This week, we couldn't resist talking about the newly announced Zen 3 CPUs from AMD, along with the tease of their upcoming Big Navi-based graphics cards, plus a bonus segment on that PlayStation 5 teardown. We're just a couple of big old hardware nerds at heart, what can we say? Support the Pod! Contribute to the Tech Pod Patreon and get access to our booming Discord, your name in the credits, and other great benefits! You can support the show at: https://patreon.com/techpod
This week, we couldn't resist talking about the newly announced Zen 3 CPUs from AMD, along with the tease of their upcoming Big Navi-based graphics cards, plus a bonus segment on that PlayStation 5 teardown. We're just a couple of big old hardware nerds at heart, what can we say?
Support the Pod! Contribute to the Tech Pod Patreon and get access to our booming Discord, your name in the credits, and other great benefits! You can support the show at: https://patreon.com/techpod
Links!
Brad: [00:00:00] You think you could take Alleve on an empty stomach?
Will: [00:00:02] I don't believe in Alleve.
Brad: [00:00:03] Why not? What? What's not to believe. What's not to believe about Alleve
Will: [00:00:08] it's too much of a time. It's too much of a commitment. Like if I hurt and, and it's good for what? Like, like six days or something, 12 hours, I don't know, a long time. First off, not medical professionals here. This is a technology podcast, not a medicine podcast.
Brad: [00:00:22] you're not.
Will: [00:00:23] Um, yeah, I like it. I've always been afraid to take it. Cause if I, if I hurt, I want the hurt to go away. And if it, if I can't take another one for like 12 hours, that's too long,
Brad: [00:00:34] else are you gonna make the hurt go away. If you don't take
Will: [00:00:36] something into, I take Advil. I can take those again, like four hours.
Brad: [00:00:41] the difference between the functionally? What is the difference between Advil and Aleve? They're both pills. They're both. They're both painkillers,
Will: [00:00:47] I've never taken a leaf before, but I'm taking Advil my whole life
Brad: [00:00:50] except one is an anti-inflammatory and the other one is fill in the blank. Oh, you mean INSEAD's
Will: [00:00:57] And is that how they said, how does that have the medical professional say
Brad: [00:01:00] I have, I have an Aleve in my hand. I have a cup of coffee here. It is eight 18 in the morning.
Will: [00:01:04] well, so if you take, if you drink the coffee, you won't be an empty stomach anymore. Mission accomplished.
Brad: [00:01:08] that doesn't count as food
Will: [00:01:09] Yeah. It's food. It goes in your stomach. Isn't in your tummy.
Brad: [00:01:12] Yeah,
Will: [00:01:13] Yeah. If it's in your Tumtum it's food.
Brad: [00:01:15] it's liquid and it has like no calories in it.
Right. It's
Will: [00:01:20] It has, it has oil and nutrients.
Brad: [00:01:22] Is it literally zero calories?
Will: [00:01:24] It's literally zero calories. If you don't put milk or sugar in it,
Brad: [00:01:26] Oh, yeah, no, I'm straight, straight black over
Will: [00:01:28] black like your heart.
Brad: [00:01:31] I, I messed up my neck, something fierce doing workout yesterday and it hurts a
Will: [00:01:35] Oh no.
Brad: [00:01:36] old. This socks.
Will: [00:01:37] are you working out? You got a gym in there now.
Brad: [00:01:39] I got some hand weights and that night we talked about it, that on the self-care episode a
Will: [00:01:44] Oh yeah.
Brad: [00:01:45] that app.
The seven minutes of hell.
Will: [00:01:48] I got the ring
Brad: [00:01:49] of the app.
Will: [00:01:50] and we've literally used it like three times. I didn't like the ring fed. It turns out I waited all that time
Brad: [00:01:55] You could probably offload that thing for at least most of what you paid for it.
Will: [00:01:59] they're pretty readily available now. It's not like a vendor markup situation.
Brad: [00:02:03] you just make somebody stay then.
Will: [00:02:04] we as the resistance band, the resistance band is nice for other exercises.
Brad: [00:02:08] All right, here we go. There it goes.
Will: [00:02:12] You're going to take this empty stomach, leave 12 hours. It's a commitment. 12 hours of 12 hours of no pain. Okay.
Brad: [00:02:19] I could barely move my head right now, man. Do something.
Will: [00:02:22] I, you know, is this possible it's related? It could be related to your Dumont, dual monitors. Are you, are you spending more time looking left or
Brad: [00:02:29] well, yes, but I've been doing that for the last three year or two, two
Will: [00:02:34] But are you spending more time at the desk now than you were in the past?
Brad: [00:02:37] Yes.
Will: [00:02:38] I jacked up my neck. Pretty good, because I was looking at the chat monitor too much.
And now I just ignore chat and everything's better. Sorry. Yeah.
Brad: [00:02:45] That's good. That's probably good advice. I've also been spending some time in VR for the first time in like a year.
Will: [00:02:50] You've been squatting thing. Ooh.
Brad: [00:02:54] The PlayStation version, which is a little
Will: [00:02:57] it works. Well. The, the PA piece, the PC version has problems with high frame rate displays and VR headsets. So
Brad: [00:03:05] yes. I went with the one that would just work.
Will: [00:03:07] If I turn it all the way down, like all the way down, I can get an almost playable frame rate.
Brad: [00:03:13] Huh. Okay. Well
Will: [00:03:14] then it looks like a PS two game.
Brad: [00:03:15] I chose wisely. Um, okay. Well, I think we firmly established that neither of us is a gastroenterologist, so we don't know what's going to happen with my Aleve situation, but, uh, I'm drugged up and ready to podcast.
Will: [00:03:27] Have you ever seen alien bread?
Brad: [00:03:29] let's just move on.
Will: [00:04:00] Welcome to Brandon will made a tech bot. I'm well,
Brad: [00:04:03] I'm Brad. Hello? Hi.
Will: [00:04:05] hello, you still there? Everything. Okay.
Brad: [00:04:07] yeah, I'd start. I was about to cough, but I am suppressing it right now.
Will: [00:04:10] Yeah, no, don't cough.
Brad: [00:04:11] I didn't take a cough suppressant. I just took that a leaf, but uh, we'll see how things go. That's pause. That's that's impossible.
Will: [00:04:17] home assistant told me that's the case and no assistant does not lie.
Brad: [00:04:20] is the zero particulate matter. That is not
Will: [00:04:23] No, it was one, one particular
Brad: [00:04:25] I see.
Okay. You mean the
Will: [00:04:27] an index.
Brad: [00:04:28] you mean the PM 2.5 was one or the, the ozone was one or the full aggregate was
Will: [00:04:34] you had the full aggregate, the American air quality Institute of America said you're you're we, we, they looked deep into the heart of our air and they said, Hey, this is a one dog.
Brad: [00:04:44] well, maybe the particulates were zero and the ozone was bringing that number of
Will: [00:04:47] It could be, it could be a car or, you know, it could have been some sort of methane emission in the immediate vicinity of, uh, of my house.
Brad: [00:04:54] sensor, right? Oh, well, we're in the green right now. So I think some windows are going to get opened around here after this.
Will: [00:05:01] Ooh,
Brad: [00:05:02] If let's, let's see if we'll see if we're still green in say 45 ish minutes. Anyway.
Will: [00:05:07] that cross ventilation going. You get windows open on both sides of the apartment, like blow straight through.
Brad: [00:05:11] give you that crossbar use, man. Just turn, turn this place into a wind tunnel.
Will: [00:05:14] yeah, please. Uh, what are we talking about today, Brad?
Brad: [00:05:17] Chips giblets.
Will: [00:05:19] Oh, man. I love giblets. It's almost time for Thanksgiving. We're going to make that gravy.
Brad: [00:05:23] carve them up, serve them in the sea. You get the gravy boat dumping all over the triplet. No, that's not what we're talking about.
Will: [00:05:29] Oh, you said triplet.
Oh, well that's fun
Brad: [00:05:33] central processing unit Litz
Will: [00:05:35] Ooh, I do like these. This is a cool idea.
Brad: [00:05:38] yeah. Uh, yeah, AMD announced there is in three, um, line of wealth.
I should say he's in three micro architecture, the new raisins, the new Ryzen 5,000 series based on these in three marker architecture.
Will: [00:05:52] Did they, did they name Zen based on the half-life one for the final levels of half-life one
Brad: [00:05:57] E N. Wasn't it?
Will: [00:05:58] look, it's impossible to say.
Brad: [00:06:00] I don't think it's impossible. It's a difficult and by difficult, I mean a Google search away.
Will: [00:06:08] Oh yeah. It's Z and you're absolutely right.
Brad: [00:06:11] X, E N in half-life Zen. I mean, like whatever you look at, all the, uh, look at all the, the imagery they use around the marketing of these chips. There's a lot of like, sort of Eastern looking brushstroke type animations and stuff.
Like, I think that is very much leaning into a, you know,
Will: [00:06:29] a little bit of this, a little bit of that.
