 Hello and welcome to the fourth episode of Techpad. My name is Jackson Statz, and this is my co-host, Blake Smith. Hello, and today we are going to be talking about AMD technologies with our special guest, Kyle. Hello. Just introduce yourself. Yeah, my name is Kyle Taylor. I'm just a technology enthusiast of building PCs and working with hardware. Cool. So what are we talking about tonight? So, I was quite interested on some new developments with AMD, and they invented this, I would say invented. They developed this new technology, or I don't know the much history behind it, but it was just something that caught my eye. This V-cache technology, where they did this really cool thing of, and there should be, we have some different images to bring up, about how, just by adding more cache onto their CCX, which in a CPU, a processor in your, any of your technology, any of your laptops and devices, you have things called the cores, of course, and AMD does this chiplet design, where they have multiple groups of cores, and they call those a CCX. And so each CCX has cache on it, which is like a lot of really fast storage, well, a lot relatively, it's only a couple megabytes, of really fast storage that the processor and has really easy access to. Very helpful when you're doing things like gaming and other workloads, where you want to have really quick turnarounds on different things. So like the cache is like the CPU RAM kind of thing? Yeah, I would say more like an SSD. Cool. And so by adding more of it, you are able to just have a bigger buffer for things. So as let's say you're playing Cod Warzone or something like that, and you need to get into this zone that has a really high texture input, the CPU can then take that information, store it really fast, have its access to it, and it has a much bigger buffer before it fills up and has to start throttling or lowering your frame times. And so then if you get into a much lighter area for processing, the CPU can then start lowering down and processing through that data efficiently enough and fast enough that it can drain that storage. So would you say that more cache is proportional to performance or does it have like diminishing returns? I would say proportional in the fact is like if you if you have 10 exits, you're going to get a lot more performance because you can only make it only needs really so much. But at this point, we're at the point where more is better because it can use more. But there's going to be a point where just having a a barrel of available storage is going to be too much. And there's no net. Yeah. So it's like a point of diminishing returns. Right. And we just haven't reached that yet. So the coolest thing from, of course, tech power up and all the other tech sites I was reading through was the fact that they basically to do this, they took their the CCX chip. They basically flipped it over because when you make a die, it's just laser engraving pretty much on the top, all the transistors and different things like that, they basically took it out to turn it around. This one's saying where was it? I think it was like 98 per 95 percent. The silicon is not even used because it's you don't need to make it that thin. It's hard to work with, but they take it. They take the die, flip it over, and then they just shave down different sections of it until it's pretty much just barely touching the where the top was. Yeah. And then they just add more cash onto it. They just inscribe a new like cash system in that chip, basically pretty much just on top of the actual cores. And they effectively triple the cash on the chip. So with their usually per CCX, it's and a CCX is up to eight cores. So with with AMD's largest chip for North with their rise in line, being at 16 cores that has 72 megabytes of cash. Wait, no, 74, sorry, from the 30, no, 64 megs. I wasn't saying AMD slides. I'll have to look at them. So if they have 32 megabytes of cash per CCX by adding this to add an extra 64 megabytes up to 96 megabytes of cash per CCX, it's a big boat. So they're new if they if they do continue with the same naming scheme and they go to like the 6000 series for their new stuff, then they'll be at the 16 core will jump up to 192 megabytes of cash. And that's that's a substantial amount. And it's had CPU cash two terabytes per second. That's how fast the cash is. And that's incredible. And here I can actually look up some slides of the of the let's see the 50 X. So just so the viewer knows, is this already out? I mean, is this technology already being implemented or is it for future? It's being implemented as in all there's no. Of course, there's no confirmation on anything. Yeah, these are all just until it's like right, I mean, right. Until there is well, AMD did have a slide presentation presentation in an investor's meeting talking about their technologies to, of course, get money because it's an investor meeting. Yeah, it's an investor meeting. They want the show. Hey, that we're working on stuff. But this is definitely going to be in either their 6000 series with their new sockets and everything like that. Or they might do something where they did last time with the Matisse architecture, where in a 3000 series, they released the XT chips, where those were just tuned differently, but they might release a 5000 series XT variant where it has this technology implemented, which would be substantially faster. Because from their numbers, they were looking anywhere from 4% increased frames in League of Legends, which makes sense because League of Legends is already it's so easy to run. You can really it's really hard to boost it even more because you're in the you're getting into the 400s of frames per second in these titles versus some other games that were running like 25% faster just by this increased CPU cache. So it varies differently per game, but that's jumps. You would see like generationally in the past from Intel and of an AMD in general of that around 20% increased performance just by this new