKERlike you said, the traffic hopping, the bookkeeping

30:30SPEAKERto keep everything in sync and that sort of thing?

30:33SPEAKER- Probably some of both, but I don't know.

30:35SPEAKERI'm not an expert, that is a level of granularity

30:38SPEAKERthat I do not have on this topic.

30:40SPEAKER- I mean, yeah, I'm just sort of like throwing that question

30:42SPEAKERout into the ether, but I would wonder

30:44SPEAKERif they're using like some kind of ASIC

30:46SPEAKERor something like GPU ask almost

30:49SPEAKERto handle the heavy compute aspect of that task.

30:53SPEAKER- So my understanding is that they're set up in a way

30:56SPEAKERthat they, that you have processors

31:00SPEAKERthat are tuned for specific things.

31:01SPEAKERSo like you'll have x86 cores

31:03SPEAKERfor doing lots of big floating point math.

31:05SPEAKER- That's what I was getting at.

31:06SPEAKER- Like that is pipelineable and deep,

31:09SPEAKERand then you'll do GPU cores for massively parallel stuff

31:13SPEAKERwhere you need thousands upon thousands upon thousands

31:15SPEAKERof cores, not hundreds of cores.

31:17SPEAKERAnd then, yeah, I think the ARM processors

31:19SPEAKERare gonna be the kind of making sure everybody gets fed

31:23SPEAKERbasically, all the different cores

31:26SPEAKERare getting the data they need in the right time.

31:30SPEAKERSo, okay, so like, why does Nvidia want ARM?

31:35SPEAKER- Why does Nvidia want ARM?

31:37SPEAKER- Well, yeah, they like having market share.

31:41SPEAKER- Yes.

31:42SPEAKER- And if you look at, Nvidia is right now,

31:47SPEAKERas a video game graphics company, sure,

31:49SPEAKERthey very much wanna be the machine learning,

31:52SPEAKERAI hardware provider, and you can tell that

31:57SPEAKERby looking at how they're pushing Ampere

31:59SPEAKERto workstations and data centers.

32:01SPEAKERYou can look at the fact that they're putting 10,000

32:04SPEAKERshader cores on these GPUs

32:07SPEAKERwith like dedicated machine learning units

32:09SPEAKERand ARM also you can get dedicated machine learning

32:13SPEAKERacceleration on ARM chips now.

32:15SPEAKER- Yeah, and like we should mention,

32:17SPEAKERlike a lot of Nvidia's own messaging

32:19SPEAKERaround this deal pertains to AI specifically.

32:22SPEAKERLike they, like the line I pulled out here

32:24SPEAKERis that ARM is based in Cambridge

32:26SPEAKERand they said that they want to turn the Cambridge location

32:28SPEAKERinto quote, a new global center of excellence in AI research.

32:33SPEAKER- Yes, well, and like if you think about Nvidia's path

32:38SPEAKERover the last 20 years, 'cause this is like

32:41SPEAKERa 24 year old, 23 year old company now,

32:43SPEAKERthat they looked at the PC graphics market

32:47SPEAKERand were like, look, we can build,

32:48SPEAKERwe can start building these,

32:50SPEAKERA, there's a market for PC graphics,

32:52SPEAKERbut B, this market provides a path

32:55SPEAKERto massively parallel machine learning accelerators,

32:59SPEAKERray tracing accelerators, all of that stuff.

33:01SPEAKERAnd they use the market demand for games and 3D graphics

33:08SPEAKERto do, to pay for the R&D to get to the point

33:12SPEAKERthat they had machine learning processors.

33:13SPEAKER- Can we just generally use the terms AI

33:17SPEAKERand machine learning interchangeably in this context?

33:19SPEAKER'Cause they use AI in most of their messaging.

33:23SPEAKER- I think AI is the more consumer friendly term, right?

33:27SPEAKER- But just to be clear, they mean basically the same thing.

33:30SPEAKER- Yeah, I mean, we could just call it,

33:31SPEAKERyou know, linear, statistical linear math,

33:34SPEAKERbut it's, you know, people are gonna be

33:36SPEAKERless interested in that.

33:37SPEAKERBut yeah, machine learning and AI, essentially, I mean not--

33:39SPEAKER- Yeah, I'm just trying to clarify.

33:41SPEAKER- I would say machine learning is probably a subset of AI

33:44SPEAKERif we wanted to be highly technical,

33:46SPEAKERbut I don't think it matters.

33:47SPEAKER- Okay.

33:48SPEAKERSo Nvidia has sold ARM chips in the past.

33:52SPEAKERThe Tegra is in the Switch.

33:54SPEAKERIt was in the Nvidia Shield and the Shield Pro,

33:57SPEAKERI think the iron. - It still is, yeah.

33:58SPEAKERYeah, it still is.

33:59SPEAKERI mean, yeah, I looked at the Tegra roadmap before this

34:02SPEAKERand like they've got a new one on the way.

34:05SPEAKEROf course, I mean, well, you know, whatever.

34:06SPEAKERThere are rumors about a new Switch next year.

34:08SPEAKERSo like there would be a new one in that, for example.

34:13SPEAKER- Well, but like they haven't made big inroads

34:15SPEAKERinto the phone market.

34:16SPEAKERThere were a couple of phones that shipped on Tegra

34:17SPEAKERa few years ago.

34:18SPEAKER- Was that Razer phone on Nvidia?

34:21SPEAKER- I feel like maybe--

34:23SPEAKER- The gamer phone.

34:24SPEAKER- It seemed like their performance per watt

34:27SPEAKERwas not, which in turn meant you had bad battery life

34:31SPEAKERon the phone, wasn't where you wanted it.

34:34SPEAKERBut it's been pretty good on like the Switch,

34:37SPEAKERwhile the initial launch Switch

34:39SPEAKERdidn't have the best battery life,

34:41SPEAKERit's more than sufficient for like what I've been using

34:44SPEAKERover the last four years or five years,

34:45SPEAKERhowever long it's been.

34:46SPEAKER- So one thought I had about this acquisition,

34:48SPEAKERlike obviously there's data centers

34:50SPEAKERand super computer stuff.

34:51SPEAKERThere's machine learning research.

34:53SPEAKERLike there's a lot of applications for this,

34:55SPEAKERbut like you tell me if this is an accurate perception

34:58SPEAKERof the mobile market.

34:59SPEAKERLike Apple's arm-based chips are like insane, right?

35:03SPEAKER- God tier, yeah.

35:04SPEAKER- Yeah, like they hired, they went out and hired

35:06SPEAKERlike some of the best chip designers in the world, right?

35:09SPEAKER- And spent billions of dollars.

35:10SPEAKER- Because they have unthinkable amounts of money.

35:13SPEAKERBut my understanding of the Android market

35:15SPEAKERis like the Qualcomm's and like the Android,

35:17SPEAKERor the ARM chips in that space just don't keep up, right?

35:20SPEAKERWith that.