Brad: [00:06:30] of, yeah. Eastern kind of Buddhist sort of styling. Here's what they've got
Will: [00:06:35] I guess, I guess the Buddhism is probably where it comes from then. Uh,
Brad: [00:06:38] I think they're saying these processors are so fast and efficient and massively parallel that you will go to a, a
Will: [00:06:46] A better
Brad: [00:06:47] state of a state of Nirvana widen when you use them.
This is what the, and to say,
Will: [00:06:50] That sounds pretty good. I could, I liked Nirvana.
Brad: [00:06:53] Yeah. Um, I'm pretty,
Will: [00:06:56] Yeah. But know what's going on.
Brad: [00:06:57] pretty excited about these I'm on, uh, go on, go on.
Will: [00:07:02] These, these look really good. They look like very competent to best in class at their price points, for sure. And even at the high end now, um, chips, assuming the benchmarks like the, the, so this is like the paper launch where they tell us what the what's inside the chips, not where they give reviewers a bunch of CPU's.
Um,
Brad: [00:07:21] barely that's coming pretty soon. Cause they're, they're launching November 5th and I saw at least one hardware site saying that their, their review samples are in the mail now.
Will: [00:07:30] Uh, yeah, so I've heard from multiple people that review samples, if they haven't, if they haven't arrived yet are on the way. And, um, they were saying, so basically the, the top level stuff is that. This is a new, a front-back redesign of the, of the Zen architecture, the triplet design that's been in, uh, the first two generations of Ryzen chips and it's, uh, running on the seven nanometer TSMC process.
And, uh, there they are look, I'm always skeptical of vendor benchmarks, but the way that they did this looks fair. If you know, I would describe it as maybe selectively chosen, but not outright cooked. Um, in terms of the benchmark numbers, this, they, they said it's 19% more instructions per clock cycle, which is yeah.
Which is an aggregate number based on, as they said, the geometric mean of 25 benchmarks.
Brad: [00:08:25] So I'm not as opposed to other types of mean, I'm not sure what makes that distinct, uh, other than being geometric, but like that, that is the big story here, because that is the one place that AMD was still lagging behind Intel, right?
Will: [00:08:38] Well, as a single core performance in gaming, they had problems in
Brad: [00:08:42] those are kind of hand in hand,
Will: [00:08:43] Yeah.
Brad: [00:08:43] Because a lot of games, a lot of games still are not making like a massive use of multiple cores. Right.
Will: [00:08:48] That's it, um, most games, even if they do use multiple cores. So like I often use pubs jesus' example, uh, even though it's not particularly CPU bound, but you, you have one main thread that needs a really fast processor. And then a lot of other stuff runs on secondary threads, like audio and physics and things like that.
Brad: [00:09:08] Right.
Will: [00:09:09] So if that main threat is slow, then it chokes the whole thing and you run it the speed of the main thread rather than, you know, you're, you're not, if you're playing most games, you're not taking advantage of six, eight, 12 CPU cores with two threads, a piece
Brad: [00:09:27] as opposed to like a Photoshop or a Maya or a, you name your media heavy productivity application, which would.
Will: [00:09:34] So Photoshops weird. Cause Photoshop doesn't isn't it's multithreaded but not like there are filters that are multithreaded, but I feel like if you're using, like, if you're just resizing your photos and stuff like that, you're probably not going to see a whole lot of multithreading. Um,
Brad: [00:09:47] I mean like a static image processing got fast years ago. That's not a huge thing.
Will: [00:09:52] yeah, but like video and code CAD stuff like that video and code is we'll use as many courses as you can throw at it. Well, and.
Brad: [00:09:58] actually like big for me with all the stuff we're doing at home now. Like I was. I was working with a 175 gigabyte pro Rez file on my ass the other day.
And I think that thing has a 6,700 K in it. And I actually needed to crunch that video down to like a usable H two, six, four, right.
Will: [00:10:15] Yeah, that's it. It's going to take a minute,
Brad: [00:10:17] to upload to people. And yeah, that took a while on that old, that five-year-old quad-core CPU.
Will: [00:10:22] right? It's time to upgrade.
Brad: [00:10:24] I of want to put one of these things in my nest now.
Will: [00:10:27] I mean, that's a really assuming they're supported properly on, I mean, look, I think when you move to Linux, that makes a lot of sense. I wonder what the BSD support for Ryzen is like yeah.
Brad: [00:10:38] my, my superficial understanding is that Linux is pretty good on AMD these days, the free, free BSD, less. So in fact, the truth also with my boy, this is going off in a million tangents. We got to use, we have to get used to saying true. NASA said a free NAS
Will: [00:10:53] Oh, right.
Brad: [00:10:53] free NASA is about a week or two away from not being a going concern anymore.
Will: [00:10:58] Oh, wow. Okay. Well,
Brad: [00:11:00] 12 is coming out any day now,
Will: [00:11:02] stayed away from the scene drama. We need to, we really need to think about spinning up that second podcast called scene drama. That's just us digging into the scene drama in different, uh, open source projects.
Brad: [00:11:10] to be trawling the discussion forums and mailing lists of open-source projects, looking for people sniping each other all day long, just dramatic, dramatic readings of people. Getting mad about library and compatibility
Will: [00:11:23] Right. If somebody is somebody upset about your,
Brad: [00:11:26] interests converging.
Will: [00:11:27] if somebody's upset about your header file formatting on an open source project, send us a note to scene drama dot content downtown. Um,
Brad: [00:11:34] I think we've talked about it before, but I should, like the early example of that is, is I've complained about this when FFmpeg pulled the NDI support out, uh, because new tech screwed around and violated their license.
Will: [00:11:46] Uh, yeah, I mean, rightly so they pulled their support.
Brad: [00:11:48] it's somebody. Yeah. High level from new tech in that thread, like the CTO or somebody, like I looked him up, it was like, this is a major figure at this company that makes the TriCaster being lambasted by like half a dozen people from the FFN big project, just like yelling at this guy.
And he's just like flailing around trying to defend what they did and like, like smooth things over. So their relationship can continue and these guys are just not having it.
Will: [00:12:13] This is a tangent and I apologize, but like the first time that it happened, it was like logic tech or somebody that did that. Like one of the early, early cases that I read about on slash in the early two thousands or late nineties even was, I want to say it was a logic tech had used some open source code in a project, and then didn't.
Didn't do the, do the diligence. Yeah. And, um, people, people got really like, I'm sure that there are still people in the Linux community. It might not have been the largest check if it is. I apologize. Um, but yeah, there, I'm sure there's still people in the open-source community. Like, fuck those guys. I'm not using their hardware ever again.
Brad: [00:12:47] grudges last, I mean, I get it, man. If you're gonna put that much work for free into something, you kind of want people to respect the work, right.
Will: [00:12:54] Yeah. Uh, so, okay. So, uh, this is third generation center. Um, if you don't know what Zen is basically it's, uh, uh, they, they looked a few years ago and said, Hey, you know, Moore's law kind of slowing down the process technology slowing down. How can we continue to make CPU as fast without relying on, you know, less power, more?
More computation, uh, that you kept from shrinking process size and they said, Oh, okay. So we're going to put, we're going to put these CPR CPU's on these triplet designs, which a helps them maintain yield so they can, they can, uh, uh, you know, build smaller, easier to manufacture chips and then jam them onto, uh, onto a die together.
Brad: [00:13:34] Yeah. As opposed to the Intel model of just one giant dive that has a record on it. And then if they get it. No think it chips off the wafer that aren't bad. The, the don't hold up, you know, that they've been poorly. You just kind of have to toss it, right.
Will: [00:13:46] Well, or they burn out, you know, they disabled big chunks of it
Brad: [00:13:50] I don't mean they literally throw in the trash, but you know what I mean?
Like it's not, it's not up to spec for the thing they would like to sell it for. Whereas like they're, they're just making, they're making the components of these chips in, or these CPU's and smaller parts that can then be combined together. Uh, which barely, I guess, I guess that had some kind of latency implications that they have improved on in this design, which we'll get to you.
Will: [00:14:09] Well, yeah, so, so basically they would put the triplets would be centered around an L three cash, uh, where they would have the compute you the, you know, the, the part that does the math, the compute unit centered around a little chunk of memory that was like eight, 10, 16 megs or gigs rather.
Brad: [00:14:23] the L three, the fastest one or the slowest one,
Will: [00:14:25] Slowest one. So the L one is the fastest, but usually one is measured in kilobytes or maybe megabytes now.
Um, and so, so these caches are where these caches are kind of like the wild things are happening on the CPU. Here's the memory that's being used. So each cache. Uh, so, so you have your system Ram and then L three cash is faster, like probably an order of magnitude faster than the system Ram. And then L two cash is faster than that.
L one catches faster than, than the L L
Brad: [00:14:55] also like with there's an inverse relationship between the speed and the size, right? The faster the cash, the smaller it is. So the less it has to work with, but, but, and also like the closer to the diet is right. Like I assume the is right there in the core.