processing just by this new increased cache type, which is pretty incredible that they didn't change anything with the architecture. They no shrink at all to five nanometers. We're on seven now. They just added this new. Thing that they created and it's incredible cash. So with this, where does it like put them in comparison to other companies like Intel? Well, Intel right now is they're really fighting it out. It's there with the 11th gen. They're really close, almost a margin of error. It's mostly game dependent because some games are optimized for AMD and the Vulkan API, and then some are optimized for DirectX 11 and 12, which Intel has a stronger field in. And so you have to look online and it's really a game by game basis on who wins. AMD definitely has more cores in the space. So they've got their 16 core at the 5950X, where Intel really tops out at eight cores on their 11900K. But it's still a very fast gaming processor because Intel still has the lead, even though they've still be on they've been on the 14 nanometer plus, plus, plus process. It's getting pluses on the back now. But it was actually pretty incredible when they announced the 11th gen, because after that, the announced that 12th generation Alder Lake is actually going to be released at the end of the year towards Christmas, which is quite incredible that they're kind of double generationing here this year. So I feel like anybody that's going to buy 11th gen is going to kind of just be out of their money because just for nine months later, you're going to get a whole new motherboard there. It's going to be have DDR5. There's rumors of PCIe 5.0 being on it, which is weird. But DDR5 is the biggest thing there. And then there are new processors and the 12th gen. And there are actually a lot of rumors going around that 12th gen with their new with Intel finally getting to 10 nanometers on their new Superfin design is going to be incredible because they're also implementing this thing called Big Dot Little Design, where they have really power, the powerful cores. They only have a couple of them, mostly those. And then they have a couple of power efficient cores. And that's very similar to what's happening in the mobile space for the past long time. Because quite similar to Apple's in one chip. Right. I mean, we have a bank of. Dedicated high performance cores. And then for like your web browsing, watching YouTube and Netflix, you have the super low power draw cores, right? Right. The power efficient cores. And Apple does that. Qualcomm does that for any. Yeah. The cell phone. Yeah. Yeah. You'll have your cores that can boost higher for, yeah, for booting up applications, opening things. And they can direct them down to the background task to the smaller cores for background tasks. So I mean, going back just a second to the the new cache architecture. Right. Would. You know, because because AMD manufactures server chips, their epic lineup. Right. Do you think they'll integrate this cache, you know, better in better architecture into would it be of use to have a server that could have more cash? Yeah, servers with their high processing of things. They're always shuffling through data. The huge thing in the server market is just these new fast SSDs. They're going straight there first to have those really high just transfers of getting data just in and out everywhere. And so having more cash, being able to have a bigger basically queue for the data to get through. And also this new this new cache technology is even faster than the old L3 on the on the we're using. And so that will increase the use there as well. I'm pretty sure. So we just talked about the Apple in one, how would you say if you know, I mean, it might be a little bit out of your range of knowledge. How would this compare to the Apple in one? The M one is this like supposed to compete or is it kind of a different product? I guess it's really a question. I don't know specific on that. The M one. Yeah. I can speculate saying that the M one is kind of an illegal zone because it's it's kind of hard to compare as well because AMD chips are X86. Yeah, they're different. They're different architectures. Apple has amazing power efficiency. This one go with that. But again, this is really for gaming in those high intensive workloads that would need the cash like this, where with just rendering applications, that's more that's much more just raw CPU power. Yeah. Then needing cash because and also if the technology is not out yet, then with programs that don't take advantage of it, you'd have to program it to take advantage of more CPU cash. And also, I just don't know that much about Apple Silicon. I just haven't researched that much about that. That's fine. I just wondered if you had a had a thought on that. And oh, yeah, I found the thing while I was getting confused. So it has. So the 5950 X has a 64 megabytes of L three cash. Yeah. But combine that with the eight megabytes of L two cash is where I was getting that 72 number, which is what was advertised on the slides there. I remember seeing. Yeah. So yeah, I forgot about L two. So yeah, and so it was. Just incredible to hear about that. They they I forgot I forgot there's so much silicon there on the CCX because it is just laser engraving for creating for creating the chips. And of course, there's different processes than that in chemicals and things, but it's it's extremely thin, comparatively to the actual thickness of the wafer. And the fact that they just decide to, hey, we're going to flip it over, shave it down and shove some more cash on there. And it just kind of worked. So there's better space efficiency, really. Oh, yeah, instead, because normally they increase in the X direction. Yeah. And just make it a larger chip. And if you make it a larger chip, then you have to make it a larger PCB, your sockets, which you can only get so large because motherboards are standard sizes and you don't need to have room for everything else. And then that comes into the itx