00:00SPEAKERDo you do you like Adam Murray or Adam Patrick Murray or Adam or Adam from PC world Adam Patrick Murray
00:05SPEAKERI don't Patrick Murray. I got a I got a funny story about that. So that's not that funny. It's just SEO really
00:11SPEAKERYeah, I get that. It's when when your name when you have it when you have two common names stuck together
00:15SPEAKERIt's not smart for people to find you who knew yeah when I first went to go make my website
00:20SPEAKERThere was at least three other famous Adam Murray's like there was a soccer player or football whenever there's a politician
00:27SPEAKERThere was somebody else. I was like, alright, you know what? I'm just gonna start going by Adam Patrick Murray
00:31SPEAKERLet me google it and at the time there was like no hits now
00:35SPEAKERThere is one other Adam Patrick Murray that you can find and it's a Florida man. Oh, no
00:40SPEAKERStory about a Florida man. So no, that's not me
00:43SPEAKERThat's the whole thing where like criminals and hate to say this like serial killers
00:47SPEAKERThey always use like a middle name to try to differentiate from other people. That's not you know, I've been told that as well
00:53SPEAKERLike hey, you know you realize you're setting yourself up for failure on that one
00:56SPEAKERLike well, maybe or maybe I'm hiding in plain sight. You just don't know. There's another Adam Patrick Murray out there
01:01SPEAKERWho's a ticking time bomb just waiting to destroy your reputation?
01:05SPEAKERBut you should just switch to a symbol. I was thinking about switching to a symbol. Yeah, it worked for Prince
01:09SPEAKERYou know, I make it I'll design something
01:11SPEAKERI got let me let me use let me use it one of our recent tips for my past episode and hit window
01:15SPEAKERPeriod and pull up. Okay, you got the emoji picker. What's your I feel mad this morning if I'm being honest, you're gonna be
01:22SPEAKEREmoji picker, you know, do you not use the emoji picker? No, it was period. Oh, we're about to blow your mind
01:28SPEAKERHey, there it is. Oh, wow
01:30SPEAKEROh, I like the ninja cat if I this is the first one that popped up that I was like
01:34SPEAKERYeah, I can I dig that one. That sounds good. All right old ninja cat Smith smiling face with sunglasses Brad and ninja cat
01:41SPEAKERMade a tech pod. Yeah, Brad and ninja cat made a tech pod. I want to be merman
01:44SPEAKERI want to be merman with with no shirt though. Sorry merman is taken ask me how I know
01:50SPEAKERcould I interest you in a piece of fruits or
01:54SPEAKERHere we go. How about how about a potato? Hey, what do you think will I mean potato?
02:20SPEAKERYou
02:28SPEAKERWelcome to Brad and will made a tech pod. I'm will I'm Brad. Hello Brad
02:32SPEAKERHi, hi, your browser window was like three windows deep. I didn't actually see you. Oh, it's okay. Call me best prize
02:39SPEAKERIt was like invisible will well, I mean that had sometimes I'm here. Sometimes I'm just in your brain
02:43SPEAKERThis is this is the beautiful thing about having when I walk around with my air pods in all day. I
02:49SPEAKERJust occasionally like something will start playing or I'll have like chill music going and then I'll get a call
02:54SPEAKERIt's like ah bones ringing. This is terrifying. I haven't heard the sounded years. Oh, right
02:58SPEAKERI have machines in my ears and they do that sometimes. Yes. Yes. Yes. Um Brad today. I'm really excited
03:04SPEAKERWe have as you guessed from the cold open
03:06SPEAKERWe have a guest on the show our friend Adam Patrick Murray from PC world is on
03:12SPEAKERfolks who are longtime listeners may remember
03:15SPEAKEREarlier this year and late last year you and I both went over to PC world where we built PCs with Adam and cordon
03:20SPEAKEROver there, they kind of guided us through the process of the modern
03:24SPEAKER13th gen slash Intel slash 5th gen AMD is that is it 5th gen or 4th gen?
03:31SPEAKEROh my god, it's Zen. It's Zen 4 on am 5 which is
03:37SPEAKER5th gen socket 4th gen Zen something like it look it's kids complicated. But um, Adam is a
03:44SPEAKEROften we have Gordon on during this time Gordon's out right now
03:48SPEAKERSo this week Adams on the show to talk to us about the stuff that Intel rolled out at their innovation event earlier this week
03:55SPEAKERincluding new CPUs next next year CPUs and then
03:59SPEAKERGetting further into the weeds a little bit with things like neural processing and longer-term plans from from the big blue chipmaker
04:08SPEAKERFoveros and angstrom and there's a lot of other terminology in there. You take your angstrom and shove it Brad
04:17SPEAKERWelcome to the show Adam. We're so glad to have you here. Oh, man. I I gotta tell you I'm honored to be here
04:23SPEAKERI've I followed both of you guys work for for a long time and
04:27SPEAKERYeah, the the fact that you wanted me on the show. I'm just like what me who what is it? Oh, yeah
04:33SPEAKEROh, I'm I'm excited to be here. Thank you for coming all your you're the keeper of the forbidden knowledge
04:38SPEAKERyou're you know about tiles and
04:41SPEAKERFoveros and
04:43SPEAKERWhatever they're calling the CPUs now. So hey, we're getting chip like crazy up in here the the information
04:49SPEAKERI just happen to stand in front of the information and it hits me, you know
04:54SPEAKERAnd some of it sinks in some of it just like goes past so that's how it goes. But yeah you you this
05:01SPEAKERYou you covered Intel innovation, which is I think this is like they're developed. They don't have a developer conference really
05:07SPEAKERBut it's like they're hey, here's what we have coming
05:10SPEAKERSoon and later and they do it every year. It seems like they didn't used to call it Intel innovation
05:15SPEAKERThey just just call it a press event in the old days now. It has a name
05:17SPEAKERWell, actually the the thing that Gordon likes to point to which I had never been to my sorry my coworker Gordon mong
05:24SPEAKERYou may remember him such publications as maximum PC and PC world
05:29SPEAKERBrad's and Brad's PC build video. Yeah
05:33SPEAKERHe he likes to to kind of link it to IDF the Intel developer forum it like in a lot of ways
05:40SPEAKERHe's like, oh, you know what hat this that was Pat's thing and then IDF kind of trickled away when Pat was gone
05:46SPEAKERAnd now that Pat's back. He's like, yeah, let's make this let's have a developer forum
05:51SPEAKERSo, I mean they talk a little bit about hardware
05:53SPEAKERI mean, we definitely got our hands on the first meteor like laptops
05:56SPEAKERBut at the same time it was all about about getting developers developers developers, you know that kind of stuff