Will: [00:15:08] So, yeah. Is so in this case, the L three cash is how the different processors, the cores talk to each other. So, uh, like in the old days, when, when Intel. Uh, started doing dual core CPU, where there were just two CPU dyes slapped onto the CPU baggage in order to talk to each other, they couldn't talk on the dye or on the package.
They had to go all the way up to the PCI express bus and into system memory, and then all the way back out. So it was slow.
Brad: [00:15:36] horribly inefficient.
Will: [00:15:37] It wasn't great. Um, we, we have learned a lot about how to design chips in that, as it turns out, not the way to do it, I'm talking across all three caches is very effective and, and, and works well.
And so the big, and it seems like the big thing in this design is that, uh, previously the triplet, uh, what is it? They call them compute cores
Brad: [00:15:58] CCI core complex, I
Will: [00:16:01] court. Yeah, the
Brad: [00:16:01] that they're using. And now that's what I mean, that's kind of a marketing or, you know, they came up with that term themselves. That's
Will: [00:16:06] Yeah. So.
Brad: [00:16:07] kind of call. Or whatever you want, but, uh, core
Will: [00:16:10] it's, it's basically like the smallest unit of the
Brad: [00:16:13] Yes.
Will: [00:16:13] that you can get. Uh, and previously those were four cores with 16 megs of cash, I think. And now each, so if you had an eight core design, you'd have two of those. And then the L three cash, four of the course could talk to one set about three cash.
The other four cores would talk to another L three
Brad: [00:16:30] yeah, so they, they put out
Will: [00:16:31] to go across the dye to yeah.
Brad: [00:16:33] sorry. They, they, they put out a slide along with this presentation that was incredibly illustrated for me. Um, and this is, this is getting into ultra nerdy hardware territory. But if you're not into that stuff or are you listening to the podcast? I guess so here. So here we go.
Um, but I had naively assumed before I looked at this slide that, uh, every triplet was its own chip, like physical chip on, on the package. Right. And that is absolutely not the case. I guess that makes sense. Cause that's just play probably. Too small of a thing to fabricate
Will: [00:17:03] Well, yeah, in some cases like on those thread rippers, they are though, like there, there will be multiple triplets on underneath the heat spreader.
Brad: [00:17:10] Okay. Well, yeah, so that, I guess it's kind of what they're doing here, but it's, it's more multiple cores per triplet is the thing it's not, it's not one core to one. Triplet is, which is what I thought it was. Does that make sense?
Will: [00:17:21] Yeah, no, no. That's yeah. You're right about that. Yeah. So the, the, the triplets are the, are centered around the L three cash. That's the, that's the thing
Brad: [00:17:28] Well, well, the, the, I think the L three cash is part of the ship. What isn't is what I'm getting from this, like a triplet quote, unquote is like mold is a number of cores and the cash in one, one die. I believe.
Will: [00:17:40] that's correct.
Brad: [00:17:42] So what they've done here is up to the number of cores per triplet. If that makes sense, right?
Excuse me, rough
Will: [00:17:48] Yeah, I th um, so yeah, they've upped the number of cores per triplet and, and changed the way that they talk to each other inside those triplets. So in the, in the last gen triplets, if you had an eight core triplet, you would get two, essentially two discrete, four core units centered around discrete. L three
Brad: [00:18:09] That's what
Will: [00:18:09] Now it's one monolithic 32 megabyte, LTE, a gigabyte Meg Pega. Right? I can't remember
Brad: [00:18:15] Yes. So
Will: [00:18:16] 32 megabyte, L three cash.
Brad: [00:18:19] it's basically the upshot is to get an eight core chip here in the past on two, you had. Two triplets that were four cores and some cash each. And now you just have one triplet that has eight cores and one giant pool of cash each.
So it's, the cores are more tightly integrated and interconnected than they used to be, which is a big part of where they're saying the performance gains are coming from.
Will: [00:18:40] So then Mike, my question is, I think what the high-end chip in this is a 12 core, sorry, sorry. Uh, yeah. 12, 16 thread, 16 core 32 threads. So that's two of these triplets. There's a 12, six core 12 thread, which I assume is one of one and a half of these
Brad: [00:18:58] no, no. So they are making they're fabricating to triplets. There's an eight core triplet and a six core triplet. So, so therefore the four CBS, they announced are six, eight, 12, and 16 cores.
Will: [00:19:08] Hmm.
Brad: [00:19:08] So for the, you know, for the six core, what does that the 5,600
Will: [00:19:12] That's the 5,600, if that could be one that doesn't been,
Brad: [00:19:17] may possibly. But, so that's what they're saying is that that one would be a one six core.
Triplet. And then the 12 core, which is, I think is the 5,900 X would be two of those six core
Will: [00:19:28] Okay. Interesting.
Brad: [00:19:29] they're saying that's way big part of where the performance gains are coming from is that these triplets just are not talking across as many bottleneck interconnects,
Will: [00:19:37] Interesting.
Brad: [00:19:39] my understanding
Will: [00:19:40] okay, so, so, I mean, look, anytime you can take out a bottleneck and let the cores talk to each other better, the multithreading performance is going to get better. That's that's the lesson of the first 10 years of multi-core 15 years of multi-core CPU design. Um, The, the, the other stuff that's changed.
I mean, it's pretty much like the, the, so double L three cash per core complex, which is good. Um, they're saying that they've made the floating point. Energy units were wider and more flexible with lower latency than they had before, which I assume is the latency is related to the not having to go across buses to talk to different cores.
Um,
Brad: [00:20:24] out a whole breakdown of like, here's where this 19% higher IPC that we're claiming is coming from. And it's like, Oh, 4% is from this. And you know, two and a half percent is from this. And it's a lot of, you know, branch prediction and like hardcore CPU design stuff that I barely understand.
Will: [00:20:39] yeah, the branch prediction stuff is always interesting to me because like the, basically what they're doing with that is they just, they, they calculate multiple potential outcomes of what the next math is going to be, that you request. So that it's just done when you ask for it, which is.
Brad: [00:20:53] You just pull up the, the correct answer from the pool of answers instead of having to do it on the spot. Yeah. That
Will: [00:20:58] Yeah. It's it's a trip.
Um,
Brad: [00:21:00] it worth stepping back real quick and talking about like, I dunno if it's getting in the weeds or if it's too basic, like the difference between like instruction set architecture and micro architecture, because I feel like a lot of people just shorthand both of those to architecture.
Will: [00:21:15] I think those are basically the same thing. Aren't they? I
Brad: [00:21:17] no, no. Like instruction set architecture is something like x86 or arm or power PC. Like that's, that's literally just the instructions that make up this type of CPU.
Will: [00:21:25] So it's like the machine, if you're writing machine code for a x86 64 computer. That's what you're writing
Brad: [00:21:31] Essentially. Yes, but it's, you know, it's even a lower level than that in that it's just the spec that includes those instructions, you know, writing the machine code is how you inter interact with the instructions, but you get what, I mean, like x86 is an instruction set architecture, and then, and then there's like x86 64 on top of that, which are extensions to that, et cetera, et cetera.
We boy, skinny pig breakfast.
Will: [00:21:50] King of the pigs.
Brad: [00:21:52] I know we talk about that. Some with the, uh, the arm. A Nvidia episode, a micro is something like Zen ones and twos and three, or like on Intel. Skylake is an example of that. Those are implementations of the instruction set
Will: [00:22:06] So one is a subset of the other, the, the, yeah, the micro architecture is a subset of the, of the instruction
Brad: [00:22:12] And I just see a lot of, a lot of people just abbreviate both of those two architecture, which kind of confuses the issue, I think,
Will: [00:22:18] I've always done that. Brad.
Brad: [00:22:19] I know I have to, but I think the reason I bring it up is it helps illustrate why these things can be faster than the previous generation, because they've gone in there and tinkered with how, like you said, how the, how brand prediction works and how things are executed in the internal, even though like it's still like 66.
Right,
Will: [00:22:34] well, and you're still running the same, like glow level. Like if you compile the application, the same code will run on this processor in last generation and an Intel processor. Um, it'll just may run better on one than the
Brad: [00:22:46] And part of the reason I bring that up is because I wanted to compare these chips to where Intel is at, which I guess we'll get to you a little later, but Intel has been very stagnant because all of their chips for the last, like five, like the flash five generations of their chips have all been on that same Skylake microarchitecture, which is getting extremely long in the tooth now.
Will: [00:23:04] Well, and also, I mean, look, this is a difference in, this is the thing that AMD has done well over the previous general, like both with this and with the original Apollon too, they looked at where the market was going and they were like, Hmm. Intel was, was building penny and fours at the time was building these really deep.
Uh, highly dependent on good prediction pipelines in their, in their micro architectures. And Amy was like, I don't think that's the right way to go. And they spun out this Athlon the Athlon a micro architecture, which was not as deep and was much faster. And they were able to clock it, clock it differently.
And, and, you know, they whipped the Lama's ass back in those
Brad: [00:23:40] yes. That let them take the performance crown for quite some time.