space of every square millimeter is precious to shoving another component on there. Yeah. And so you want to make things as tight as you can. And then at that point, you would have to get into other technologies and things that you're missing out on based having a smaller IO die, which means you couldn't do faster PCI lanes. And then also just maybe not having a space for course. But by going in the Z direction, they can this might just also get in the engineer's head of this is actually a way we can do things to maybe make the IHS smaller and have better thermal transfer to make their chips cooler. There's also, I remember some talk of like people designing water cooling inside the chip, like. Yes, I remember my tech tips did a video on that where they just pop the IHS off and then had the water cooler, like just put thermal paste all over the die itself and it smacked a like in a laptop. I'm talking about like mini like channels on the on the actual die. So like that water through the die. That might be cool in a gaming laptop. Yeah, it was like microscopic, like kind of how they do it in not exactly how they do it in more convection based instead of pump based, where they have like an I'm trying to think of them the name like heat pipes. That's convection based where the internal structure of wicking the water away as it gets hot. That way there's no moving parts. It just does its own based on just thermodynamics and everything just and so it would be able to do something like that as well, which would be it showed amazing results. But again, a lab testing, nothing proving that we done full scale because that's that'd be so precise. That might be a I can see like if you're making Dell or Alienware, daily and where our phones, our phones just integrating this because you can that effectively means you can shrink the the the pipe, right? I mean, this just takes up less room, which is at a premium in the in smaller devices. I it's interesting idea. It was so incredible. And then also that kind of reminds me with mobile and Qualcomm. Isn't Nvidia like getting Buddy Buddy with Qualcomm for putting like AI and things into it? And since they're Qualcomm, Mari does arm. That sounds familiar. AMD is getting Buddy Buddy with Samsung, which I am in full support of because I love both those companies. But of course, whatever is faster, that's what you get. Well, more cost of faster and cost effective value. Right. I was actually talking to this one guy when I built a game PC with him and I was like, you want to get this processor from like Micro Center? They're having an amazing deal on it. Yeah. And then he was like, no, I want AMD. And I'm like, but this one's cheaper and the same speed. He's like, nope, AMD and like, there's nothing. There's no shame in going Intel. Yeah, I mean, he was just like, all my friends hate him, hate Intel or something like that. No, I mean, fanboyism kind of kills the consumer market. Like for me, fanboy, I used to be a huge anti Apple, like a Samsung and Windows fanboy. Oh, yeah, I got the S9. Got the S10. Oh, yeah. Oh, yeah. That's nine. And but it was mainly because Apple just didn't make anything good. Yeah, I got the time. They're using they're using Intel chips that were just, I think it was more Intel's faults than Apple's fault because they were innovating with making Bionics chips for their phones and stuff. Yeah, but for their computers. I mean, Intel wasn't like giving them the results they wanted because power efficiency is true. But I mean, Apple had a fair bit to play in it being a bad laptop as well. You know, like the butterfly keyboard. Oh, yeah. Where you just the keyboard would just stop working after a couple of months. And, you know, that sort of thing. I'm familiar. I'm familiar. Yeah, I did. I do have a Mac now because they fixed all the major complaints I had with it. There's still a couple like USB-C only. Oh, yeah, I'm going to buy the docks, carry the dongles. Oh, yeah. Like I have a whole bag of just dongles. Yeah. Oh, back to Alder Lake for a minute. So there was a rumor. It was a leak on Twitter by one of the leakers, Raichu, and he was quoted saying the new the 12900 KS engineering sample non-overclocked on water. So probably talking. We're talking like probably a 360 millimeter rad or something. Cinebench are 20 standard stuff. I think so much is too much aren't 20. I think that's Intel favored, though. Like I've seen a lot. I've seen like swinging results between R 22 and I think R 24 is the newest one, right? I think so. I think you're right. Yeah. So I don't know why they're using R 20 because that seems quite old compared to what we have now. And I think I remember R 20 having something like favoring Intel, like just being like more optimized kind of thing. Yeah, optimized for it. So this might they might need, of course, independent reviews. You always never trust one go around. But 28 percent faster in multi-threading than the 5950 X. And I forget it's supposed to be like a 10 core, I guess. Yeah, like, I think it was like eight big core. I don't know, six big. I don't know. The big dot little design is confused is going to confuse everyone. It's a pretty cool idea, though. Oh, yeah. I don't know why this hasn't been integrated sooner. Honestly, well, it's been in the home market, but but it's in the PC. I mean, I guess I guess the laptop. It's a great idea. Yeah. But again, well, desktop, you don't really need it, you know, right? Well, the thing I'm thinking for desktop is whether going in this way is that way they can just shove juice through the big course and keep that TDP at the same level. So they have a more power efficient ones that just kind of do a stand by and they can just get those really fast single core, which is actually here. Let me see. I mean, it might be good for streaming. Yeah, like handling. I mean, does the 16. Does gaming really, for my experience, I don't think gaming always needs a ton of like most games are optimized for two or four cores, right? I think