06:02SPEAKERSo yeah, it's funny because the GPU people have done this for a long time
06:05SPEAKERLike Nvidia does their their developer conference AMD does developer reach out days
06:10SPEAKERThe Nvidia has Nvidia is what used to be their developer conferences become that Nvidia
06:16SPEAKERLike it's all about graphical compute and machine learning and stuff now. It's kind of crazy
06:20SPEAKERThen it's usually when they roll out new hardware
06:22SPEAKERSo it's it's it's I can't felt weird when Intel kind of moved away from this and I'm I like that it's back
06:27SPEAKERYeah, like I said it Gordon kind of equates it to Pat like it was probably a Pat kind of thing
06:32SPEAKERSo I think that that makes sense, but I should point out Gordon is on
06:37SPEAKERIt was was not able to join us at Intel innovation, unfortunately
06:41SPEAKERSo I was kind of there in his stead usually for people who know us on the PC world side
06:46SPEAKERI'm the video guy behind Gordon
06:48SPEAKERso I just I feel lucky to be able to to hang out with him and learn from him and he
06:54SPEAKERUsually I have the stupid questions and he's like, okay. Well, well, this is what that means this time
06:59SPEAKERI had to kind of be the the face of PC world. So kind of a newer thing for me
07:04SPEAKERSo it's a long a long tradition of small scrappy media outlets that the video and production people just find themselves in the spotlight eventually
07:11SPEAKERYeah. Yeah. I mean, sorry, we're not even getting into innovation
07:15SPEAKERBut yeah, like I've a it's been a crazy history here to foundry is the company that employs me technically
07:21SPEAKERAnd so I technically I'm on the video team
07:23SPEAKERBut I cover I run the main b2c channels
07:27SPEAKERSo in PC world is our main one that we focus on so it's it's uh, it's been fun
07:34SPEAKER2017 was kind of when I like kind of took charge of the YouTube channels and we can go in depth on on
07:40SPEAKERYouTube, you know in the hardware ecosystem over there, but like like I said just hanging out with Gordon and learning from him
07:46SPEAKERmaking the videos doing the podcast like I've been able to kind of
07:50SPEAKERAbsorb a lot of knowledge and I'm fairly new to two pcs and hardware
07:55SPEAKERI mean, I I literally learned how to build from watching you will and and you know
08:01SPEAKERreading articles and same thing with Gordon, so it's it's been a matter of
08:06SPEAKERMe just kind of learning from the best and and continuing there. Well, I was I think you undersell yourself a little bit
08:12SPEAKERbecause you have a pretty
08:14SPEAKERyou pay so
08:16SPEAKERBack in the maximum PC days when I worked with Gordon
08:18SPEAKERLike he focused on CPU stuff and I covered GPU stuff for a really long time
08:23SPEAKERand and then everybody else covered the other stuff because those were big beats that required a lot of time and a lot of
08:29SPEAKERThere was a lot of testing and a lot of stuff
08:31SPEAKERYeah, I think the thing that you have that that frankly I never had and Gordon has had off and on I don't think I
08:37SPEAKERthink it's stuff he kind of doesn't care about a lot of the time but it's like the all the how the all the other stuff
08:41SPEAKERand what people are actually doing with stuff in the real world versus the actual benchmarks like you you have
08:46SPEAKERLike you were the first person I talked to about undervolting stuff
08:49SPEAKERYou were the first person I talked to about reducing power consumption on high-end pcs
08:54SPEAKERYou you were the first person I talked to about like what's going on with weird RGB controllers and how you can do open RGB
09:01SPEAKERnow instead of having 25 different arbitrary pieces of software installed on your computer and and
09:09SPEAKERYeah, well, yeah. Yeah
09:12SPEAKERfan control
09:13SPEAKERI think it was awesome. Fan control is amazing
09:16SPEAKERThe windows building it in thing does not give me you I don't have a lot of cut
09:19SPEAKERThat doesn't give me a lot of comfort. But anyway, that's a different topic for another day
09:22SPEAKERYeah, all this is to say I was at innovation with actually Mark Hochman from PC world like so
09:29SPEAKERJust to back it up a little bit. So Mark Hochman flew. I will say under under
09:35SPEAKERThe sponsored, you know Intel sponsored the travel to go over to Malaysia to see their their fab over there
09:41SPEAKERSo he got he got the real deep dive nitty-gritty kind of stuff like at the fab
09:48SPEAKERYeah, yeah bunny suits everything
09:50SPEAKERYeah, it was it was nuts. They offered us to go to Israel. It was late last year. Maybe earlier this year
09:57SPEAKERI can't remember. Unfortunately, we didn't go to that event
09:59SPEAKERThe Malaysia thing didn't work out for me to go either but Mark was able to go and yeah
10:04SPEAKERThey they kind of showed off all their cool stuff
10:06SPEAKERSo, you know, hopefully I get to go to a fab one day because I think that would be that'd be fun
10:11SPEAKERBut so he got all like the deep dive technical stuff and then we took that knowledge going into to innovation
10:17SPEAKERWhere where they went to talk about it? So we had some pre-release stuff and there's videos up over our YouTube channel
10:24SPEAKERWell, I said this was kind of the dot-com. Yeah. Yeah YouTube.com forward slash PC world
10:30SPEAKERbut the
10:31SPEAKERThe idea is that hey you go to innovation and I was kind of curious to see how they were gonna message all this stuff
10:37SPEAKERCuz I think that's that's the bigger thing that is exciting for me is like there's a lot of technological stuff that's going on
10:44SPEAKERbut also the messaging is kind of
10:47SPEAKERThe more interesting thing because it is so complicated and it's it's not just like oh, hey, you know what now we got four cores
10:54SPEAKERHey now we got eight cores check this out
10:56SPEAKERWell, yeah, you it used to be that the that you looked at you were like, oh the CPU clock is 20% faster
11:01SPEAKERTherefore your benchmarks gonna be like 19.5% faster easy and and it just doesn't work like that anymore. Like everything is complicated
11:09SPEAKERI'm especially coming off of the sky Lake era where it literally was just that for like what five straight years or something
11:15SPEAKERI'm like, oh it's the same thing but 10% faster this year
11:17SPEAKERI wanted to ask a real quick like like the the highest level question we can ask about this event
11:23SPEAKERWhich is you mentioned Pat a few times that's Pat Gelsinger or is our help at Gelsinger
11:27SPEAKERYeah, CEO of Intel. So like would you say this events is like the
11:33SPEAKERfirst major or the biggest showcase of the Gelsinger era because like it was kind of a