Will: [00:23:43] like five, four or five, three, four years. It seems like
Brad: [00:23:45] So like explaining all that as a way of saying that AMD has just been iterating on their design way faster than Intel has, because Zen two is only what, two years old year and a
Will: [00:23:53] they've been, they've been spinning new stuff about every 18 to 24 months. It seems
Brad: [00:23:57] Like they are, they are. Two times or more faster than Intel and bringing new generations of, of, of chips out.
Uh, and I think like the rumors are saying, or the scuttlebutt is that like a, is informed micro-architecture is only a couple of years out like
Will: [00:24:10] I think they explicitly said at the end of that press conference that like Zen, like the seeds are sown for Zen
Brad: [00:24:16] Right. Like they are, they are iterating very quickly on this stuff, which is pretty exciting and fun to see like competition in the CPU space is something we haven't had in a while is what I'm saying.
Will: [00:24:25] well, it's interesting cause I kind of want to build it's it's funny, like with the streaming stuff, it's like you said, it's all of a sudden people who are streaming have use for these big wide architectures with lots of CPU cores. Um, I. I know that the early Ryzen chips, which I didn't have any personal experience with, because I was not in a building PCs professionally kind of place then, but they, they like, it was an unwieldy architecture you had to buy exactly.
That might right memory. If your memory timings were off, then things would be unstable. It was, it was a little bit fiddly, but, but they've gotten in two generations assuming this lands. Okay. Yeah, the Zen two stuff was really solid and people, I know a ton of people in our audience are running it just from looking at the discord.
And it's, it's amazing how far it's come with a wild, new approach in, you know, three, four years.
Brad: [00:25:20] Yeah, pointing out these things around the same socket and for that's been around for the last, what, four or five years or something. So anybody with a basically relatively modern AMD motherboard can kind of just drop one of these in
Will: [00:25:33] Uh, I mean, you have to, you have to firmware update and all that stuff, but like that, that's the, I feel like the socket decision on Intel's part for a little while it was a, um, Like for a while that was tied to memory choices, but memory has been relatively stable for the last few years. Uh, I mean, JEDEC just announced the PC 3,200 spec, uh, like, um, a month or two ago when we've been buying 3,200 megahertz memory for, for like, I have two year old memory and my computer, this PC 30 that's 3,200.
It's just not gen X back. Um, I feel like, yeah, like it's, it's a. It's it's a good spot to be. If you're an enthusiast. I I'm so used to upgrading motherboards every time I upgrade CPU from being on Intel, the last three PCs that I don't even think about that as an option
Brad: [00:26:25] is absolutely just, uh, priced into any upgrade process. At this point, for me is just, well on me. If I get a new CPU, I'm gonna need a new motherboard too. So it's like, I don't know, it's alien to me somehow to think about just buying a new chip and putting it in your old system, but
Will: [00:26:38] Well, and at the same time, they're also supporting stuff like PCIE four, which Intel just got this generation right. With the
Brad: [00:26:45] Oh, it's no, the next
Will: [00:26:46] Oh, it's next one. Yeah.
Brad: [00:26:48] the rocket Lake we'll we'll talk about it. So they, they kind of Intel attempted to kind of counter programming and be a little bit earlier this week with a medium post
Will: [00:26:55] Oh yeah. That's always the, you know, things are going well.
Brad: [00:26:58] they announced that, uh, Or they kind of soft announced that rocket Lake, which is the next iteration is coming in Q1 next year, but they didn't say much about what it is.
Like they're still on 14 nanometer with even these current ships that they've got out and people are just assuming from the name, this will be another iteration of Skylake, like the sixth one in a row or something, fifth one. So like their Intel has kind of been on the back foot for quite a while.
Will: [00:27:23] I mean, usually what happens with Intel, it gets this in the weeds and behind is all of a sudden everything's, everything's bad for a while. And then the drop something that's either really, really good or really, really bad. So like I Tanium came out of one of these cycles, uh, but also the current, the, the, what became.
And inhale them came out of one of these cycles to nail them is basically what has, has begat everything all the way through the 10 generations with Skylake or whatever we're on
Brad: [00:27:51] like that kicked off pretty much like the whole core line. Right. Uh, which obviously was incredibly dominant for a decade. But, um, Intel's in a weird spot. I earlier this week, I won't really get into the details, but for like budgetary and time reasons, I was looking at a quick and dirty CPU motherboard upgrade just, it was just a situation of like, well, this purchase might happen.
If you get it in, in the next six hours. And I was like, well, that's not enough time to research like AMD memory requirements and figure out what I would want to do there. So let's just go with what I know and just throw the current Intel stuff on there. So, so I have a 7,700 K right now when I was like, well, it's the ten seven hundred K is the current equivalent of that.
Is that every saying that ten seven hundred K
Will: [00:28:31] 10.7. I don't know. It's a bit unwieldy. Naming
Brad: [00:28:35] an unwieldy name. Um, I started looking at the ten seven hundred K and. Hey, the clock speed is slightly lower than the 7,700 K I already have by like a hundred megahertz or something. Be the per core performance. Best I could tell is like basically unchanged.
Will: [00:28:52] It's like 6% better or something
Brad: [00:28:53] barely faster. The only difference is that it's an eight core chip instead of a four core. So
Will: [00:28:58] Well that's nice,
Brad: [00:28:59] I would get some performance improvements
Will: [00:29:01] but probably not in the games.
Brad: [00:29:03] not in games. Like I was looking at benchmarks between the current one I have and this chip for gender three generations later.
And. The game performance benchmarks were like basically identical. They were even even faster in like one case. They were actually faster on my current ship and that 10, 700 K has like super high power requirements. It's like a 125 watt TDP. And, uh, and I was just like, well, this is the worst possible time to be buying a CPU.
And I think I'm just not going to bother. So I ended up not putting that in the budget order. Because like, what's the point, you know, like it's so much hassle to pull your motherboard out of your CPE or you're out of your case and like rebuild the whole thing
Will: [00:29:40] It's funny.
Brad: [00:29:41] improvement. I was just like, what's the point?
Will: [00:29:43] I mean, the last time I did a CPU upgrade was last year. And, or maybe earlier this, I don't, I dunno, it was, it was the before times and it was mainly like, I got a new CPU because I'd done some, I played in a video game tournament. Um, But I had gotten by a motherboard. And the thing I was excited about was having a better USB three implementation on that motherboard than the previous shitty chips, X 99 chips at ad
Brad: [00:30:11] That's exactly how I felt in the brief window. I was looking at doing this. I was more excited about seeing what I could get on a brand new motherboard in terms of features.
Will: [00:30:18] more PCI express lanes on the chip set and, and more better USB. Anyway. Um,
Brad: [00:30:24] Intel CPU's are not interesting in the slightest right now.
Will: [00:30:27] No. Uh, so, okay. So I think let's like they, they did benchmarks when, when the benchmarks for these land at, at your, uh, Nan techs and your PC worlds and all those places, we'll talk about it in the discord. I'm sure. Um, they're pitching like 28% improvement in shadow of the tomb Raider, which is at 19 by 10, uh, gained, you know, 10 80 PhD, which is.
It has to be with all the hair stuff turned on. Um, and it must be set up to be as CPU bound, as possible to see
Brad: [00:30:58] I see. Yes, I see that you wrote here in the notes must be really CPU bound, which I got to answer to the question I was going to ask, which was a Trudeau seem like shadow of the tomb. Raider gets used an awful lot for benchmarks by everybody.
Will: [00:31:09] So AMD likes shadow of the tomb Raider, because they had it's part of their co-marketing thing. Uh, yeah, they like, it's a, it's a, you know, the, you know, Nvidia has he used to have the way it's meant to be played? AMD has had a thing that said like 10 different names,
Brad: [00:31:23] totally. But the reason I ask is that, uh, it was also an early RTX showcase for India. So
Will: [00:31:29] right. Oh, right. It was the other, it was the first tomb Raider. That was an AMD game.
Brad: [00:31:33] yes. Was AMD hair works.
Will: [00:31:35] Hair works was a tress effects was AMD hair works. His
Brad: [00:31:38] works with
Will: [00:31:39] hair works is Nvidia.
Brad: [00:31:41] Tressa trespass versus hair works. Choose your, choose your hair. Middleware, middle, middle, hate middle hair,
Will: [00:31:46] title. Um,
Brad: [00:31:48] I'm attempting to workshop that one for a little while.
Will: [00:31:50] look, I just, I'm just glad that they're doing braids instead of trying to do AAA chair.
That's all I care about. Um, I, I like, I'm always skeptical of vendor provided benchmarks for pretty much
Brad: [00:32:01] Yeah, I read the non-tech write-up of this announcement and they did say that in the past day of AMD has been quite honest with their numbers overall.
Will: [00:32:09] I was going to say, like, I think, I think it's safe given honestly, almost everybody has been pretty good faith with numbers over the last three to five years. There's people vendors have realized that there's no benefit to cooking numbers because of the people who run benchmarks on parts. When they get out in the wild are really good at it.
And like the blowback on cooking your, your release numbers is such that it's just, it's easier to just make the chip good than to try to make it look good with bad benchmarks. So, um, I think it's, I think it's, I think it's safe to assume that that like, especially the single core performance of this is going to see a pretty significant boost over Zen, too.