four cores is the what most games are optimized for. But I don't know. It doesn't for me, like it never uses 100 percent of the CPU. So that specific core, I mean, if you have an older system, it probably would. But for me, I have a rise in 2600. So I would assume that you could put one of these big, little, big dot, little chips. Whatever the name is, I forgot. Rain fart. All week. Yeah. 12th term. You could have the gaming on the more power efficient cores and then have the heavy duty encoding. On the encoding is usually handled by the IGPU with use of hardware encoding. That is true. Or just the dedicated GPU is true. But if you wanted your GPU to do the focus more on gaming, I guess Nvidia has the hardware. Oh, yeah, here it is. It's supposed to be a 16 core, the 12900K with the first one to two cores out of the eight. So there's eight big, eight little with the first one to two being able to do like the 2.3 gigahertz, which again, I'm surprised that by going to 10 nm, they're still keeping above five gigahertz. Like they've been working on this for a while. They're just actually able to make it. Yeah. So they've been, they've been refining this architecture for a while. As you get smaller, can't you more easily increase in speed? Well, that's the thing. There's usually like this thing called a TikTok architecture where you go, you trade a smaller node for a decreased clock speed. And you have to kind of, and then you kind of have to mess with it a little bit to get the clock speeds back up. Or actually, TikTok's more IPC and then clock speed. You ticking that. But usually when you go to a smaller node, you have, you have to decrease clock speed and then work with it and tune it a little bit over time. But they've been working on this for years. So it's just been delay, delay. So they're still above. So 5.3 gigahertz for the first one to two cores and the rest of the eight are supposed to be at five gigahertz. And then you can tell, yeah, there's humping electricity. Oh, my gosh. And then the last the next group of one to four are going to be like 3.9 gigahertz. And then the rest of them are going to be like 3.7 at the minimum. 30 megabytes. Isn't slow. Cash. That's not slow. But still, I mean, it's not. That's much different than five gigahertz. And I honestly, it just sounds like it sounds a little bit like just rebranding what they already do. Because Intel, like it's a little up to like every not every core gets the 5.3 gigahertz on the 10, 900 K or something like KS or something like that. It's only the very, it's only the ones that can. Like they like not all not every like processor does that. But I guess if they can reduce, if they can like hardware lock some of the other ones, they could have a lot more electricity because this thing's I don't know what PL1 and PL2 means, but PL1 saying 125 Watts and then 228 Watts. So I don't know what that specifically means, but I could understand 125. I just think that maybe they do I don't know. Oh, OK, so I'm going to say 125 Watts. Watts TDP. That's definitely does top grade chip. And then you got the I7 looking at as a 12 core. And then you got the I5 as a 10 core. Yeah, I guess, which again, so a six core. And then there's like some things like where the there's going to be hyper threading on the big cores. So the the thread count is going to be whack where I think it's it's going to be called. Yeah, it's going to be called like a six core, but it's and so it's got six big four little and then those six are going to be hyper threaded. So it's going to be like 18 threads on the I5. But you're going to have to lose power because people are saying that like each little core is like a quarter of a real core, I guess. Yeah. In IPC because they're so they're made to be power efficient, not IPC, even if the clock speeds high. So again, it's really hard to talk about things when it's all rumors. Yeah. Well, we're running out of time. So any last last questions, thoughts, concerns? Well, with this preformance boost, pretty big. What do you think the if the rumors are true? What do you think the price would be? Oh, it's what they really put the price really high up. Just I think they'd just be they would probably just be on par with what AMD's price prices are now AMD. So they're just going to they're just going to swap it out. I think so. That's better. But they might do a surcharge of like 50 bucks extra. But AMD already did that and that's going to make people mad because AMD bumps like an extra 100 bucks on some processors and extra 50 bucks on some other processors with this new 5,000 series because because they could. Like that's it's great. That's the reason. And you can just go on Amazon and it's like $100 more for the for the core. And it's you can you could say that. Oh, the 5950 X sorry, the 5600 X is only $50 more than last generation of the 3600 X, which was 250. But you really can't say that. It's actually $100 more because there is no 5600 non X. And so you can't that is this. You have to pay $300 for a six core nowadays. Worth last generation, you could pay $200. So it's actually $100 more expensive if you look at that. Because you can compare that all you want. But if you want to buy a six core new, you are paying $100 more. And then one one last thing of single threaded performance of those big cores, the 5950 X is a single core at 650 on Cinebench. And then the new 12900 K is 810, which if that's if that's true, that makes it the fastest CPU in the world. Yeah, first single threaded. That's like those are record breaking numbers. And imagine if you could if you could look at nitrogen cool that thing. Oh, yeah, it may boost even more. But all right, well, we're just about out of time. So I'd like to thank you for watching this fourth episode of Techpad. I'd like to thank Kyle here for coming in to LPM to record this and be our special guest. Thanks for having me. Yep. And I hope you learned a little bit from Techpad. And I hope to see you next Wednesday at 6pm. Thanks for watching.