11:37SPEAKERIt seemed like kind of a big deal in the press when he came back as CEO because like, you know
11:42SPEAKERthere was this like sky Lake stagnation period where they didn't have a lot going on and then it was you know,
11:46SPEAKERThe narrative has been like oh, you know
11:47SPEAKERHe's he's an engineer like he knows where this company should actually be going and like they debuted so much new forward-looking stuff at this
11:55SPEAKERevent
11:56SPEAKERLike I know that, you know
11:58SPEAKERCEO sorry CPU roadmaps are like, you know steering aircraft carriers, you know, they take years to
12:04SPEAKERJust kind of change the trajectory and get moving
12:07SPEAKERso like he's been back for what two three years now as a CEO is this kind of this like a taste of the future under
12:13SPEAKERhis I mean that I would I would say innovation last year like he came up there and was like hey, you know, we
12:20SPEAKERReally aggressive roadmap, you know what four different nodes over five years or I can't remember the the exact thing
12:27SPEAKERHe was talking about but he made it very obvious that he wanted to move really fast
12:32SPEAKERAnd do a lot of a lot of things now
12:35SPEAKERThere's been a lot of talk of like hey, they're they're not executing that well
12:39SPEAKERYou know, they're they're getting you know held up in in certain process types and and things like that
12:44SPEAKERBut in there, oh man, wow, they're having to work with TSMC, you know
12:48SPEAKERnow TSMC is gonna make at least a couple of the tiles and meteor lake like
12:53SPEAKERso I mean
12:55SPEAKERLike you can kind of interpret it different ways
12:57SPEAKERBut the way I take it is that I can tell that that he's passionate and he wants to like really push things forward and is
13:04SPEAKERTrying to be forward-thinking. I mean I like yeah
13:07SPEAKERHe's he's kind of a nerd and like he doesn't have the best stage presence, you know
13:11SPEAKEREspecially compared to like Jensen right from Nvidia, but like I just like that. He's he's he's a nerd
13:17SPEAKERHe gets up there on stage. He's self-deprecating. He he has fun with it
13:21SPEAKERLike I I kind of empathize a little bit more with him than than somebody like a Jensen
13:26SPEAKERHe was just like so smart and so cool
13:29SPEAKERbecause I'm not
13:31SPEAKERJensen just because Jensen has been practicing for 25 years now and is like
13:36SPEAKERHe didn't they don't all start out. They don't also they usually don't start out like that. So
13:42SPEAKERIt's it's but it's it's interesting to see because like 13 1300 series 13,000 series
13:47SPEAKERThe 13th generation was kind of like it's high and low right because performance is pretty good power consumption is really high
13:53SPEAKERBut it's not like it hasn't been they kind of got stomped by AMD in a lot of a lot of ways
14:02SPEAKERIt's just they're not we're there they're a little bit on the back foot for Intel and have been for quite a while at this
14:06SPEAKERpoint it feels like and and it seems like they're finally making decisions that are going to
14:12SPEAKERYou know, let them shift their shift their direction a little bit more
14:15SPEAKERI mean a lot of people want to to kind of put down alder like 12th gen
14:21SPEAKERRaptor like 13th gen and you know and say hey
14:24SPEAKERThey they're stagnant on the the big power cores and they had to kind of put throw in these these crappy cores to
14:32SPEAKERMake up for it in cinebench numbers, but also so I mean, okay, maybe to a certain degree. Yes, but the
14:39SPEAKERphilosophy of having two different kinds of
14:43SPEAKERperformance
14:44SPEAKERTo kind of lean on and now now with meteor like three different kind of cores and like to me
14:50SPEAKERI think it's more of like hey
14:51SPEAKERYou know what?
14:52SPEAKERThat was the first indication of things kind of need to change a little bit and obviously the on the AMD side
14:57SPEAKERThey've been working with chiplets and and kind of trying some different things there which is a great success obviously
15:03SPEAKERSo I think it's more like hey sure, they're they're a little slow moving but at least with with meteor like now
15:09SPEAKERIt's like hey, we're we're we're diving in
15:12SPEAKERAnd things are things are gonna get crazy in terms of like just performance
15:16SPEAKERLike how do you how do you understand engage performance back, you know Gordon, you know 10 20 years ago
15:23SPEAKERWould you run a cinebench and be like, okay, it's faster. Cool. How many ops does it make? Yeah, right
15:29SPEAKERYeah, so it's and that that's where that's where I'm excited to talk about
15:33SPEAKERit's just like I feel like this disaggregated architecture is like
15:38SPEAKERJust nuts and I mean it has huge implications
15:41SPEAKERWe should explain chiplets to people because I think I think it's a relatively it's not new on the AMD side
15:46SPEAKERBut it's still a relatively new concept to a lot of people
15:48SPEAKERyou know in the old days you would buy a CPU and the CP the entirety of the whole CPU would be cut out of a
15:54SPEAKERsingle wafer of silicon it would be etched in at a
15:58SPEAKERLike the traces would be carved by a certain wavelength of light which made them a certain width
16:03SPEAKERYeah, and that and that's what we refer to as the process size often. So this triplet thing is a different approach, right?
16:10SPEAKERYeah, so on the AMD side chiplet here on the Intel side. They're calling it tiles
16:16SPEAKERWhich you know what whatever blip by any other name still?
16:20SPEAKERYeah, the the idea that you have you have separate things doing
16:26SPEAKERDifferent kind of tasks and then you kind of bring it together like like the whole
16:30SPEAKERYeah
16:30SPEAKERlike you said the whole idea of a Northbridge Southbridge that that kind of stuff is going away as well and and it it allows
16:37SPEAKEROn the AMD side, right? They've been able to mix and match chiplets or oh, hey, you know what we have a package
16:43SPEAKERWhich is one of the the compute chiplets on it
16:45SPEAKERso and then Brad, you know your your new PC you've got two two chiplets and one of them doesn't have
16:52SPEAKERIt's just the normal standard cores. The other one has the the 3d stacked
16:57SPEAKERL3 cache or whatever and so it's just a matter of like
17:01SPEAKERIn the future, which I know we'll talk about naming a little bit later too because like that that's where things get real
17:07SPEAKERHarry is that Intel seems like hey, you know what with these
17:12SPEAKERIrrigated tiles you're gonna be able to mix and match ad do you want you want you know?
17:17SPEAKER10 10p cores or do you want 2p cores? Do you want to you know a bigger?