Brad: [00:32:48] say 19% improvement of resent to, for the IPC instructions per clock, which is the single core performance. And like in crucially, they claim that is significantly faster than Intel's best offering right now. So like, This is essentially them saying we have the fastest CPU on the market now in every possible use case,
Will: [00:33:05] Yeah. And, and, and, and that's that wasn't. The fastest CPU in each of the price points that they're releasing CPU that's, they're saying the CPU is faster for normal consumer business than, than what Intel has at the, at the different price points. So, um, let's run down to the parts. I w we there's, if we have a little bit of time at the end, I'd love to talk about, uh, the PS five stuff
Brad: [00:33:27] Although the tear downs.
Will: [00:33:28] the tear down.
Brad: [00:33:29] such a weird thing.
Will: [00:33:31] What a, what an unbelievable, well, we'll get to it. Okay. Um, so for, for chips, for CPU's coming with Zen for Zen three, rather, um, the, at the low end for 300 bucks, you got the Ryzen nine 5,600 X. It's a six core 12 thread.
Brad: [00:33:47] hold up. Is that it? I think it's a Ryzen five. Isn't that right?
Will: [00:33:51] Did I say Ryzen nine? I wrote Ryzen nine on all of these.
Brad: [00:33:54] Yeah, I believe it's swimming and hang on. I've got the chart here. I know it's confusing. And there's too many numbers.
Will: [00:33:59] Well, I mean, the important thing is it's a 5,600 X there's only one 5,600.
Brad: [00:34:03] yes. It's especially confusing because they put the Zen three name out there, but the previous is in two chips were Ryzen 3000 series.
Will: [00:34:11] they skipped
Brad: [00:34:12] a lot of, yeah, well Ryzen 4,000 is a mobile part, I
Will: [00:34:16] or AP their AP use. Yeah. So.
Brad: [00:34:18] Yes. So.
Will: [00:34:19] Uh, APU is our, the, the GPU, and it's like an SOC for, uh, for a month for a PC. Um,
Brad: [00:34:26] upshot is there, there's a lot of confusion in the numbering between the various parts of their
Will: [00:34:30] well, so I think Ian at an Antech, um, pointed out that if by skipping 4,000, they have an opportunity to unify all the Zen three stuff on the 5,000 line AP use NCPS, which is good.
Brad: [00:34:43] Yes. Well, yeah. So yes, Ryzen five 5,600 X is the name of that low end
Will: [00:34:47] well, I might put nine on all of these,
Brad: [00:34:49] two 99.
Will: [00:34:50] um, yeah,
Brad: [00:34:51] I think it's, I think, I think that is that's just to match Intel on the core. I five I 709 naming rate. I think that's, they're just piggybacking off of that
Will: [00:34:58] Oh, is that it?
Brad: [00:34:59] I believe that's the case. Yeah. It's just purely a marketing thing.
So like at Ryzen five, you can assume is kind of the equivalent of a, of a core I five
Will: [00:35:07] Interesting. So, okay. So yeah, a 4.6 gigahertz, 3.7 gigahertz, uh, boost and, and base clock. Uh, so the boots clock is what you get when you have a single or dual threaded application. It's like, it's like, um, Intel's turbo boost, uh, stuff, which does the same thing,
Brad: [00:35:25] that was one of the questions I had written down here was if you couldn't help me understand exactly what the boost clock or how the boost clock works or, or the turbo boost or whatever, and telecalls it as well, like is,
Will: [00:35:36] Yeah,
Brad: [00:35:36] you? Right. So basically if you're in a single threaded situation, they'll boost the frequency because they're not having to run all the cores at the same
Will: [00:35:42] it's, it's not just even single-threaded it's that the chip is aware of what the thermals are on the processor and what it's capable of. So if you have like a, if you have a 12 core chip and you're using four cores heavily, it'll spike the clock on the four cores that are being used and turned down the other ones.
Brad: [00:35:59] it can,
Will: [00:36:00] Balance out the TDP, this, the power consumption on these on modern CPU's as well. Monitor you brought it in chips in general is bonkers. Like they can turn off the stuff that's not being used and turn up the stuff that is being used.
Brad: [00:36:11] Crazy. I would not have thought they'd be able to run the, the individual cores at different clock rates.
Will: [00:36:15] Yep. I mean, your Intel chip to your 7,700 does that too.
Brad: [00:36:19] Right. I just didn't know that I, I assumed they all had to run in the same frequency just to kind of stay in sync or whatever, but
Will: [00:36:24] Well, see, I mean, you can choose to do that if you want. And there's performance benefits to doing that as if you can handle cooling it. But, but I mean, the upshot is, uh, you'll see faster perf uh, on single threaded apps at that boost clock.
Brad: [00:36:38] Basically the fewer cores that are being worked, the faster those cores can run is what you're saying.
Will: [00:36:42] Exactly.
Brad: [00:36:43] Okay. So in that, and is that sustainable for as long as, as necessary?
Will: [00:36:47] be, yeah.
Brad: [00:36:47] So, so
Will: [00:36:48] Assuming your cooling's right.
Brad: [00:36:49] what you're saying is that is incredibly useful for games that are not using all the course.
Will: [00:36:54] like you could have called this game mode for all intents and purposes, because yeah.
Brad: [00:36:58] exactly what this is. That's okay.
Will: [00:37:00] Yeah. So if you have a, if you have an unreal engine game that uses two threads or four threads and is hitting two cores really hard, the other cores will slow way, way down. And it'll crank those up as high as 4.6 gigahertz in this case.
And usually you can adjust that and make it even go faster. If you want in the bios, if you have the cooling discipline, like if you put a water cooler on it, you can boost a little, like, for example, On my I 9,900 K I can. The boost clock by default is I think two cores at five gigahertz. I can set all of them because I'm raising a water cooler and it'll run at five gigahertz or all of them, but one maybe, um, anyway, uh, 35 mega megawatt cash on this one and a 65 watt TDP.
Uh, so the TDP is the thermal design profile AMD and Intel measure them a little bit differently, I believe. But the, the, um, You know, this isn't going to make, this is a relatively low wattage part. Uh, and, uh, this is also the only one of these chips that comes with a cooler there. They, they specifically said like market data shows that people who are buying $500 CPS don't use the cooler that we provide.
So rather than just have somebody have some stuff that you have to Chuck in a drawer or send to the landfill, they're just not going to put it in the box anymore. You can buy your own.
Brad: [00:38:17] Yep. That's smart.
Will: [00:38:18] Um, the Ryzen is this a Ryzen seven 5,800 X is an eight core 16 thread at 3.8 gigahertz base clock 4.7 gigahertz boost, so ever so slightly faster, but the big jump is two additional cores and four additional threads.
Uh, another 30, 36 megabyte cache. This one's a lot hotter, 105 watt TDP, uh, and it's 450 bucks.
Brad: [00:38:43] Yeah. That's, that's like, that's a little toasty, I guess. Like my 7,700 K is like a 90 something. What I believe TDP. So it's kind of in the ballpark, but like I said, these, these current eight core Intel ones are like 125 and more.
Will: [00:38:56] Uh, around a hundred is kind of like the mid range, like where you are and like, this is you can air cool. Any of these, it seems like a hundred and 105 waters and easy. Like one of those cooler master Nakisha, uh, air coolers. We'll do one of those big towers will do well, but you can, you can also put water on it.
Brad: [00:39:13] I want to talk to her next time. I do a build very excited.
Will: [00:39:16] it's funny, I'd forgotten. I put one of those cooler master, uh, you know, the El cheapos on my stream machine and I opened it up to do some work the other day and was just really pleasantly surprised at how UNGASS it's gotten. Uh, I didn't even have to blast it out. Uh, anyway, uh, it's been doing good work for a few years and then the 5,900 ex uh, Ryzen nine 5,900 X, 12 cores, 24 threads, 4.8 gigahertz boost 3.7 gigahertz stock, 70 megabyte, L two plus L three cash.
Um, and another 105 watt TDP at 550 bucks.
Brad: [00:39:50] this one seems like the sweet spot to me. I mean, granted, granted it's more expensive. Like if you're price conscious, you know, something like the 5,600 X is the sweet spot, but like, Bang for the buck, you
Will: [00:40:02] like if you're, if you're playing games, I'm not, I, so, I mean, just to see how these overclock, because if you can turn up the boost clocks on these than, than the 5,600 for gamers is a smoking deal. Um, I I'm, I feel like like 500, 550 bucks. This is a. I, I would assume that this is a CPU that will last you for a big chunk of the next console generation.
Brad: [00:40:28] Yeah, I think you're totally right. I do wonder though, in that context, is it important to try to have parody with what's in these consoles in terms of number of cores?
Will: [00:40:37] Matters. Having more is better than having fewer. Obviously
Brad: [00:40:41] right. Because I think that the new consoles are eight cores.
Will: [00:40:44] they are. I'm really interested to see if we ever see all eight of those cores spike on, on P on console games.