35:21SPEAKER- I don't think it's that they spec up,

35:24SPEAKERlike if you look at-

35:25SPEAKER- Not saying they're bad necessarily, it's just that-

35:27SPEAKER- Well, it's a different situation

35:30SPEAKER'cause like Apple knows what their software team is doing

35:33SPEAKERand they have the software-

35:34SPEAKER- Yeah, that's fair, that's fair.

35:35SPEAKER- They turn their software roadmap

35:35SPEAKERto the hardware roadmap so they can like really

35:39SPEAKERget everything out of every bit of silicon

35:40SPEAKERthat they put in that chip.

35:41SPEAKER- Right.

35:42SPEAKERIf say, if you're LG and you're buying a Qualcomm processor,

35:47SPEAKERyou probably aren't gonna spend the billion dollars a year

35:49SPEAKERthat you need to turn that off the shelf processor

35:52SPEAKERinto the custom processor like Apple does.

35:55SPEAKER- Totally, but I mean, the point I'm getting at

35:57SPEAKERis like to use the most recent example

35:59SPEAKERwhen the iPhone SE came out,

36:00SPEAKERlike all the reviews were just like,

36:03SPEAKERit is insane the performance advantage

36:06SPEAKERthat this budget Apple phone has

36:08SPEAKERover the current flagship Android phones.

36:10SPEAKERSo what I'm getting at is like,

36:12SPEAKERI wonder if Nvidia maybe sees an opportunity

36:14SPEAKERto become like the dominant SOC supplier

36:18SPEAKERin the Android space off of this.

36:20SPEAKER- So my guess is that they're probably going to pull out

36:24SPEAKERof the Android SOC space as part of the regulatory burden

36:28SPEAKERfor this deal to happen.

36:29SPEAKER- Interesting, yes, I actually, yeah,

36:30SPEAKERthat's totally a good point.

36:32SPEAKERThat's like antitrust stuff, not withstanding,

36:34SPEAKERcould they be possibly be looking to do that?

36:37SPEAKER- I think Nvidia is more interested in the data center

36:39SPEAKERand making sure that things like TensorFlow cores

36:42SPEAKERend up in other people's phones.

36:44SPEAKER- Okay, 'cause this exact idea was floated as well

36:48SPEAKERin some of the Android press that I looked at

36:51SPEAKERor something going to the Android media market

36:52SPEAKERand like people speculating that like,

36:54SPEAKERthere's kind of nothing stopping Nvidia

36:56SPEAKERfrom holding the best of like, they can keep licensing.

37:02SPEAKER- I think the regulatory stuff in China and the UK

37:05SPEAKERwould probably be the thing that stops that.

37:07SPEAKER- The idea to expand on it is just that

37:09SPEAKERthey could keep licensing the arm designs

37:10SPEAKERthat they currently make to the same customers,

37:13SPEAKERbut that they could like have a slightly competitive,

37:17SPEAKERslightly more competitive version.

37:19SPEAKER- Yeah, they'll keep the good one for themselves.

37:21SPEAKER- Right, or something along those lines.

37:23SPEAKER- I would be shocked if that is allowed to happen.

37:27SPEAKERLike in the US, I wouldn't be surprised

37:28SPEAKERif it's allowed to happen.

37:29SPEAKERThe EU and the UK, that's not going to fly.

37:32SPEAKER- Maybe this is just expressing my complete lack of faith

37:35SPEAKERin any kind of antitrust action to ever take place again.

37:39SPEAKERUnless it's politically motivated.

37:41SPEAKER- I think, well, here's the thing.

37:43SPEAKERGiven that Nvidia has essentially no market share

37:45SPEAKERin that space right now, my guess is that they sacrifice

37:48SPEAKERthe potential future there for licensing out instruction set

37:53SPEAKER'cause here's the thing is if they do something

37:55SPEAKERtoo anti-competitive, there are other similar,

37:59SPEAKERbut not directly compatible instructions

38:02SPEAKERthat we use out there, including some open stuff.

38:05SPEAKER- Yes, that stuff is probably,

38:07SPEAKERit's not even arm related necessarily,

38:08SPEAKERbut that's probably the actual most fascinating aspect

38:11SPEAKERof this topic, but we'll get to that.

38:13SPEAKER- Yeah, so I don't, I think my guess is that Nvidia

38:18SPEAKERwill protect the switch and shield business

38:21SPEAKERfor set top boxes and 'cause honestly,

38:25SPEAKERlike if you're thinking about Nvidia,

38:26SPEAKERif Nvidia's Tegra problem is that they perform really well,

38:29SPEAKERbut their performance per watt and battery life are bad,

38:33SPEAKERtablets make a ton of sense like the switch,

38:35SPEAKERlike the shield tablet, set top boxes make a ton of sense

38:39SPEAKERwhere your performance per watt is less important

38:41SPEAKERas long as you can passively cool it, nobody cares.

38:43SPEAKERSo I think we'll see them sacrifice the ability

38:47SPEAKERto be in phones 'cause it like, it's not like they're not,

38:49SPEAKERis it, here's the question you need to ask is,

38:52SPEAKERis it better for Nvidia to make a little bit of money

38:56SPEAKERon every single one of the one or 2 billion phones

38:59SPEAKERthat are sold every year or to make a lot of money

39:02SPEAKERon some small percentage of the market

39:04SPEAKERbecause they piss everybody off

39:05SPEAKERand they all switched to MIPS or whatever.

39:07SPEAKER- Sure, sure, yes, that is an actual exact question I have

39:10SPEAKERwhen we get to the MIPS and RISC-V stuff,

39:12SPEAKERbut I'll save that for now.

39:15SPEAKER- I think it's, like I said, I think it's much more likely

39:17SPEAKERthat we'll see it go the other way

39:18SPEAKERwhere Nvidia makes tensor cores and things like that,

39:23SPEAKERlike specific parts of the current GPU architecture,

39:27SPEAKERthings like their H.264 and H.265 live encoders,

39:31SPEAKERwhich are really good,

39:32SPEAKERmake them available to ARM licensees

39:35SPEAKERalong with the other IP.

39:37SPEAKER- What you're saying is it's less likely

39:38SPEAKERthat Nvidia is going to take the ARM tech

39:40SPEAKERand kind of pull it behind their curtain

39:42SPEAKERand more likely that some of the Nvidia tech

39:43SPEAKERis gonna bleed back into the publicly available ARM stuff?

39:46SPEAKER- I would really, really hope that the regulatory,

39:50SPEAKERone of the three or four regulatory agencies involved

39:54SPEAKERwould ensure that that doesn't happen.

39:56SPEAKER- Right, so like an example is like ARM has a GPU design

39:59SPEAKERthat they can license along with the CPU stuff,

40:03SPEAKERbut you're saying that like Nvidia graphics

40:06SPEAKERcould end up replacing that Mali GPU design, for example.

40:10SPEAKER- Even some subset, like small bits and pieces.

40:13SPEAKERThe other thing is like their PowerVR

40:16SPEAKERhas a whole ARM GPU division,

40:19SPEAKERlike they sell a buttload of ARM GPUs too.

40:21SPEAKERLike there are, there have been three companies

40:24SPEAKERmaking ARM GPUs for a while, four if you count Apple,

40:27SPEAKEReven though they don't sell those external vendors.