17:23SPEAKERGraphics performance. Do you want a bigger NPU? You know, the sky's the limit
17:28SPEAKERBuild your own CPU
17:29SPEAKERI guess we should we should mention real quick just for people who don't know like the the whole point of doing things this way
17:34SPEAKERIs for manufacturing efficiency because it's not every CPU that comes out of fab is good, right?
17:39SPEAKERLike that's what the process of binning is is basically testing CPUs to see if they work right after their fab
17:45SPEAKERSo when you're when you're making the CPU package out of multiple chips
17:49SPEAKERfrom different different fab processes you can
17:53SPEAKERyou know, you don't have to throw a whole CPU away because one part of it doesn't work right right like you can you can mix and
17:58SPEAKERmatch the
17:59SPEAKERYeah, I mean it's that which obviously directly impacts cost the the less waste they have the the or the better year yields
18:06SPEAKERThey have the less money. They're throwing away
18:09SPEAKERBut then also it lets them iterate faster to like the the other thing which kind of is talked about here
18:15SPEAKERIs that for meteor lake and it's gonna launch on the laptop?
18:19SPEAKERFirst the Intel did confirm meteor lake for desktop in 2024
18:24SPEAKERTo us and the the idea is that hey the the tiles the tiles can be upgraded in different ways
18:33SPEAKERSo like generational naming what does that even mean anymore?
18:36SPEAKERLike if it were if if meteor lake was 14th gen and then next year, they're like, hey, you know what?
18:42SPEAKEREverything else is the same but we replaced the tile
18:45SPEAKERfor the the SOC part of it or we updated the
18:50SPEAKERThe graphics tile like is that a whole new generation?
18:55SPEAKERNo, not really, you know
18:56SPEAKERLike so like the the way that the fact that they're getting away from generation naming is kind of like kind of a big deal
19:02SPEAKERIn a lot of ways for good and for bad
19:05SPEAKERSo like so this isn't I mean my first question for me about me
19:09SPEAKERYou're like was this isn't just it because they're they're also reducing the process size for this
19:13SPEAKERSo the you generally when the cert when the when the process gets smaller
19:16SPEAKERThey the cores you can do more compute in the same
19:19SPEAKERBut then you can fit more transistors on the same amount of space and that means that things get faster and they use less power
19:25SPEAKERPer per like the power per compute is is less
19:31SPEAKERIt's not just a process this is a whole new this is a whole new architecture kind of kicking off with the triplet design
19:37SPEAKERRight. Well and and each tile has its own process
19:40SPEAKERthat's why I'm saying like like the fact that that they can update maybe one process out of the whole chip and
19:46SPEAKERand and just kind of
19:48SPEAKERReintroduce that into the market is kind of a different way of thinking of it too. Like it's once again
19:52SPEAKERIt's not a generational thing because maybe we'll get a an update to one tile, but not the other
19:59SPEAKERAnswered oh, sorry Brad real quick
20:01SPEAKERI this is from the the verge story on this like the I believe is so in this iteration of meteor lake like for example
20:07SPEAKERthe graphics part of the chip is on TSMC is five nanometer process, but the
20:12SPEAKERIO and SOC tile are on TSMC's
20:15SPEAKERN6 whatever process that is and then part of it is made by Intel
20:18SPEAKERSo like that's your that's to your point right that exactly they can literally source parts of the CPU from other
20:25SPEAKERFabrication companies and just kind of slap it all together
20:28SPEAKERExactly. This is a response to the 10 nanometer process being kind of plagued by issues over the last few years, right?
20:35SPEAKERRight. I mean, I you know as as as much as people
20:39SPEAKERWill give them grief rightly. So it probably was one of those things where they were they were not happy with it
20:45SPEAKERI'm sure and they were like, okay, how the hell do we change this how to how the hell do we get out?
20:50SPEAKERYou know in front of this and I think in the future who knows who knows but I would think that this would allow them
20:57SPEAKERTo be more lean and be like hey, you know what we can we can control each tile that's included here
21:03SPEAKERAnd and kind of have different different ways of packaging it and upgrading it
21:08SPEAKERBut yeah for me you're like there's there's four specific tiles
21:12SPEAKERYou have the compute tile you have the SOC tile, which is very different the IO tile
21:18SPEAKERAnd then the system on a chip in this case. Yeah
21:22SPEAKERyeah, and the reason the reason why they call it that is because that SOC tile now has
21:27SPEAKERTo low power e cores. So the compute tile has p cores and e cores, which you know was introduced with Alder Lake
21:34SPEAKERBut now the SOC tile has we're calling it shorthand LPE cores
21:39SPEAKERSo the idea and this is more for the laptop side because that's mostly what they're talking about here
21:44SPEAKERIs that if you need to wake up the computer or you need to run something that's really low power
21:49SPEAKERYou don't need to light up that whole
21:52SPEAKERthe whole compute core you can just light up the SOC core because it has two low power e cores and
21:59SPEAKEREspecially that's where the media engine is is as well
22:01SPEAKERSo you can you can play back video and just only use those two cores and the compute
22:07SPEAKERYou know tiles just sitting there dormant and not using power
22:11SPEAKERWhich is is definitely a huge advantage on the laptop side how it works out on the desktop side is
22:17SPEAKERYet to be seen but yeah, that's that's that's interesting. It seems like they're taking that bottom-up approach to
22:22SPEAKERTask assignment whereas like my impression of 12th and 13th gen especially like looking at you know
22:27SPEAKERLinux kernel development and how they're handling it there
22:29SPEAKERIt seems like it was very much prioritized the p cores and have things trickle down to the e cores as necessary
22:34SPEAKERAnd again, obviously like the power needs are different on desktop versus laptop
22:38SPEAKERBut like to hear that to hear that they're there, you know, it's this it's this tiered like, okay
22:42SPEAKERAre we gonna just light up these like super low power ones? Like do we have to even activate the compute tile like okay?
22:47SPEAKERYes, we do. But now we've got the e cores there like that. That seems really smart
22:52SPEAKERYeah, we have a really good interview with
22:54SPEAKERMichelle Johnston Holt house who's the kind of the EVP and GM of the client computing group?