Brad: [00:40:51] That's fair. That's fair. You might need, you might be just fine with a six core CPU.
Will: [00:40:56] that's a question for our, for our friends who are like hardcore graphics developers, if you are, if you're a hardcore graphics developers, that is the question.
Send us a question to tech pod at content downtown. We would love to know if you think that you're going to be able to use eight cores in a modern console. Cause I w I'm skeptical.
Brad: [00:41:12] Even better. If your employer will allow it and you'd like to appear on the show, please let us know. We'd
Will: [00:41:16] We're not talking to one specific person here.
Brad: [00:41:19] I would love to pick the brain of, uh, any working game developers who can shed some light on this stuff. But, um, the reason I say that 5,900 X seems like sweet spot is that there's one more above that, which is the 59 50 X, which is another four core.
So it's 16 cores and 32 threads, but similar, similar clock speeds, same cache, but way more expensive. It's like another $250 to get those extra four cores on not much else.
Will: [00:41:43] so it, this seems to me. I think, I feel like, yeah, I don't know. So they said they're doing six, six core triplets.
Brad: [00:41:56] Yes,
Will: [00:41:57] I missed
Brad: [00:41:57] and eight, six and eight. There are two, two chip chip complexes. One which six cores. One with eight,
Will: [00:42:03] I wonder if these triplets are just torched out when two, two of the, of the cores don't been, I like I'm, I'm super curious about that. Somebody who, when somebody takes a look at with electron microscope they'll know.
Brad: [00:42:15] like how many, how many, uh, Chiclets are on this 16 core
Will: [00:42:19] Yeah.
Brad: [00:42:20] You mean?
Will: [00:42:20] Um, or it would be probably just with a regular microscope actually, but, um, we're we're so yeah, the 59 50 X is 16 cores up to 4.9 gigahertz boost, um, and 72 megabytes of two and all three. So
Brad: [00:42:35] one is that one is $800 compared to five 50 for the next one
Will: [00:42:38] yeah, I mean, look, there's always a high end.
Brad: [00:42:41] Yeah, it's it's the 59 50 X is sort of the, the tight, the G-Force Titan of this thing, of this situation, right? It's the, this is the best one, but you're going to pay significantly for some slightly small gains.
Will: [00:42:52] It's kind of like the 30, 80, I think. Right? Like it's like, it's like, it's expensive. You can get something that's very nearly as good for like 300 bucks less. Yeah. Anyway, these, these seem really compelling. I'm interested in what benchmarks are. I think we're going to get a guest on when they've had a time to do benchmarks and who can talk about that stuff
Brad: [00:43:10] That would be fun. I am a guy it's been five years since I was excited about CPS. And I'm excited about CPS again. So that'd be fun to talk to somebody with a hands-on experience here, but
Will: [00:43:20] Um, and this is as, as, as, uh, I think, uh, they said at the start of that press conference, this is the fifth year that the triplet designs have been in the market. So, uh, and then at the very end of that, they teased the big Navi, the RX 6,000 radio on RX 6,000 announced, which is happening in like two weeks ish
Brad: [00:43:40] uh, yeah. End of the
Will: [00:43:41] 26th, I think.
Brad: [00:43:42] Which, uh, they didn't say a lot, certainly very little in the way of numbers, but they did mention the 4k 600 basically 4k performance in borderlands three and the numbers they put out there seem relatively competitive with the somewhere in the 30, 70, 30, 80 ballpark.
Will: [00:43:58] and I think, look, AMD's graphics cards have played low end and mid range, firmly low range, the last few generations. So I, I. Like they're, they're writing the, the fat part of the curve is that seems to be their GPU strategy.
Brad: [00:44:14] price to performance kind of out value rather than like hitting the top end of the performance.
Will: [00:44:19] yeah. They're building stuff for the places that people spend money versus the places where people want to spend money, but don't so,
Brad: [00:44:25] Uh, but I mean, I'm sure that, you know, they are riding the wave of all the R and D they put into these new consoles, right?
Will: [00:44:31] I, you know, it's, that's a good question. I don't know if they're allowed to, like, I know that I know that when they build these console, GPU's those teams are really siloed off from the rest of the company. So my understanding is that at least in the, in the beginning, when they started this relationship with the Xbox three 60, I guess
Brad: [00:44:49] yes. Yes. Original original X-Box was in video. And then they went AMD from there.
Will: [00:44:54] Yeah. So IP moved from the PC slash radio on team to the Xbox three 60, but not back the other way.
Brad: [00:45:02] my guess is that this would come down to whether they are licensing that architecture to the console makers versus whether they sold them, the IP for it. Right.
Will: [00:45:09] I'm sure are licensing.
Brad: [00:45:11] Also in that case, I mean, they've said the consoles are our DNA to based, and I assume these graphics, the big Navi is as well. Right. I think I've said as much.
Will: [00:45:18] I guess. That's true. Yeah.
Brad: [00:45:19] gotta be at least some overlap in any way at any rate. They've certainly learned a lot in working on these new
Will: [00:45:24] Well, yeah. So, so think if you're thinking about it in terms of like open-source projects, the PS4 five and the Xbox one elite, is that that's it X-Box elite
Brad: [00:45:35] X-Box series X.
Will: [00:45:37] X-Box series ax. Yeah, there you go. Whatever it's called the cube and the, and the fins.
Brad: [00:45:42] expert naming
Will: [00:45:43] Yeah, like, look, I bet fucking do this professionally. I don't know
Brad: [00:45:46] paying dividends here.
Will: [00:45:48] Um, uh, they're they're forks. So they forked the Zen with the architecture that they were using for the GPS at some 0.2, three years in the past five years in the past. And now they've diverge from there.
Brad: [00:46:02] this is totally my assumption. I have no like background to, to make this claim, but I would guess that any custom work they did for these console makers that was requested by them can't necessarily be rolled back in,
Will: [00:46:12] That is.
Brad: [00:46:13] the basis of the GPU architecture that they provided to them, they can use, I would
Will: [00:46:17] Yeah, that's fine. Assumptions. So speaking of the console architectures, uh, Sony released something that was unprecedented, I think in there.
Brad: [00:46:27] them.
Will: [00:46:27] Yeah. Like they, they had the lead architect for the PS five, sit down in a white room and take apart a PS five that wasn't, it looked like a production one. It didn't look like it had paste thermal paste and everything on it.
It wasn't like,
Brad: [00:46:42] Yeah. I think it's the final one. He was from he's from the mechanical design department at Sony, which is like the most like Japanese corporate name for a department. I feel like
Will: [00:46:51] Have you, have you ever heard, have you, you ever heard stories about like what? So when we were at CS, Wasn't years ago, 10 years ago. We wanted to go see an early HMD, like video HMD that Sony had. And they had had it set up, uh, in our special room. When I talked to the Sony PR rep, they were like, yeah, we didn't.
They just told us Japan just told us to block off a 20 by 30 foot room with eight stalls. And that's where we were going to do. They're going to bring some hardware. And some guy flew from Japan with like a case chain to him that was in a seat next to him. And, and like, you know, like mission impossible rolled in two days before the thing and set it up and like none of the U S people, none of the people running CBS had seen what was going to go in there at any idea what was going into that part of the booth.
Brad: [00:47:39] That that is very classically Japanese parent company. In my understanding, especially as has changed somewhat in recent years, but back in the day, very much like the, the, the, the home company in Japan called all the shots, they made all the decisions
Will: [00:47:51] Oh, yeah.
Brad: [00:47:52] the international divisions were just there to facilitate what they were doing in other territories.
And you know, that, that's why it was not really getting off tangent here. That's why it was such a big story when Howard stringer took over Sony.
Will: [00:48:03] Oh yeah.
Brad: [00:48:03] it was the first time someone not from Japan was running Sony for some period of time. Uh, and like now, you know, now you've seen them, um, you've seen them seed some actual authority to, uh, other parts of the PlayStation business in different territories.
You know, like that's kind of, it's largely being run out of Europe now. So it's changed much more. Like the PlayStation branch of Sony is a much more international operation now, but
Will: [00:48:25] it was, and it seems like the product is better because everybody's talking to each other, at least on occasion.
Brad: [00:48:31] you know, different regions have different skillsets, you know, like Japan is very good at hardware and product design. Like, you know, you've maybe maybe better like network engineering and UI software stuff in America or Europe, just as an, as a
Will: [00:48:43] Services, not Japan, please. Yeah.
Brad: [00:48:46] yeah. So, uh, anyway, this video is weird to me cause like, I don't know.
It's cool that they did it. It's just.
Will: [00:48:55] it's very unsafe.
Brad: [00:48:57] Yes, that's kind of what I'm getting at. Like it's just so unlike them to, like, they're, they've typically been very secretive about their products. So to just have them taking apart their own box, like a month before launch pretty, pretty unexpected.
Will: [00:49:08] I mean, and there's a lot, you can glean from this. Like there's a lot to see in here cause they don't, they don't usually when, when they, when they have done something like this in the past, they have kind of obscured the really interesting bits, especially if it was before launch, I would have assumed.