40:31SPEAKERLike there's, the beautiful thing about ARM

40:34SPEAKERis if LG wants or Qualcomm wants Nvidia's GPU,

40:40SPEAKERthey can do that.

40:40SPEAKERIf they want their own GPU, they can do that.

40:42SPEAKERIf they want PowerPC, they can do that.

40:43SPEAKERAnd you can assemble the chips however you want.

40:45SPEAKERIt's up to you.

40:47SPEAKERIt's like Lego, but way more complicated

40:49SPEAKERand expensive to build.

40:52SPEAKERSo yeah, like I think that's it.

40:54SPEAKERI think, and so the outside edge case for this

40:58SPEAKERis that I think we're gonna see Nvidia dropping ARM CPUs

41:03SPEAKERin weird places in the future.

41:05SPEAKERWe saw a little bit of this with the Ampere stuff

41:07SPEAKERand the direct storage API to accelerate storage,

41:12SPEAKERwhich in reality isn't going to start happening

41:14SPEAKERfor like two or three years

41:15SPEAKERbecause it takes the API's a while to roll out.

41:19SPEAKERI think we'll see in the next five or 10 years,

41:23SPEAKERwe'll see Nvidia dropping ARM CPUs on GPUs to reduce the,

41:30SPEAKERsimply because there's often math

41:32SPEAKERthat you need to do in games that is latency dependent.

41:35SPEAKERLike this is one of the benefits consoles have,

41:38SPEAKERis that because the consoles use a shared CPU memory pools,

41:42SPEAKERwhere the CPU and GPU are both talking

41:44SPEAKERto the same memory all the time,

41:45SPEAKERit's really inexpensive for the GPU to pass something off

41:49SPEAKERto the CPU and then come back to the GPU again.

41:51SPEAKER- Also the CPU and the GPU are on the same die,

41:53SPEAKERso there is no--

41:54SPEAKER- Often, yeah.

41:55SPEAKER- There's a system bus to communicate over.

41:58SPEAKER- Right, and whereas on the PC,

42:00SPEAKERthe GPU has to talk across the PCI express bus,

42:03SPEAKERwhich like on a GPU that's running at 3,000,

42:06SPEAKERthousands of megahertz,

42:08SPEAKERyou're looking at sometimes five or 10 or 20 clock cycles

42:11SPEAKERon the GPU before a communication can happen

42:14SPEAKERacross the PCI express bus.

42:15SPEAKERAnyway.

42:16SPEAKER- My one question about that idea of including

42:18SPEAKERlike arm designs on a graphics card, for example,

42:22SPEAKERis like, I mean, would you agree that Nvidia

42:24SPEAKERprobably employs a lot of the best integrated circuit

42:27SPEAKERdesigners in the world at this point?

42:29SPEAKER- I, it would be hard for me to say that.

42:31SPEAKER- Okay.

42:32SPEAKER- But sure, the results are good.

42:34SPEAKER- That's what I mean.

42:35SPEAKER- Yeah.

42:36SPEAKER- The proof is in the pudding there, it seems,

42:38SPEAKERbut like, do you think like a general purpose,

42:40SPEAKERlike arm set up on a graphics card would make more sense

42:43SPEAKERthan like, or let me back up.

42:45SPEAKER- They seem really good at designing logic on their cards

42:49SPEAKERbuilt for purpose, right?

42:50SPEAKEROr designed for purpose, like this direct storage stuff.

42:53SPEAKERLike you assume that there is hardware on those GPUs

42:56SPEAKERthat is very good at decompressing that data.

42:59SPEAKERLike would it make sense to have a more general purpose unit

43:04SPEAKERon there as opposed to just designing whatever you need?

43:08SPEAKER- Look, it's not outside the realm of possibility

43:09SPEAKERthat we're going to be plugging NVMe drives

43:11SPEAKERonto your video card in five years.

43:15SPEAKER- Okay, so at that point, a very small low power arm core

43:18SPEAKERthat handled that IO would make sense or something like that.

43:21SPEAKER- I see, yeah.

43:21SPEAKER- Okay.

43:22SPEAKER- Like that's kind of what I was getting at

43:25SPEAKERa couple of weeks ago when we did the 3080 stuff.

43:28SPEAKERWhen I mentioned, I mentioned that the graphics card

43:31SPEAKERis rapidly becoming like the central feature of a PC.

43:34SPEAKER- Hey guys, plug your mouse into your video card

43:37SPEAKERfor the best performance.

43:38SPEAKER- Right, like have your video card

43:39SPEAKERhandle your storage IO now.

43:41SPEAKERYou know, like at some point the graphics card

43:42SPEAKERis the computer.

43:44SPEAKER- Well, I mean, that's what Jen Sun's been trying

43:46SPEAKERto make happen for 20 years.

43:47SPEAKER- That's what I mean.

43:48SPEAKERLike, I feel like we're moving toward that

43:50SPEAKERwhere like what does this, I mean, obviously you need a CPU,

43:53SPEAKERbut like it's increasingly marginalized it seems like.

43:56SPEAKER- It was literally a talking point,

43:59SPEAKERI think around the like the GeForce,

44:01SPEAKERprobably the 5800 launch, 5900 launch maybe

44:04SPEAKERwhen they started doing compute

44:05SPEAKERon like general purpose compute on GPUs.

44:09SPEAKERHe's like, yeah, this is going to be the most important

44:10SPEAKERprocessor in your computer from now on.

44:12SPEAKER- Right.

44:13SPEAKER- And you know, turns out probably sounded hyperbolic then,

44:17SPEAKERbut it was probably correct.

44:18SPEAKER- Like I'm trying to conceptualize in my head

44:21SPEAKERthe weird fuzzy shifting line between the graphics card

44:25SPEAKERand like say the motherboard.

44:27SPEAKERLike for expansion purposes, it would never make sense

44:31SPEAKERto turn a graphics card into a motherboard

44:32SPEAKERand literally have the GPU on the board

44:35SPEAKERand just plug a CPU into it.

44:36SPEAKERBut like, it feels like we're kind of like,

44:38SPEAKERthat distinction is becoming less meaningful.

44:42SPEAKERDoes that make sense?

44:44SPEAKER- Yeah, well, when you look at those big compute clusters

44:46SPEAKERwhere they have like a huge board

44:48SPEAKERwith a shitload of PCI slots and lanes,

44:50SPEAKERand then you just drop eight boards into it with one CPU,

44:54SPEAKERlike the amount of power that goes to doing the GPU math

44:59SPEAKERversus the CPU math is strongly, strongly, strongly

45:03SPEAKERin favor of the GPU side.

45:04SPEAKER- Right.

45:06SPEAKER- It's a weird, like computers are going to be real weird

45:08SPEAKERin about 10 years.

45:09SPEAKER- Yeah, totally.

45:10SPEAKER- That's the TLDR.

45:11SPEAKER- Totally, that's kind of what I'm getting at.

45:12SPEAKERBut like at some point, if the graphics card

45:13SPEAKERis handling 90% of your system functionality anyway,

45:16SPEAKERwhy not just make that the whole system?