23:02SPEAKERShe's awesome. She she is great on camera a good interview with that mark did on our channel definitely go watch it
23:08SPEAKERBut that was one of the things
23:10SPEAKERThat was kind of talked about was
23:12SPEAKERHey thread director has been updated to prioritize
23:16SPEAKERPutting things on the low power e cores first if it can and then and then go that way which makes sense on a laptop
23:23SPEAKERBut on a desktop, do I really want to throw everything to the low power e core first, you know, they say
23:30SPEAKERIntel says
23:32SPEAKERThat they that actually can be tailored and changed depending on on use case
23:37SPEAKEROnce again, we don't know meteor like on desktop isn't here yet
23:41SPEAKERSo that's that's something that I'm kind of curious about but on the laptop side
23:45SPEAKERI mean that is definitely like a huge advantage to be able to just you know, use less power
23:51SPEAKERI mean that was that was kind of the other big takeaway is that all this is done for efficiency there
23:55SPEAKERThey're saying the actual compute tile in there compared to 13th gen raptor lake is is pretty much even there
24:02SPEAKERThey're kind of already going out there and saying hey, you know what like set your expectations like the actual compute
24:08SPEAKERIs not going to get a huge uplift
24:11SPEAKERBut it's all about efficiency
24:13SPEAKERyou know and lower power the the place that is gonna get a big upgrade is the
24:20SPEAKERIt's like the graphic style, which is is the first time they're gonna have arc graphics
24:25SPEAKERon
24:26SPEAKERThe on the chip it previously was XE or however you want to say it. Yeah, I just say XE
24:34SPEAKERBut yeah, so the arc performance they're saying like two times performance
24:38SPEAKERon that and then of course, there's the NPU we haven't even gotten into which is the
24:45SPEAKERThe AI stuff quote-unquote
24:48SPEAKERYeah, it's it's interesting. So what's the block map look like for this?
24:51SPEAKERIs that does everything run through the IO ports and then everything's?
24:54SPEAKERSpokes off of the IO or is it the IO to SOC and then everything else out to from the SOC?
24:59SPEAKERSeaport or did they get into that level of detail?
25:01SPEAKERYes, they did SOC SOC is the the main driver
25:05SPEAKERSo it'll touch even the the the DRAM is is plugged into the SOC
25:11SPEAKERSo everything kind of goes into the SOC and then gets desegregated out from there, which is why they're there especially for laptop stuff
25:19SPEAKERThey're like hey them the more we can do on this this SOC tile the better because it's it's the most power efficient
25:25SPEAKERYou know and it doesn't need for laptop you okay?
25:29SPEAKERI should say for normal laptop users right the person just writing emails or or taking it to Starbucks
25:35SPEAKERKind of thing like that's what they want that that's where Apple has been able to kind of kill it right as efficiency and and whatnot
25:42SPEAKERSo like that they they feel like in this new kind of SOC tile that that's we're gonna they're gonna have the big win
25:50SPEAKERDo they talk much about the interconnects between these between the tiles because that's like that's the big downside for people who don't know between
25:56SPEAKERTwin between or the downside of this desegregated architecture, which is how Intel is referring to this setup is that?
26:03SPEAKERIf everything's on the same die it talks to each all the parts talk to each other
26:07SPEAKERextremely fast and when they're not on the same die you have to have some other way of
26:12SPEAKERFacilis facilitating communication between them on AMD. It's called the infinity fabric
26:17SPEAKERThey talk much about what their version of that is like how fast it is like how much it's because you run into things with
26:23SPEAKERa on AMD where it's like oh, you know the the infinity fabric lock can only go so fast and that's kind of affected by your
26:29SPEAKERMemory speed as well and like that can hold you back and bottleneck you in some ways
26:32SPEAKERLike did they talk about how there's how there's works and stacks up?
26:36SPEAKERYeah, so they typically they've they've run on a thing called a ring bus. So it's it's a you know
26:43SPEAKERAll on one chip. It has different stops and then it's just easier to kind of
26:48SPEAKERAnd I put stuff around there and they kind of famously made made fun of AMD and just said oh, it's glued together quote-unquote
26:56SPEAKERin this in this new architecture
26:58SPEAKERIt's mini rings. So essentially each tile has kind of a somewhat of a ring bus that talks to each other there is
27:06SPEAKERThere is a penalty to that
27:10SPEAKERespecially on the power I'd
27:11SPEAKERIt's it the engineering side of it is going way deeper than and I can talk to but like it
27:18SPEAKERIt is not it is substantial, but it's not
27:21SPEAKERLike not a huge thing
27:24SPEAKERIt seems like that the it's something that they're they're touting to to be not not that big of a problem
27:30SPEAKERbut I mean
27:30SPEAKERI guess I guess we'll see that's that's where the the testing runs it too because if if you have any sort of
27:36SPEAKERLatency or penalty to to fire up the big compute tile, especially for power users. I mean that's gonna be a problem though
27:45SPEAKERthe
27:46SPEAKERSo you alluded this earlier the the they've said
27:51SPEAKERWe've talked about North bridges and South bridges going away multiple times over the last decade and change, right?
27:57SPEAKERBut it seems like this time it actually is for real happening because of the in the way the the chiplets are architected
28:04SPEAKERNorth bridge and South bridge for people who don't know
28:06SPEAKERPreviously and I always get this wrong
28:08SPEAKERso please correct me when I when I mess it up but
28:10SPEAKERin the in the in the old days the South bridge was the thing and that like the slow stuff connected to it was like
28:17SPEAKERSATA ports and and
28:20SPEAKERMaybe if you had ISA slots or PCI slots on a board with some PCI Express slots
28:24SPEAKERYou'd plug those into the South bridge
28:26SPEAKERMaybe your USB bus in the pre USB 3.1 3.2 days would plug in there all that kind of stuff
28:31SPEAKERand the North bridge was the kind of
28:34SPEAKERmemory CPU
28:37SPEAKERVideo card PCI Express slot like the handful of places that you wanted to be real fast these days
28:43SPEAKERIt would be SS like your NVMe SSD stuff like that
28:48SPEAKERDid I do it
28:51SPEAKERYeah, that makes sense
28:54SPEAKERAnd so this is going away. This is all kind of getting sucked up into that SOC. I those SOC and IO
29:00SPEAKERChiplets now, right
29:02SPEAKERYes. Yeah. So the IO tile specifically holds the things like PCIe on Meteor Lake
29:08SPEAKERIt's Thunderbolt 4 which is kind of weird because they did just introduce Thunderbolt 5
29:12SPEAKERSo that's kind of I'm a little bummed that this doesn't have Thunderbolt 5 support
29:17SPEAKERThey always like Thunderbolt a generation like they have like they're always a year back on Thunderbolt new chipsets. It seems like though
29:24SPEAKERYeah, yeah, and I mean just to take a side tangent like Thunderbolt is not a deal breaker for me like, you know