And like, it's all there. You can stop, you can freeze, frame this video and go through it. And I, I spent some time doing that yesterday. Um, I also want to shout out, uh, so it was Yasuhiro Tori I think is the, is the engineer's name? Um, His screwdriver collection is exquisite.
Brad: [00:49:39] Yeah,
Will: [00:49:40] has this delightful, full wouldn't handle, the like Phillips head.
And then there's a prior that I think is great. Like a little bog standard eyeglass screwdriver. And I don't know what the orange handle, uh, what the two mid-size handles are in there. But.
Brad: [00:49:56] he's got these guys that is the widest Flathead screwdriver I have ever seen. I don't even know if that qualifies as a Flathead at this
Will: [00:50:02] So that's a prior that's
Brad: [00:50:03] that also the is also well
Will: [00:50:05] the
Brad: [00:50:05] also uprising. Is that basically similar to a sponger ASU,
Will: [00:50:09] Well, so a sponger is usually nylon or something, so it doesn't Mark up the finish. I assume that's to get into a specific spot on this and maybe pop off the big side fins. I don't think they actually showed it him using that in the video.
Um,
Brad: [00:50:22] When you take the screws out to take the base off, you can store the screws in the base,
Will: [00:50:26] Oh, um,
Brad: [00:50:27] wild, little detail.
Will: [00:50:29] look, I read, I read, I read the polygon article about that. Um, you don't want to lose the screws, the support calls, man.
Brad: [00:50:36] Yeah, God, you can. You rotate the, the, the inner compartment of the stand to obscure the screw once you store it in there. So it doesn't get out.
Will: [00:50:43] what a delight? Um,
Brad: [00:50:46] I hate that. I hate the way this thing looks, but I am pretty impressed with the engineering and thought that went into the design of it so far.
Will: [00:50:52] Well, so the, the thing, the thing to me about this whole design is it seems like, whereas with the PS4, they were designing it, like they built the hardware and they were like, wait, we have to move how much air through this? And then just put a huge loud fucking fan on it. This thing seems like it was designed from the get, go to have good air flow in and out and blow from front to back.
Brad: [00:51:13] Yeah. Like people, you know, people are laughing, maybe somewhat justifiably about how giant Ganek this thing is. And it is. By all appearances, the biggest console ever made.
Will: [00:51:23] Look, I'm going to put it on my entertainment center and never look at it again,
Brad: [00:51:27] as long as you have room for it, it's fine. But I mean, that's the thing I was going to say is that Sony has always been obsessed with these like sleek, small compact designs at the expense of cooling and noise.
And honestly, yes, I will absolutely take a gigantic box if it is silent over something that is too tightly integrated and just roars like a jet engine.
Will: [00:51:45] please, please make it quiet. I love that the stand also is like, It works in both horizontal and vertical position. So that in two years when I'm like where the, I want to make it vertical, whereas my stand, I can find it.
Brad: [00:51:58] well, I mean, the fact that you need a horizontal stand is maybe a little suspect. Like I said, I hate how lopsided this thing is, but, you know,
Will: [00:52:05] I see. I really like the way it looks. I wish it was black. I feel like this is white is going to get gross really quickly,
Brad: [00:52:11] maybe, so that was actually one of the big rev fun revelations for me of this video. Is that the panels, the fins, the big white. Size, uh, a come off very easily, be up here to just be, you know, injection mold, plastic or whatever, which means that, you know, within two or three months of release, the aftermarket is going to be cranking out replacement panels for this thing.
Will: [00:52:31] Oh, yeah, we'll get all sorts of horrible stuff.
Brad: [00:52:33] you're going to be able to get your wood paneling on if you want or whatever, like whatever. I just want a smaller and the black set of side panels.
Will: [00:52:42] wow. The dust catcher. I hadn't noticed. That's interesting.
Brad: [00:52:45] that's cool.
Will: [00:52:45] Um, I want, I want, um, I want like VIN diesel, unlike an airbrushed van painting for my,
Brad: [00:52:53] I see you have one a half, whatever you want. You get gigantic weed Leafs leaves on the
Will: [00:53:00] on one side and a bunch of weed leaves on the other.
Brad: [00:53:02] That's. That can be, that can be yours.
Will: [00:53:04] they showed peeling off the security sticker. Um, this fan design is really neat. Like the fact that it just lifts out is pretty rad. I guess we
Brad: [00:53:14] a much, much, much higher volume fan than in the past or much, a much larger fan. So like, you know, the, the bigger the fan, the choir is
Will: [00:53:21] well, and it's, and it's, I don't, I don't remember the name of it, but it's like a water wheel type fan, not a, not a blade fan. So, um, it, it, it, it works a little bit differently and it's tough to suck in the flat side and blow out the sides or maybe the other way around. Um, and he, he specifically says airflow goes front to back. Um, and then when you open this thing up, basically the, the lid underneath it is almost all like there's so much surface area on the heat sinks in this thing.
Brad: [00:53:54] Yeah. Uh, we, we, we should mention the, they, they, they show off the, the uh, expansion slot for, for drives, which is like, it's, it's relatively accessible, right? I mean, you do have to take the, the big white plastic panel off, and then there's a smaller kind of metal. Cover over that slot that you have to
Will: [00:54:13] It seems precisely as accessible as the PS4, his hard drive slot was.
Brad: [00:54:19] uh, yeah.
Um,
Will: [00:54:22] I mean, honestly, I'm just excited that it actually supports an SSD. This go round.
Brad: [00:54:27] yeah.
Will: [00:54:28] so disappointed when we opened up the PS4 put an SSD in and then it did nothing because it had an old satisfac, um, Like th like looking at how this board comes out, it is so designed to be maintained.
Um, in that, like all the screws are accessible from one side. There's no daughter boards. There's only a handful of, of floppy wire connectors, uh, ribbon connectors that go to the front panel and the, and the other stuff. And then all the back panel connectors are just plainer with the board. Um, the CPU core is, is, is a, um, there's no heat spreader on it.
It's just a raw chip design.
Brad: [00:55:06] the dye is just exposed.
Will: [00:55:08] Uh, they, one of the interesting thing is when they, when he pulls the board out of the back and I'm going to go back and freeze frame for a sec, you can see there's a little heat pipe on the back, back on the, on the shield, on the back of it. Um, that goes, it looks like over things like the SSD and this and the system memory.
Uh, and there's a couple of little, uh, he sinks on that, but there's not, there's a bunch of stuff on the backside of this board. And then there's not a lot of, um, Uh, actual cooling back there. So I'm interested to see how that works. Like the system memory doesn't seem to have thermal paste on it. It doesn't seem to have coolers on it.
So either it's being cooled from the other side somehow. Uh, I I'm going to be interesting. Yeah, that
Brad: [00:55:47] Yeah. Yeah. Is there any benefit to having the memory on the opposite side from the CPU? I guess not, right.
Will: [00:55:54] I mean, you could, they could put it wherever they wanted. So I assume that there's some benefit or they would have done it on the other side. Cause it's it's I mean, it's, uh, I don't, I'm, I'm interested to see.
Brad: [00:56:04] It could just be a concession to cooling that they needed the space.
Will: [00:56:07] well, yeah, or that the other side, the, yeah, exactly. That they, they were worried that they could have been worried about anything that would cause them to do that.
I don't know why they would do that. Um, but once you pull the board out, so, and you know what, there is thermal paste on the, on the memory
Brad: [00:56:21] yeah, I was going to say that there is, there is a close up shot where you can see that there's a little bit of smear on there.
Will: [00:56:26] Yeah. And then they also have some goo on the, on the voltage regulators and stuff like that off on the side. Uh, and then you can also see it on the SSD there's memory chips on the SSD kind of on the top right side of that board when he's taking the heat, the back support off of the CPU bracket. Um,
Brad: [00:56:45] I feel like a lot has been made of the liquid metal cooling thermal interface
Will: [00:56:50] so.
Brad: [00:56:51] the, uh, on the SOC,
Will: [00:56:52] Yeah, those have been, they've been shipping for
Brad: [00:56:55] Yes, it sounds, it sounds fancy, but the liquid metal, uh, thermal material has been in use for what probably the better part of 10 years.
Will: [00:57:02] We've I've there were, there was a bit of commercially available for at least five years, three, three or five, three to five years. I know gamer's nexus, I think did a one year. Like stress test of it a few years ago to see how it worked. Um, supposedly that stuff is rated for at least like the, the stuff you can buy is rated for five years.
Um, the big problem with it is that if you don't keep it contained, it can short out stuff on your motherboard, which is bad, obviously.
Brad: [00:57:32] And I think it doesn't have a, it doesn't have an issue with drying over time as well.
Will: [00:57:36] I mean, all thermal
Brad: [00:57:37] Yeah. Yeah. The regular stuff does to you, for sure. I mean, if you look around the, the SOC on this thing, they've got some kind of gasket there, some sort of seal.
Will: [00:57:44] well, that's that's to keep it from running off onto the backside of the CPU socket.