45:18SPEAKER- Well, yeah.

45:20SPEAKERI mean, it's, I mentioned that there are people

45:25SPEAKERin our audience I know who are further into this,

45:28SPEAKERdown this rabbit hole of what computers look like

45:29SPEAKERin 10 years than we are.

45:31SPEAKER- Yes.

45:31SPEAKER- I'm really curious to hear their reaction

45:33SPEAKERto this conversation.

45:34SPEAKER- I literally copied and pasted a conversation

45:36SPEAKERout of the discord a few days ago

45:37SPEAKERabout like where CPUs are going in the next 10 to 15 years

45:40SPEAKERthat like frankly made my head explode.

45:44SPEAKERSome really wild stuff in there.

45:46SPEAKER- Well, I mean, and this isn't like, this isn't,

45:49SPEAKERit's not idle speculation.

45:52SPEAKERLike look at what a computer looked like in 2000.

45:55SPEAKERWhen I moved to California,

45:56SPEAKERI think I had a dual core Celeron machine.

46:00SPEAKERSo two physical packages,

46:03SPEAKERsocketed into two sockets on the motherboard.

46:06SPEAKERAnd like the difference in that computer

46:09SPEAKERand the computer that is in my pocket now is astounding.

46:12SPEAKER- Orders of magnitude, right?

46:14SPEAKER- Yeah, it's because of ARM.

46:15SPEAKER- Yeah.

46:16SPEAKER- And like everything, the wireless internet connection

46:18SPEAKERis better than my wired internet connection was.

46:20SPEAKERMy screen is way higher resolution.

46:22SPEAKERI can do more things at the same time than I could then.

46:26SPEAKERLike it's, yeah, it's bonkers.

46:27SPEAKERThe whole thing's crazy.

46:28SPEAKER- Yes.

46:29SPEAKERYeah, the acceleration of development of technology

46:31SPEAKERhas been pretty dizzying for me.

46:34SPEAKER- It's a lot of, yeah, a lot of that is on ARM.

46:37SPEAKERLike the rise of phones that move billions of units a year,

46:44SPEAKERlike left the opportunity to really push the technology

46:48SPEAKERbecause there was a lot of money at stake and yeah.

46:52SPEAKER- But just, I mean, not to get off on some old man rant,

46:55SPEAKERbut like if you started on,

46:57SPEAKERif you started out with parallel ports

46:58SPEAKERand text mode only displays and that sort of thing,

47:03SPEAKERlike where we are now versus where we are then

47:05SPEAKERis like, quite frankly, hard to grapple with sometimes.

47:08SPEAKER- It has been a long road.

47:09SPEAKER- Yeah, but I'll finish a short one.

47:13SPEAKER- I mean, yeah, you remember,

47:15SPEAKERI remember taking hours to download a one megabyte driver.

47:20SPEAKER- Yeah, yeah.

47:23SPEAKER- So Brad, what do you think this means for phones?

47:25SPEAKER- 300 baud, 300 baud.

47:26SPEAKER- I never had a 300 baud.

47:27SPEAKERI started with a 14.4, I think.

47:29SPEAKER- Yeah, so did I.

47:30SPEAKERI didn't either.

47:30SPEAKERI just, I know it existed.

47:31SPEAKERI'm sorry, what were you asking?

47:33SPEAKER- Look, I couldn't hear you.

47:34SPEAKERMy audio coupler came off the phone.

47:36SPEAKERSo we got to re-sync this modem call.

47:40SPEAKER- Jam that thing back in there.

47:41SPEAKERYou got to rerun your init string.

47:44SPEAKER- What does this mean for phones?

47:46SPEAKER- For phones?

47:47SPEAKER- For phones.

47:48SPEAKER- Oh, you mean like right now?

47:49SPEAKER- Yeah, right now.

47:49SPEAKERIf V is buying ARM, shit, is my phone going to be bad now?

47:52SPEAKER- We covered my apparently abortive idea

47:55SPEAKERthat they were going to try to dominate the Android space.

47:57SPEAKERSo probably not a ton at the moment.

48:01SPEAKERMaybe the Android, like you said,

48:03SPEAKERthe Android SOCs get better GPUs or something like that,

48:07SPEAKERor more machine learning acceleration, something like that.

48:10SPEAKERBut like that's probably five, maybe not five,

48:14SPEAKERbut like that's probably like minimum three,

48:16SPEAKERfour years away from showing up and shipping products, right?

48:18SPEAKERWith the regulatory overhead and the R&D to product cycle

48:24SPEAKERlead times and stuff like that.

48:25SPEAKER- I think they suggested this is going to take 18 months

48:28SPEAKERto get approval and maybe close.

48:30SPEAKER- Right.

48:31SPEAKER- So yeah, I think my assumption is,

48:35SPEAKERmy assumption is there's not going to be too much work

48:37SPEAKERhappening by other parties on anything that comes out of

48:41SPEAKERthis until that deal is closed.

48:42SPEAKER- This is, here's like a really high level legal

48:45SPEAKERand business question.

48:46SPEAKERIs there anything to stop them from beginning R&D

48:48SPEAKERwith this purchase in mind prior to the deal

48:51SPEAKERbeing fully approved?

48:53SPEAKER- I have no idea.

48:54SPEAKER- Yeah, right.

48:55SPEAKERLike, I don't know any, I quite literally don't know anybody

48:57SPEAKERwho would know the answer to that question.

48:59SPEAKERWe need like high level, like IP lawyer types

49:03SPEAKERand regulatory, like antitrust law types

49:06SPEAKERto answer a question like that, right?

49:08SPEAKER- The amount of anxiety I had when I had bought a house,

49:11SPEAKERbut hadn't closed on it yet and had gotten the keys

49:13SPEAKERand started knocking holes in the walls.

49:15SPEAKER- Right.

49:16SPEAKER- Was very high, I can't imagine.

49:18SPEAKER- That's literally that question, right?

49:19SPEAKERLike, is it okay for them to start knocking holes

49:21SPEAKERin the dry wall before the--

49:24SPEAKER- I don't, I honestly don't know.

49:25SPEAKER- Before they've got the deed in their hands.

49:27SPEAKER- I mean, do they exchange IP?

49:29SPEAKER- I guess, I mean, probably not.

49:30SPEAKER- Probably not.

49:31SPEAKER- 'Cause it's all proprietary and secret.

49:32SPEAKER- Yeah, you're probably right, you're probably right.

49:35SPEAKER- I mean, but I assume they're talking,

49:36SPEAKERI assume like they start, I assume anytime you have,

49:39SPEAKERlike I know in media spaces, 'cause I've worked

49:41SPEAKERwith some people who went through a giant merger recently,

49:44SPEAKERlike they were talking to, everybody was talking

49:45SPEAKERto everybody, but nothing was happening,

49:48SPEAKERbut everybody knew what was coming.

49:50SPEAKERSo I'd say that's probably what happened.

49:51SPEAKERI don't know, I'm curious to hear what people think.

49:53SPEAKER- Do you ever--

49:54SPEAKER- If you're in a video, send us an email or signal me.