29:31SPEAKERThere's been a lot of testing the US before and things like that and I feel like that's hopefully gonna get to to a good
29:37SPEAKERSpot but the the one spot which I'm just a weird pervert for is e GPUs
29:42SPEAKERAnd Thunderbolt 3 and Thunderbolt 4 hasn't been great for e GPUs
29:45SPEAKERSo supposedly Thunderbolt 5 is maybe where we're gonna have another e GPU revolution
29:50SPEAKEROh, I don't know. I'm excited about that. But that's that's the separate thing. So that's a weird thing to get into Adam
29:56SPEAKERI don't know your
29:59SPEAKERBut the idea of having a thin and light laptop that you plug into an e GPU like hell, yeah
30:04SPEAKERThat's uh, you know, I mean there's people I use it today and understand the downsides and still enjoy the benefit
30:10SPEAKERBut man, if if you didn't have those massive downsides like I could see a lot of people
30:15SPEAKERAppreciating it more. I think I think unfortunately it's always gonna lag but um, but okay
30:20SPEAKERSo so PCI Express all of your main I/O stuff is gonna presumably plug into that I/O
30:26SPEAKERTriplet memory and then the I/O I/O
30:29SPEAKERInterconnect and then go to the SOC and then the compute cores of the graphics cores and all the other stuff hooking to the SOC
30:36SPEAKERI'm I am
30:38SPEAKERWildly curious to see how this benchmarks because it took a while when AMD switched to this triplet design
30:43SPEAKERfor Windows to kind of know what to do with a
30:46SPEAKERvariety of like the performance and efficiency cores and
30:48SPEAKERhow to handle the new bottle like how to work around work with the bottlenecks that that changed when this stuff changed and
30:55SPEAKERIt's I mean, I guess it's gonna be I guess we'll see laptops sooner
30:59SPEAKERBut when we get to the big metal in the in the in the springtime, hopefully next year or whenever they release meteor lake
31:06SPEAKERWe'll we'll see more
31:08SPEAKERwell, yeah, and I mean that's the thing right and
31:12SPEAKERThe the secret sauce at least according to Intel is thread director
31:16SPEAKERbecause that is something they had to work on for 12th gen and 13th gen and
31:21SPEAKERKind of gives them a little bit of a leg up is to say hey on our side we can we can control
31:27SPEAKERWhere things need to go we don't have to to have windows
31:31SPEAKERI mean it works in conjunction with windows obviously, but that is something AMD does not have
31:36SPEAKERis is that that layer in between and there's a lot of people who don't like that idea of the thread director and
31:42SPEAKERWant to be able to kind of control it themselves
31:44SPEAKERBut I mean I think for the general purpose person like it's it's gonna be advantage and like I said
31:50SPEAKERThey they have we're already thinking about these updates, and how it's gonna make sense here for the laptop side
31:57SPEAKERonce again I the desktop side that that's a
32:01SPEAKERWay bigger question that that for me like I'm just like I don't know exactly how that that's gonna work out
32:06SPEAKERDo you guys actually manage what gets on a p-core and what gets on a core on on your machines that use these processors?
32:14SPEAKERPersonally no
32:15SPEAKERYou know like there's plenty of people out there who if they just game and they're on Intel systems. They'll turn off their e cores
32:23SPEAKERYou know and get a benefit. There's like there's definitely people doing things like that, but for me I
32:30SPEAKERLike to use it how like an interest experience it out of the box and see where it fails there
32:36SPEAKERFor for Gordon when he benchmarks this stuff once again. He doesn't he doesn't necessarily turn anything off
32:43SPEAKERI mean he does certain benchmarking that will say okay
32:46SPEAKERYou know what I'm gonna stress the p cores, and we'll see how the e cores add to it
32:51SPEAKERBut for the most part. I mean general people don't do that especially on laptops either
32:58SPEAKERYeah, it's it's interesting cuz like I looked just to see what was running right now and right now
33:03SPEAKERIt's like the things running on e cores for me are the two browsers
33:06SPEAKERI have open and
33:07SPEAKERLike a power toy that just is kind of always in the background and and it isn't particularly used that often so I
33:14SPEAKERIt's just something I don't even think about and I wondered if other people are thinking about that more yeah
33:18SPEAKERI did so AMD does not do e cores yet. They're they're
33:23SPEAKERThey just have one core architecture across all chip. Let's the thing I have though is what Adam mentioned earlier
33:28SPEAKERI've got two triplets and one of them has way more cash for games and does windows know which
33:33SPEAKERChiplets who put a process on if it's game or not sometimes not always mostly
33:39SPEAKERSo
33:41SPEAKERThey've had to work with Microsoft, and I mean it definitely gets better right like I mean that there was a lot of headache around
33:50SPEAKERZen Zen one launch like a lot of people were having problems with with latency and things like that and over time
33:57SPEAKERI mean it is just gotten better and better
33:59SPEAKERfor sure
34:01SPEAKERSo the the addition of the the AI co-processor that they call men cores is that right?
34:08SPEAKERthe the the the NPU lives on the SOC tile and
34:12SPEAKERAnd this is just like the normal. It's like the tensor core equivalent on Nvidia GPUs
34:17SPEAKERIt's the thing that that accelerates the kind of math that happens in machine learning applications
34:23SPEAKERYeah, and we saw a nice little talk by Robert Halleck which used to actually be at AMD for 12 years
34:31SPEAKERAnd he jumped ship now. He's on team blue
34:33SPEAKERWe also have an interview on our podcast the full nerd
34:36SPEAKERThat you can go check out, but yeah for them. It's it's kind of it's kind of two things
34:41SPEAKERit's like hey now we have a
34:44SPEAKERProcessing unit in here, and they they say it's it's more similar to the GPU than the CPU in terms of computational stuff
34:52SPEAKERWhich makes sense, but it's gonna be better at certain things than each of them
34:56SPEAKERNow me being naive. I just assumed hey, you know what all the AI quote unquote stuff is just gonna be best suited for the
35:03SPEAKERNPU but in in the various demos that they showed at innovation a lot of the people that were talking to us so like hey
35:11SPEAKERNo, it really depends on the software the implementation what you're actually doing with it and all the demos
35:17SPEAKERYou know usually you go to a trade show and they want to show like the the latest and greatest thing right not all the
35:22SPEAKERDemos were actually using the NPU
35:25SPEAKERSome of them were for sure but a lot of them were using the disk aggregated
35:31SPEAKERArchitecture of it and saying hey, you know what this is that this one function is better better suited over here
35:36SPEAKERthis this one functions better suited over here and and
35:39SPEAKERThe way they're kind of talking about it at least on Intel side is that this is truly a third pillar a new third
35:46SPEAKERpillar in computing have the CPU the GPU now the NPU and I mean that's crazy like
35:53SPEAKERWhen was when was the last time there was a new pillar?