Brad: [00:57:50] Maybe that also helps with it maintaining its kind of state longer.
Will: [00:57:54] I mean, it'll keep it, keep it definitely contained to the dye and the, between the diet and the interface. Um,
Brad: [00:58:00] For some context, like where you see this in use is like, if you buy an Intel chip and you've got, you got the giant heat spreader on there. They'll use this kind of liquid metal thermal material between the actual CPU die and that heat spreader, which you're not meant to take off like people do it.
That's what the lidding is. You know, we talked to, we talked to West Finland about the wild, wet and wild process of deleting your CPU a few weeks ago.
Will: [00:58:21] It's a bad idea.
Brad: [00:58:23] And that's where you'll find stuff like the liquid metal is underneath that heat spreader, but it's not the type of stuff that your Arctic silver is providing or something like that.
You know, your kind of classic thermal paste. So this is, this is like, this is like high quality stuff, but it's not, it's not unheard of in the PC space
Will: [00:58:35] Well, especially for the direct dye to thermal. I mean, the interesting thing here is there's no heat spreader, so they're going straight from the dye to the, to the heat pipes. They're not doing a vapor chamber or anything like that, which is what most GPU's and stuff are doing these days. So presumably.
You know, they felt like there's some efficiency to doing no heat spreader. Um, and then it does have a, just a boatload of heat pipes, which there's a little bit of a phase change in them usually, um, that, that run to an enormous, just a truly, truly terrifying amount of surface area on these heat things.
Brad: [00:59:14] it is. It is huge. And I think gigantic heatsink. Is it interesting to you at all? That they. Chose to go with an internal power supply considering how big this thing already is. Like that's kind of Sony like that. I'm trying to remember the last
Will: [00:59:28] Did the PS, the PS two slim had
Brad: [00:59:29] I was, I was just about to say, I
Will: [00:59:31] PS three slim,
Brad: [00:59:33] I never had the, the last PS three slim, so it may have, but the PSU sullen is the last one.
I remember having an external power supply. You might be right about, you might be right about that last, the third model of PS three. Might've also had one.
Will: [00:59:45] the, the, so I had the second model of PS three and it, it did not, it had a, it was a normal plug.
Brad: [00:59:52] Yeah, it
Will: [00:59:52] I look, I'm a big fan of internal power supplies. The X-Box external power supply is always loud, cause it always gets dusty and gets gross underneath entertainment center.
Brad: [01:00:02] I've never heard it cause I sit far enough from the TV, but it does, it does pick up to us for sure.
Will: [01:00:07] Um,
Brad: [01:00:07] Uh, but that doesn't cause any extra thermal concerns really doesn't have having the power supply internal.
Will: [01:00:13] Not, I mean, look, if you're looking at the things that cause generate heat in there, the power supply is like a CU quartenary issue combined. The, the APU, the SSD, probably the wifi chip and, and you know, there's a bunch of other stuff in there it's going to get hotter. Um, it's interesting. Cause they cut back and forth between either a cleaned up.
Board with no thermal goo on it. Um, and then, then, uh, a stock one throughout this video, um, the SSD they're connecting to the heat pipe as well, which is good. That means it'll hopefully, hopefully last longer like this, this seems like a very serviceable machine for a
Brad: [01:00:53] Yeah, it seems, seems pretty well-made and, and again, you know, like they have, they just been, they have been so obsessed with like small and compact and tightly integrated consoles that are way too God and loud for so long. Uh, it is nice to see them go the other direction and create something that's going to be quiet and hopefully long-lasting.
Will: [01:01:09] I hope so. I mean, it remains to be seen, but, but, um, I, as far as I know, nobody's actually seen these in the real world yet.
Brad: [01:01:17] Uh, they had, um, Sony had a few, like a handful of Japanese influencers and YouTubers in to play the system and capture some footage a week or two ago. Those videos hit, I think, last weekend. So there's just a little bit of footage out there of people just kind of smiling and mugging next to a PS five and like playing some Astrobotic and some godfather.
Will: [01:01:37] It looks like, like this PSU is sealed up enough that the cockroaches won't be able to get in,
Brad: [01:01:42] Well, good.
Will: [01:01:42] which is important because that is for whatever reason, don't, isn't that a New York thing that like cockroaches loved the inside of PlayStation.
Brad: [01:01:49] I would believe it. I can tell you that ants, just those ants love the inside of cable modems.
Will: [01:01:56] you
Brad: [01:01:57] yeah,
Will: [01:01:57] know? Oh, okay. Uh, I think that's it probably for us, huh?
Brad: [01:02:01] That's just a bunch of new hardware. It's a lot of stuff here at the end of the year,
Will: [01:02:05] It turns out I thought we were gonna have, uh, yeah.
Brad: [01:02:07] new new video cards from both AMD and Nvidia. There's just a giant stack of hardware to talk about,
Will: [01:02:14] Yeah. I, um, I wasn't, I was kind of, I wondered what the, well, in next week, iPhones, presumably.
Brad: [01:02:20] right? Yeah.
Will: [01:02:21] Yeah, I was, I was kind of expecting a quiet year, given the Rona and the manufacturing situation, but it seems like everybody's ramped back up and is working full speed again.
Brad: [01:02:30] man, the, uh, the investors wait for no one,
Will: [01:02:34] Foxconn Foxconn doesn't sleep right?
Brad: [01:02:36] shareholder and other shareholders don't sleep.
Will: [01:02:38] Well, yeah,
Brad: [01:02:39] shareholders expect product cycles to be met and earnings to be generated and revenue targets to be sought. Yes.
Will: [01:02:49] Uh,
Brad: [01:02:50] yeah.
Will: [01:02:51] Yeah, but, uh, so this is the part of the show where we think our patrons, um, as always, uh, one of my, the best part of my morning is getting up and firing up the, uh, tech pod discord and finding out, uh, what people are talking about this morning, this morning. It was well, there's a bunch of look, we don't need to get into the specifics cause there's a bunch of inside baseball stuff in there.
Um, apparently I just found out that I'm on penny arcade today from our discard. Uh, well, Gary and I had been playing golf with the penny arcade guys on the rag.
Brad: [01:03:24] Wait. We're like, like in an illustrated form. Did they put you in a
Will: [01:03:27] They did not draw me, which is the disappointment. They just use my name. They called me the budget will Smith, Brad.
Brad: [01:03:34] Oh, okay.
Will: [01:03:36] I don't know, I feel like I'm being watched right now,
Brad: [01:03:39] going to, I'm just gonna leave that one alone
Will: [01:03:40] yeah, just walk away.
Um, but yeah, if you want to get, uh, get on the Patrion discord, the Tecplot discord, you can by backing us at the $2 level. Um, and you can do that by going to patrion.com/tech pod. Uh, your money is used to support the show and, and, uh, you know, keeps, it, keeps this thing going.
Brad: [01:04:00] Yeah. Y'all are y'all are the reason we're still doing this?
Will: [01:04:02] Yeah, we started out doing this just because we wanted to have an excuse to talk.
And now it turns out it's Brad is one of the only human faces I see on the day-to-day outside of my immediate family. So. Um, but yeah, thanks as always to our executive producer level patrons, Andrew Cotton, David Allen and Jacob chapel, as well as everyone who supports the discord and the Patrion and the podcast.
Um, and like the conversations I, I was in, I had some, some PCI express lane problems earlier this week and some problems early this weekend popped into the personal computers and. It's it's personal computers and other maddening devices channel, and people gave some real good advice that was like, Hey, don't just get rid of the second M two card in your computer and just use a SATA SSD because for what you're using your port, it's going to be fine.
And then my problem went away because instead of obsessing over small details, I just, I just moved on and, and swap some hardware around and everything's better now.
Brad: [01:05:00] Generally asking a question on that server and then doing what the consensus says to do is a pretty safe bet. I have found
Will: [01:05:06] Well, and, and you learn about other things as well. The reading room, which we added after last week's Hitchhiker's guide episode was, is, has been very popular and I'm, I've got a giant list of books to read now. So thanks everyone. Um, What if you want to find out more about the Patrion you can, by going to patrion.com/tech, uh, we'll do emails in a couple of weeks, probably first week in November, given that there's hardware coming at the end of the month that we might want to talk about.
Uh, but you can find out about that. Uh, you can send emails in to tech pod@content.to town, or if you just want to let people know that the podcast is awesome, send them to us pod.content.town. This is all very simple and straightforward. It's a perfectly chromium naming scheme, right? Uh, I guess that'll do it for us,
Brad: [01:05:54] What do you, where's the, what's our outro. What he got, man, send us.
Will: [01:05:59] I guess. I'll see you next.
Brad: [01:06:00] This is the danger of the early am podcast. At some point, the brain might just short circuit.
Will: [01:06:04] Yeah. I ran out of coffee about the time we started talking about boost clocks. So
Brad: [01:06:08] to say this cup can only hold so much coffee and I haven't gotten to re up yet. So let's just, let's just boost clock on outta here. That's what I got.
Will: [01:06:16] See see you all next week. Bye everybody.