49:58SPEAKER- Do you ever have this tendency when you see

49:59SPEAKERa deal of this magnitude?

50:00SPEAKERI mean, like you said, this is like basically

50:02SPEAKERpractically the biggest deal that's ever been made.

50:04SPEAKERDo you ever have a tendency to look at that

50:06SPEAKERand try to ascribe some kind of like organization

50:08SPEAKERand method to the whole thing?

50:10SPEAKERLike think like, oh, there must be a plan,

50:12SPEAKERlike fully written down and they know exactly

50:14SPEAKERwhat they're doing and they're gonna execute it.

50:15SPEAKERBut like then what you just described

50:17SPEAKERis the actual reality of, it's actually just a bunch

50:19SPEAKERof human beings flailing around, trying to figure it out

50:23SPEAKERfor a period of months to years, right?

50:25SPEAKER- I mean, you don't spend $40 billion

50:27SPEAKERif you don't have a plan for what you want to do

50:29SPEAKERwith the thing you're buying, right?

50:31SPEAKER- Right, but it's not all set in stone.

50:33SPEAKER- You know, I mean, I think everything's fluid

50:36SPEAKERand like there's some parts of the arm business

50:37SPEAKERthat are getting spun off that Nvidia didn't want,

50:40SPEAKERstuff like that.

50:41SPEAKER- Yeah, I mean, there's much more of a human,

50:43SPEAKERlike in the moment kind of human elements

50:46SPEAKERof the decision-making, right?

50:47SPEAKERThan you might think.

50:48SPEAKER- Yeah, yeah, yeah.

50:49SPEAKER- Yes.

50:55SPEAKERI was a little, $40 billion is a staggering amount of money.

51:00SPEAKERGiven the scale of that arm business,

51:02SPEAKERI was a little bit surprised that it was only $40 billion.

51:06SPEAKER- I could see that, sure.

51:07SPEAKER- Like when arm sold a few years ago for 20 or 30 billion,

51:11SPEAKERwhatever it was, I was kind of surprised that it was only,

51:16SPEAKERI assume that the licensing fees aren't onerous per chip.

51:20SPEAKER- Okay, well that is a good segue into the next topic then,

51:23SPEAKER'cause I was wondering about that exact thing.

51:26SPEAKER- Yeah, I mean, I'm curious.

51:30SPEAKERI mean, these are very not public numbers.

51:32SPEAKERIt's funny, it's even hard.

51:35SPEAKERI was trying to figure out how much

51:37SPEAKERof the global Microsoft processor business,

51:40SPEAKERlike across all architectures

51:42SPEAKERand all markets arm represented,

51:45SPEAKER'cause it has to be a significant portion.

51:50SPEAKERBut like that stuff is even all buried

51:52SPEAKERin really expensive analyst reports

51:54SPEAKERabout microprocessor, if I wanted to pay five or $6,000,

51:59SPEAKERI could have gotten a report that would let me know

52:00SPEAKERexactly how much microprocessor business arm gets

52:04SPEAKERand what their licensing model looks like.

52:05SPEAKERBut I didn't spend that five grand, Brad, I'm sorry.

52:08SPEAKER- Sounds like we need a $5,000 tier on the Patreon next month

52:11SPEAKERto get this analyst report in our hands.

52:14SPEAKER- Give us the 5,000 bucks,

52:15SPEAKERwe'll buy one analyst report a month.

52:16SPEAKER- That's right, that's right.

52:17SPEAKER- And we'll put it right back into the pod.

52:20SPEAKER- You can just send us the report if you have it.

52:22SPEAKERI won't tell anybody.

52:23SPEAKER(indistinct)

52:27SPEAKER- So yeah, I guess that brings us

52:29SPEAKERto kind of the last big topic here,

52:30SPEAKERwhich you have written these notes

52:32SPEAKERas are there alternatives to arm for phones?

52:35SPEAKER- Not really right now, but maybe soon.

52:37SPEAKER- But yeah, it seems like it's getting there.

52:38SPEAKER- Well, I mean, Intel has been trying to push an x86

52:41SPEAKERthat part for phones for a really long time

52:43SPEAKERand nobody's biting.

52:44SPEAKER- Good luck with that, as we said.

52:47SPEAKER- But I think MIPS and RISC-V have both shown.

52:53SPEAKERRISC-V.

52:55SPEAKER- Is it RISC-V?

52:56SPEAKER- Yeah, it's RISC-V.

52:57SPEAKER- Oh, I thought it was RISC-V.

52:58SPEAKER- It's the fifth, yeah, I agree, that sounds better,

53:00SPEAKERbut it is the fifth generation of RISC.

53:02SPEAKER- If it was, why is the hyphen in there if it's five?

53:05SPEAKER- According to them.

53:06SPEAKERIt's right there on the Wikipedia page, man.

53:07SPEAKER- I hate this.

53:08SPEAKER- I don't know what to tell you.

53:09SPEAKER- I regret, look, I was pro RISC, so RISC-V is-

53:13SPEAKER- RISC-V pronounced RISC-V.

53:16SPEAKER- God.

53:17SPEAKER- That's literally how the Wikipedia page starts.

53:18SPEAKERThat also came out of Berkeley, actually.

53:19SPEAKERSo I looked into it while we were talking about it.

53:21SPEAKERWe were both right earlier.

53:23SPEAKERThe concept of RISC started at Berkeley,

53:25SPEAKERbut then MIPS was the company

53:27SPEAKERthat spun off of that research

53:29SPEAKERand was the first commercial entity to offer a RISC.

53:33SPEAKER- But it seems like there was parallel development

53:36SPEAKERhappening in other places at the same time,

53:37SPEAKER'cause DC had a MIPS chip, a RISC chip, and all that.

53:41SPEAKER- Is that like how calculus was created

53:43SPEAKERby more than one person simultaneously?

53:44SPEAKER- Leibniz and Leibniz and Newton, yes.

53:48SPEAKERBut-

53:49SPEAKER- Newton got the win, though,

53:50SPEAKERwhen he spoke English, it turns out.

53:51SPEAKER- Yeah, yeah.

53:53SPEAKER- But he had a better publicist, I think.

53:55SPEAKER- Yeah, that whole Apple thing is very easy to remember.

53:57SPEAKER- Yeah, also apocryphal.

53:59SPEAKER- Yes, I'm sure, I'm sure.

54:00SPEAKERBut there's a little bit of public myth-making

54:03SPEAKERgoes a long way in getting people to remember you,

54:05SPEAKERit turns out.

54:07SPEAKERSo yeah, which one do we wanna talk about?

54:09SPEAKERWell, it's kind of the same topic.

54:11SPEAKER- Well, RISC-V is gonna be, is an open source architecture.

54:15SPEAKER- It's literally an open source instruction set

54:17SPEAKERin the way that open source software is freely available.

54:21SPEAKER- You can take that instruction set

54:23SPEAKERand add a few billion dollars and hundreds of people

54:27SPEAKERand maybe thousands of people,

54:29SPEAKERand you can have a CPU in a few years.