35:58SPEAKERWe should we should say in stands for neural just for people neural processing unit is what they're saying
36:02SPEAKERIt's funny how things worked out because there was that massive Microsoft leak the Xbox leak earlier this week like the day before
36:10SPEAKERInnovate I think or innovation
36:12SPEAKERI think and I can't remember the exact date on the on the Xbox slide about the next Xbox the one that's gonna come out
36:18SPEAKERIn 2028. It was it was a very speculative bullet list of like, you know, are we gonna go arm or go or stick with?
36:23SPEAKERAMD's x86 stuff but in that slide
36:26SPEAKERThey also used the term NPU and that was the first time I had seen it and that was in
36:30SPEAKERPresumably an AMD context and then here we are the next day
36:33SPEAKERIntel talking about their NPU. So I gather that's going to be the generic industry term
36:38SPEAKERProbably going forward for here. This is our AI compute unit. Well, they're not gonna call it a course
36:44SPEAKERI think I think early on it was referred to as a VPU
36:48SPEAKERLike I know I know a or a Apple. I think Apple refers to it as an NPU like I mean
36:55SPEAKEREssentially, it's kind of the same thing
36:59SPEAKERBut yeah, I mean the the idea of having a whole separate
37:03SPEAKERcompute
37:04SPEAKERlike like
37:06SPEAKERLike a like a place to do computing is wild and especially here to innovate one of the big things that Intel people kept saying
37:13SPEAKEROver and over again. I mean the question is okay, that's great
37:16SPEAKERBut what the hell can you do with it, especially for the normal user right now?
37:22SPEAKERYou know who's going to buy a laptop? Why do they care about a quote-unquote AI PC?
37:27SPEAKERWhat is it gonna gonna do for them and obviously there's demos
37:30SPEAKERI mean they showed it off, but at the same time Intel's like hey listen, we want to get it out there
37:35SPEAKERThey they had a number of like I can't remember like 10 million units
37:39SPEAKERYou know and in the next year is everything Intel based is gonna have an NPU in it of some sort
37:45SPEAKERso they they they want to get the the hardware out there and then and then let the
37:52SPEAKERThe market kind of decide what happens with it, and how people are gonna use it well
37:57SPEAKERThat's interesting for two reasons right one is that the the the most it's interesting
38:03SPEAKERThey put in the SOC versus in the compute chiplet because if they put in the compute chiplet
38:06SPEAKERIt'll only go out on the high-end stuff or in specific subsets on the SOC presumably that means it's gonna be on literally everything
38:12SPEAKERThey sell that it's a meteor lake chip
38:14SPEAKERFor the next year, so my guess is that the SOC is the gate on what the generational shifts like when that SOC shifts over
38:21SPEAKERThat's when we see them say okay. This one's a 14th gen. This is a 15th gen. This is a whatever gen
38:27SPEAKERThat's a guess. I don't I don't I don't know for sure they haven't said either as far as I can tell
38:32SPEAKERWhat what kind of what kind of demos did they show like what were what were the actual use cases for the hardware?
38:38SPEAKERyeah, so some of them were simple kind of things that we had seen before like
38:44SPEAKERWhat is it like like video collaboration
38:48SPEAKERYou know enhancements, and this is actually something AMD showed off with as well
38:52SPEAKERAMD would also like to point out there the first ones with an AI accelerator on a chip came out was our
38:59SPEAKEREarlier this year so a couple of their laptop chips had a dedicated
39:05SPEAKERProcessor on there, I can't remember the the exact details of it, but it's from their Xilinx acquisition
39:10SPEAKERSo AMD is working on this stuff and technically they're their first to it on the consumer side
39:16SPEAKERBut what they showed off is kind of the same thing that Intel was showing off was that that video collaboration stuff hey
39:22SPEAKERIt's a lot more power efficient if you can run background blur
39:26SPEAKERLocally you know on on this this processor
39:30SPEAKERSame thing was like denoisers and things like that
39:34SPEAKERAMD worked with
39:35SPEAKERMicrosoft to build in this thing called Microsoft studio effects, so it's actually something that just shows up in your your little your little pain
39:44SPEAKERWhen you click on it, and it's like oh a studio effects turn on black
39:48SPEAKERbackground blur and all that kind of stuff
39:50SPEAKERAnyway, so the they showed off that in the form of X split so in the the actual background blur and the face tracking was
39:58SPEAKERAll happening on the NPU and then they were feeding that feed into OBS which was running on the the GPU
40:05SPEAKERAnd it's it's faster because of the two times performance on the ARC graphics chiplet
40:10SPEAKERSo there was that one they had a film aura, which is a video editing software
40:15SPEAKERIt was like real-time playback of effects that they were doing and that was happening on the NPU they showed
40:22SPEAKERAdobe Lightroom they had a denoiser that just works quicker on the newer arc GPU that one was just on the GPU
40:30SPEAKERthey had GIMP running stable diffusion a
40:35SPEAKERBunch of different plugins, and this is all enabled by open vino
40:39SPEAKERBut stable diffusion running inside GIMP, and they had the the different
40:45SPEAKEROptions that you could have in the plug-in be like hey
40:47SPEAKERDo you want to pick the CPU GPU or NPU the demo where they were showing off was a mixture of all three?
40:53SPEAKERAnd then one that was actually more near and dear to my heart was audacity they had a stable diffusion plug-in
41:00SPEAKERActually a couple plugins, but the one that I liked was a stable division variant called
41:06SPEAKERrefusion
41:07SPEAKERWhich you know for me like when I first thought I was like wait. I thought stable diffusion was image-based
41:12SPEAKERThis doesn't make sense so what it's doing is like taking taking the audio
41:18SPEAKERspectrogram of it which is a a
41:22SPEAKERWhat is it's like a visualization of the frequency
41:27SPEAKERSpectrum yeah, it's like that like for the for where the where the data is in the in the audio spectrum
41:33SPEAKERWhich is is technically is I mean it's an image right so that you feed that image into stable diffusion and
41:40SPEAKERIt changed it spits out a new image then gets translated into audio and and what they were showing well one of the things that
41:46SPEAKERWere they were showing for that was hey, you know what create take this song like a like a Taylor Swift
41:53SPEAKERYou know kind of song and make it make it EDM
41:57SPEAKERSo it would it would take a spectrogram of that song
42:00SPEAKERRunning through stable diffusion which would spit out a different image wow that and it would play back
42:06SPEAKERI mean it wasn't great. It was it was fun. It was fun to mess around with but yeah, you're not useful, baby
42:13SPEAKERYeah, yeah, I mean it could be like I was thinking I was like man. We've had just a little inside baseball
42:20SPEAKERwe've had some like licensing problems with our
42:22SPEAKERthe music on our podcast like we had a needle drop service that we didn't re-up with and we went to a different service and
42:30SPEAKERIt was a so we're like well crap. Okay. We got to change the podcast music, but now I was like well crap
42:37SPEAKERI mean if I can if I can actually make
42:40SPEAKERPodcast music you know through refusion to be like hey
42:44SPEAKERYou know what take kind of like that song and and and maybe spit something out so you know
42:48SPEAKERI mean it's the right thing, but yeah, there's a whole can of worms Adam. Yeah
42:54SPEAKERSo did they did they give you a sense of what the API?