54:31SPEAKER- Yeah, I guess that's the big difference here,

54:32SPEAKERis that a single programmer can take

54:35SPEAKERan open source repository and do something cool with it,

54:39SPEAKERand the barrier is very low.

54:40SPEAKER- Depends on the repository, but yeah.

54:42SPEAKER- Well, you know what I mean, though.

54:42SPEAKERA single person can develop software pretty effectively,

54:45SPEAKERdepending on what they're doing and how skilled they are,

54:46SPEAKERbut to take an open instruction set

54:50SPEAKERand actually turn it into a tangible product

54:51SPEAKERtakes unthinkable amounts of money and skill

54:56SPEAKERto design the chip.

54:58SPEAKERGetting chips made is incredibly expensive and difficult.

55:00SPEAKERIt's not like some little garage startup

55:04SPEAKERis not gonna take this and change the world, right?

55:06SPEAKER- I mean, they might.

55:07SPEAKER- Well, maybe, maybe.

55:08SPEAKER- That's not out of the question.

55:09SPEAKER- I guess not.

55:10SPEAKER- It's not, this isn't...

55:14SPEAKER- I still firmly, you know, that's a really good question.

55:17SPEAKERLike, is there, is this still a market

55:20SPEAKERthat a garage startup or even like a billion dollar startup

55:24SPEAKERcan compete in?

55:25SPEAKER- Maybe.

55:26SPEAKER- Probably.

55:27SPEAKER- The thing that stuck out to me in reading about RISC-V

55:30SPEAKERis it's something about ARM, actually,

55:32SPEAKERthat I didn't really, it had never occurred to me,

55:34SPEAKERwhich is that like RISC-V being open,

55:36SPEAKERyou can take the literal base instruction set

55:38SPEAKERand do whatever you want with it.

55:39SPEAKER- Right.

55:40SPEAKER- I did not realize that an ARM license

55:42SPEAKERis not anywhere near that permissive.

55:45SPEAKERYou're essentially, like you can extend the ARM designs,

55:49SPEAKERbut not that much in comparison.

55:51SPEAKERDoes that make sense?

55:52SPEAKERLike, it sounds like you're largely kind of taking the cores

55:55SPEAKERthat ARM offers with some customizations

55:57SPEAKERthat you can specify, but you can't like,

55:59SPEAKERyou cannot just take the ARM instruction set and license it

56:01SPEAKERand then make your own shift from the ground up with it.

56:03SPEAKER- No, you can do that.

56:04SPEAKERThat's what Apple did specifically.

56:06SPEAKER- Oh, okay.

56:07SPEAKERThis thing I was reading was wrong about that then.

56:10SPEAKER- Well, what you're saying is you can't license

56:14SPEAKERthe instructions for ARM and then build your own chip.

56:21SPEAKERThey'll, like a lot of businesses, if the price is right,

56:25SPEAKERyou can do the thing you're looking for, right?

56:27SPEAKER- Sure, but maybe that's like written

56:29SPEAKERinto the licensing agreement and then it costs more

56:31SPEAKERor something like that, right?

56:31SPEAKER- Right.

56:32SPEAKER- That's what I'm getting at is like,

56:34SPEAKERyou're not just gonna pay ARM a flat rate,

56:35SPEAKERtake those instructions and then go off to the races

56:37SPEAKERand do your own thing.

56:38SPEAKER- Yeah, if you license the Cortex-A9,

56:41SPEAKERyou're licensing the chunks that make up the Cortex-A9

56:44SPEAKERand then how you assemble them is up to you.

56:45SPEAKER- That's what I'm getting at is you,

56:47SPEAKERwhereas the difference is with something like a RISC-V

56:49SPEAKERor actually now MIPS, which we're getting to,

56:52SPEAKERyou can take the literal instructions

56:54SPEAKERand like build your own CPU core from the ground up

56:57SPEAKERis my understanding.

56:59SPEAKER- And that is, so I'm gonna be interested to see

57:04SPEAKERif open source projects can handle this.

57:07SPEAKERMy guess is that probably they can't.

57:10SPEAKERLike it seems like too big a thing for the typical chaos

57:13SPEAKERof an open, of a big open source project.

57:16SPEAKER- I could see that.

57:18SPEAKER- I also don't know.

57:20SPEAKER- Should we mention the MIPS stuff before we get too much?

57:22SPEAKER- Yeah, the MIPS stuff is interesting.

57:26SPEAKER- They have opened, they have open sourced

57:28SPEAKERtheir architecture, which like you said,

57:29SPEAKERhas been around since the '80s, right?

57:30SPEAKERLike MIPS, my router has a MIPS 64 CPU, for example.

57:36SPEAKER- Yeah, like a lot of networking hardware

57:39SPEAKERis powered by MIPS processors.

57:40SPEAKER- So it's really interesting, especially interesting

57:42SPEAKERin the context of my router, the edge router four,

57:44SPEAKER'cause like I SSH into it, it's just running Debian,

57:47SPEAKERbut it is a, you know, it had obviously by nature,

57:50SPEAKERit has to be a MIPS 64 compiled version of Debian.

57:54SPEAKER- Yeah.

57:55SPEAKER- So like when I'm going on, when I go on there

57:56SPEAKERand add like the app repository to look around

57:58SPEAKERfor like software to install on there,

58:00SPEAKERlike there have to be MIPS 64 compiled versions

58:03SPEAKERof those, of those, those binaries available to actually

58:06SPEAKERdo anything with them, you know what I mean?

58:09SPEAKER- It's, it's, but like, it's an interesting,

58:14SPEAKERlike it's, if you look at ARM as embedded processors,

58:17SPEAKERwhich is what we would have called these like 10 years ago,

58:20SPEAKERMIPS is the original kind of embedded processor, right?

58:23SPEAKER- Right, right.

58:24SPEAKER- But I think there was even a MIPS version

58:26SPEAKERof Windows NT up until four.

58:28SPEAKER- Sounds right, 'cause I think they were

58:30SPEAKERin workstations back then, right?

58:32SPEAKER- Yeah, they, they were, they were early risk chips

58:35SPEAKERand people were excited about them

58:36SPEAKERso they ended up in weird places.

58:38SPEAKER- So this article, this article you linked

58:40SPEAKERbasically makes it sound like MIPS has just lost

58:42SPEAKERdramatic amounts of market share to ARM over the years.

58:45SPEAKER- Well, that's the, so the challenge is that as ARM

58:47SPEAKERhas developed more performance per watt for higher end CPUs,

58:51SPEAKERfor higher end SoCs for phones, they, they still have like

58:55SPEAKERa really deep well of designs that go back, you know,

58:58SPEAKERand can scale down to less capable devices

59:01SPEAKERthat don't need, don't need the same kind of hardware.

59:04SPEAKERBut yeah, they still, they're 10 billion chips

59:06SPEAKERchipped that use MIPS, I think is what they advertise.

59:09SPEAKER- Okay, that's not exactly nothing.

59:12SPEAKER- Like a small chip manufacturer still is selling billions

59:16SPEAKERand billions of chips.

59:17SPEAKER- Okay, that's fair.

59:18SPEAKERBut the point is they have now opened their architecture.

59:21SPEAKER- Yeah.

59:22SPEAKER- Just like RISC-V except they are a commercial company

59:25SPEAKERthat has also made this open.

59:27SPEAKER- So yeah, like there are, so the takeaway on this,

59:31SPEAKERif the, if the question we're answering is,

59:33SPEAKERare there alternatives for ARM for phones?

59:35SPEAKERThe answer kind of not right now, but maybe in the future.

59:39SPEAKER- Well, they're like, they're available.

59:40SPEAKERIt's just, nobody has done the insane amount of like work

59:42SPEAKERit would take to actually implement them in that role.

59:45SPEAKER- Yeah, and in the timescale, like if we're looking at,

59:48SPEAKERyou know, two to three years to close this deal,

59:51SPEAKERis there a way, like, will there, could Apple switch

59:55SPEAKERfrom ARM to MIPS or ARM to RISC-V, five, whatever?

59:59SPEAKERRISC-V, RISC-V, could they in the time, like,

1:00:04SPEAKERI wonder if Apple would have committed to full ARM

1:00:08SPEAKERafter this deal closed.

1:00:10SPEAKERI know that's been a thing that's been obviously

1:00:12SPEAKERin the works for 10 years on their side,

1:00:14SPEAKERbut Apple and Nvidia have a kind of inimical relationship

1:00:18SPEAKERthat has not been great.

1:00:21SPEAKERAnd I'll be interested to see how they react to this.

1:00:23SPEAKER- Yes, so that was like, when you mentioned like,

1:00:25SPEAKERmaybe the ARM licensing fees are not actually that high,

1:00:27SPEAKERlike that was the exact question I wanted to address

1:00:29SPEAKERwith this was, would Apple switch?

1:00:32SPEAKERIs it worth it?

1:00:32SPEAKERLike, you know, Apple being like one of the most

1:00:35SPEAKERcash-rich companies in the world, right?

1:00:37SPEAKERIs it, and with the number of devices they sell,

1:00:39SPEAKERlike are the licensing fees on ARM just low enough

1:00:41SPEAKERthat it's not worth the investment for them to switch

1:00:43SPEAKERto a licensing-free kind of situation?

1:00:47SPEAKER- So there's multiple components.

1:00:50SPEAKEROne is that ARM has been optimized relentlessly

1:00:52SPEAKERfor the last 13 years in a way that these other

1:00:55SPEAKERarchitectures have not, but like,

1:00:57SPEAKERthey're not as fully understood.

1:00:59SPEAKERThey don't know where the slow and the fast,

1:01:01SPEAKERwhere the bottlenecks are.

1:01:02SPEAKERThey don't, like, we don't have institutional knowledge

1:01:05SPEAKERin how to design really fast, capable MIPS processors

1:01:09SPEAKERor RISC-V processors for phones and other tablets

1:01:14SPEAKERand set-top boxes.

1:01:15SPEAKER- Like, I'm just throwing this out there,

1:01:17SPEAKERbut this would probably be at least like,

1:01:19SPEAKERwhat, a decade long undertaking to implement?

1:01:22SPEAKER- Yeah, my guess is that Apple looked at the power

1:01:25SPEAKERperformance curves on ARM in like 2007,

1:01:28SPEAKERwhen they started selling iPhones and they were like,

1:01:30SPEAKERhey, you know, if this continues at this rate,

1:01:32SPEAKERthe Intel line and the ARM line are gonna cross

1:01:35SPEAKERin about 2020, 2018.

1:01:38SPEAKERAnd at that point we could start not doing Intel chips

1:01:40SPEAKERanymore if we wanted.

1:01:41SPEAKERAnd that's been the process this whole time.

1:01:43SPEAKERThat's been in the works this whole time.

1:01:44SPEAKERSo yeah, you're looking at 10 years from now,

1:01:48SPEAKERfrom the starting line to when you start seeing

1:01:51SPEAKERviable competitors shipping.

1:01:53SPEAKERAnd like everything else, you'll start seeing

1:01:55SPEAKERlike weird dev kits two or three years before

1:01:59SPEAKERthe actual viable products get done.

1:02:01SPEAKERLike, you know, if you think back to that first Chromebook

1:02:04SPEAKERIntel released, it was slow and useless.

1:02:07SPEAKERAnd then two years later, they're pretty nice

1:02:09SPEAKERand worth having.

1:02:10SPEAKER- Yeah, but some of the, just the little bit

1:02:12SPEAKERI've been reading about these open architectures,

1:02:14SPEAKERlike analysts in that space seem pretty bullish

1:02:17SPEAKERon their future though.

1:02:19SPEAKER- I would, I think that's, I think,

1:02:22SPEAKERI think anytime you get a monolithic company,

1:02:27SPEAKERlike the people will be hedging,

1:02:29SPEAKERbut we'll be investing in these as a hedge

1:02:31SPEAKERagainst NVIDIA and against NVIDIA behaving badly.

1:02:34SPEAKER- We've got one gigantic company now controlling

1:02:37SPEAKERa frightening amount of the space.

1:02:39SPEAKER- Yeah.

1:02:40SPEAKER- These are just alternatives that might become more relevant

1:02:44SPEAKERin the future. - Yeah.

1:02:45SPEAKERSo that's, I guess that's it.

1:02:47SPEAKER- Yeah, it's a fascinating topic.

1:02:48SPEAKERIt's like, we're obviously not,

1:02:50SPEAKERneither hardware designers nor programmers.

1:02:52SPEAKERIt's like pretty-- - Nor analysts, frankly.

1:02:54SPEAKER- Heady stuff here.

1:02:55SPEAKERAnd we're just kind of trying to make sense of it

1:02:57SPEAKERas best we can, but I think it's fascinating.

1:02:59SPEAKER- Look, man, it's a confusing world.

1:03:01SPEAKERWe just want to figure out what's going on out there.

1:03:03SPEAKER- Yeah.

1:03:06SPEAKER- We've reached the portion of the show

1:03:07SPEAKERwhere we thank our patrons.

1:03:09SPEAKER- Yeah.

1:03:10SPEAKER- The currently 1,090 people.

1:03:13SPEAKER- Jeez.

1:03:14SPEAKER- Like what we do enough to give us money

1:03:17SPEAKERor want to get access to the Discord or want to hear

1:03:19SPEAKERthe patron exclusive episode every month.

1:03:21SPEAKER- Yes, yes.

1:03:22SPEAKEROr just want us to continue to continue to keep doing this.

1:03:25SPEAKER- Yeah, you can find out how to support the Patreon

1:03:28SPEAKERat patreon.com/techpod.

1:03:32SPEAKERNext week is our email episode.

1:03:33SPEAKER- Is it?

1:03:34SPEAKER- Yeah.

1:03:35SPEAKER- Okay, that's good.

1:03:36SPEAKERThat means I've got one more episodes worth of buffer

1:03:39SPEAKERbefore I have to finish that book.

1:03:41SPEAKER- Yeah. - Okay.

1:03:42SPEAKER- Yeah, you've got two weeks, maybe three,

1:03:44SPEAKERdepending on whether we do.

1:03:45SPEAKERIt's good to know what kind of timeframe

1:03:47SPEAKERyou're working with.

1:03:48SPEAKER- Look, just Sunday afternoon, Brad.

1:03:50SPEAKER- Yeah, yeah, just need to find a sunbeam

1:03:52SPEAKERand kind of relax in it.

1:03:55SPEAKER- But yeah, so we're gonna, I'll open up

1:03:57SPEAKERa Hitchhiker's Guide to the Galaxy channel

1:04:00SPEAKERin the Discord this week.

1:04:01SPEAKERSo if patrons want to discuss it,

1:04:03SPEAKERthen they will be able to.

1:04:06SPEAKERWe've had a lot of, as always, lovely conversations.

1:04:09SPEAKERI learn something new every time I dip into the Discord.

1:04:13SPEAKERAnd I learned about panorama stitching of,

1:04:20SPEAKERhow photography on Mars works the other day,

1:04:23SPEAKERwhich was amazing.

1:04:26SPEAKERJust talking about how the panorama stitched

1:04:28SPEAKERand how you do self-inspections of the wheels

1:04:30SPEAKERand the rover and all that stuff.

1:04:32SPEAKERIt was awesome, super cool.

1:04:33SPEAKER- We have someone there who works on it.

1:04:35SPEAKER- Yeah, yeah.

1:04:38SPEAKERSo if you want to find out more

1:04:41SPEAKERabout how to join the Patreon or back the show,

1:04:43SPEAKERit's two bucks to get access to the Discord,

1:04:45SPEAKERand you can find out about that at patreon.com/techpod.

1:04:50SPEAKERAs always, thank you to our executive producer-level patrons,

1:04:53SPEAKERAndrew Cotton, David Allen, and Jacob Chappell.

1:04:56SPEAKER- Yes, thank you so much.

1:04:57SPEAKER- And everybody who supports the show.

1:04:58SPEAKER- Yes, thank you, everyone.

1:04:59SPEAKER- And if you can't support the show financially,

1:05:02SPEAKERtell your friends.

1:05:02SPEAKERWe had so many good tweets last week,

1:05:04SPEAKERI had to stop retweeting them all

1:05:07SPEAKERbecause I didn't want to be spammy.

1:05:09SPEAKERBut we appreciate you.

1:05:11SPEAKER- Friendly and flattering tweets out there.

1:05:13SPEAKERI should, maybe I should start using Twitter again.

1:05:15SPEAKERI don't know.

1:05:16SPEAKER- Look, Brad, let's not get crazy here.

1:05:18SPEAKER- I don't know, man.

1:05:20SPEAKER- You know what?

1:05:23SPEAKERI have found the secret to no bummer Twitter.

1:05:25SPEAKER- Yeah?

1:05:26SPEAKER- Well, it's not no bummer, it's mostly no bummer Twitter.

1:05:29SPEAKER(laughing)

1:05:30SPEAKER- Speaking of hedging. - You turn off free tweets.

1:05:32SPEAKER- Okay.

1:05:33SPEAKER- And you do that by just going to twitter.com

1:05:35SPEAKERand typing in the mute filters area.

1:05:38SPEAKERYou do a keyword mute for RT colon,

1:05:41SPEAKERand then they go away forever.

1:05:42SPEAKER- I was going to say,

1:05:43SPEAKERI was surprised that they built that functionality in,

1:05:45SPEAKERbut it sounds like they didn't.

1:05:46SPEAKER- Oh, they don't.

1:05:47SPEAKER- They didn't, but you figured it out.

1:05:48SPEAKER- I told a Twitter engineer

1:05:49SPEAKERthat you could do this the other day by accident

1:05:51SPEAKERor an ex Twitter engineer that you could do this

1:05:53SPEAKERand I might've fucked us, but I apologize if I did.

1:05:59SPEAKERAnd then also if you filter out HTTPS colon slash slash,

1:06:03SPEAKERyou'll get rid of almost all the links.

1:06:05SPEAKERI think images still come through,

1:06:07SPEAKERbut it's nobody promoting bullshit,

1:06:09SPEAKERwhich I realized is a little bit hypocritical

1:06:10SPEAKERon my part, but whatever.

1:06:12SPEAKERAnd yeah, because the thing is people see something

1:06:16SPEAKERthat they feel like, oh man, this is important,

1:06:17SPEAKERbut I don't want to talk about this.

1:06:18SPEAKERIt's a huge bummer.

1:06:19SPEAKERI'll just smash the retweet button.

1:06:21SPEAKERI've done my job.

1:06:22SPEAKERAnd then it's just,

1:06:23SPEAKEReverybody's just retweeting shit that bums them out.

1:06:27SPEAKERSo yeah, anyway, Twitter's bad, but the tech pod is good.

1:06:31SPEAKERSo thank you all for supporting us.

1:06:33SPEAKERWe appreciate it.

1:06:34SPEAKER- We are a better Twitter.

1:06:35SPEAKER- Yes.

1:06:36SPEAKERWell, but the thing is the discord is the thing,

1:06:39SPEAKERlike I get out of the discord now,

1:06:42SPEAKERwhat I used to get out of Twitter

1:06:44SPEAKERwhen Twitter was good in like 2008,

1:06:46SPEAKERand it was just me goofing with my friends.

1:06:47SPEAKER- And in 2008, I was getting out of Twitter

1:06:49SPEAKERwhat I previously got out of IRC.

1:06:51SPEAKERSo like, whatever, this is a whole episode topic,

1:06:55SPEAKERbut I think the internet is undertaking a process

1:06:59SPEAKERof wrapping back around to smaller,

1:07:01SPEAKERmore enclosed communities again, perhaps.

1:07:03SPEAKER- Hold on, where did you get out of,

1:07:06SPEAKERwhat did you get out of IRC?

1:07:08SPEAKERDid we replace human contact with IRC in the 90s?

1:07:11SPEAKERIs that what happened?

1:07:12SPEAKER- I mean, that's very presumptuous of you to think

1:07:13SPEAKERthat I had human contact in the first place.

1:07:15SPEAKER(laughing)

1:07:19SPEAKERLook, there were not a lot of people importing

1:07:22SPEAKERweird Japanese PlayStation games

1:07:24SPEAKERin my little hodunk rural town.

1:07:26SPEAKER- That's the reason you were the anime editor,

1:07:28SPEAKERI guess, for so long.

1:07:29SPEAKER- I had to go far afield.

1:07:31SPEAKER- Fair.

1:07:32SPEAKER- To find like minds, let's say.

1:07:34SPEAKER- Fair.

1:07:36SPEAKEROkay, well, we'll see you next week.

1:07:37SPEAKERIf you have questions, send them into techpod@content.town.

1:07:40SPEAKERWe're doing questions next week,

1:07:41SPEAKERso please send us some good ones,

1:07:43SPEAKERand we will see you all next time.

1:07:46SPEAKERBye, everybody.

1:07:46SPEAKER(upbeat music)