42:58SPEAKERWe're what the access to this these end cores is like for developers. Is this something where there's a Windows or OS level?
43:05SPEAKERAPI that developers are writing to and then there's just a driver that lives between the NPU and the and the and
43:11SPEAKERThe the software and then the OS or is this something where you have to build?
43:16SPEAKERspecifically is like using CUDA cores for Nvidia where you have to have to build specifically for the for the
43:22SPEAKERCore that you're working on I mean they didn't show anybody who is building specifically for it
43:27SPEAKERBut that doesn't mean that that I mean that's probably gonna happen that it wouldn't surprise me
43:31SPEAKERBut the thing that they were touting was open vino, and that's kind of like a translation like an API essentially
43:38SPEAKERAnd it works in conjunction with things like the direct ml and things like that so I
43:43SPEAKERThat's a little bit out of my pay grade like I kind of had to quickly learn how it was working at innovation
43:50SPEAKERBut yeah open vino was the thing they kept saying over and over and over again. Is that is that just for Intel hardware? Do you know?
43:57SPEAKERYou know I actually don't know okay. I looked it up on its open source
44:01SPEAKERI pull up the github repo and they mentioned Intel hardware enough that I assume it might be specific to Intel's stuff
44:08SPEAKERBut I'm not sure yeah, I don't know
44:11SPEAKERVNN eyes is another one that's in the compute chip
44:14SPEAKERYou know so that there's there's different kinds of standards, but open vino was used
44:19SPEAKERIt's all over the show floor
44:22SPEAKERSo I think that's where they're really hanging their head this whole situation is interesting to me like the the the widespread use of the term
44:29SPEAKERNpu like you know some people might be like I who care who cares what they call it
44:32SPEAKERIt's just jargon
44:33SPEAKERBut like to me it implies a process of standardization that kind of is very reminiscent of the 90s 3d accelerator wars because like
44:40SPEAKEREverybody's coming up with their own ship right everybody now sees that there is a need for this type of compute
44:44SPEAKERSo they're all making their own thing
44:46SPEAKERBut they're all gonna have their own custom API to go along with it like 3d effects
44:50SPEAKERI'd glide back in the day, and then we're presumably we're gonna go through some open GL direct 3d
44:56SPEAKERkerfuffle of competing standards
44:58SPEAKERand then eventually there will be just like the one API like DirectX that everybody uses for this stuff or
45:03SPEAKERBalkan or open g like there's but there's never it never it's never just one, but you know
45:10SPEAKERBut it'll be settled it'll be kind of settled to the extent that like okay. This is what we use now like we draw
45:14SPEAKERThese are our options. We know what we're doing. There's no there's no format war stuff going on anymore well in the Halleck talk
45:21SPEAKERHe he was really
45:24SPEAKERPumping up like hey. This is where this is where Intel
45:28SPEAKERCan play ball?
45:29SPEAKERYou know compared to to AMD and to a certain certain agreed Nvidia is that they were like hey here?
45:35SPEAKERHere's the amount of developer support and money
45:38SPEAKERYou know that we're able to get in there and and Gordon always talks about like that's the key to success is dev rel
45:43SPEAKERRight and why MD has kind of lagged in a lot of ways and why Nvidia?
45:47SPEAKERJust just kills it in a lot of ways as well
45:50SPEAKERIt's like if you can put developers in there and spend the time and money to have them
45:54SPEAKERImplement it then you have a leg up, but a lot of people say that's cheating. He was like oh, you know
46:01SPEAKERLook it's a choice to invest money in I
46:03SPEAKERI used to feel like it was cheating 20 years ago
46:05SPEAKERAnd now that I've been on the other end of the of the stick and actually been able to call people and get help with
46:10SPEAKERStuff like it is is a competitive. It is an actual competitive advantage
46:14SPEAKERI think and so in this case Intel is saying hey like the amount of dev rel
46:19SPEAKERWe have in here the amount of support. They showed a list
46:22SPEAKERI actually don't think it's in the slides that I had sent you guys, but they showed a list of companies
46:29SPEAKERthat they're working with
46:31SPEAKERThat it was just like hey
46:32SPEAKERYou know what look at our list right now, and we haven't even launched compared to the other the other lists
46:37SPEAKERAnd look we got more names on this list
46:41SPEAKERThe list with names on it is always kind of bullshit, but the thing the thing the thing to me that says hey
46:45SPEAKERWe're committed to this is the thing that Apple did with the like what a 13 or where a 14 whatever was the first one
46:51SPEAKERThey shipped with neural engine support
46:53SPEAKERThey're saying hey
46:54SPEAKERWe're putting it in all the phones not just the good one the fact that they're putting in the SOC
46:58SPEAKERChiplet instead of the graphics or compute chiplet says hey, this is gonna be on everything that's 14th gen from the hundred dollar processor
47:06SPEAKERIt'll be out in two two and a half years like that
47:09SPEAKERThey're trying to get this to a billion users faster than as fast as they can and and that to me is the is the mark
47:15SPEAKERBecause the interesting they get like on the phone side the place that this whole technology got interesting is when they stopped having to send
47:23SPEAKERVoice stuff to the to the cloud to do the compute on on on Siri decodes or on on Google voice decodes
47:30SPEAKERand and when that happened the time to response dropped from three or four hundred milliseconds to like
47:36SPEAKER75 or 100 milliseconds and it feels much more much more instantaneous and becomes much more usable
47:42SPEAKERAnd and and that's why this is that's why this is interesting coming in the CPUs. It feels overdue
47:48SPEAKERGiven what's happened on arm in the last five years and how fast this stuff rolled out it with arm, honestly
47:53SPEAKERWell, yeah, and and there's big benefits. I mean Intel's first major benefit that they pointed out was oh, hey
47:58SPEAKERIt's a lot cheaper to make the clients do it then have everything happen in the cloud
48:03SPEAKERYeah, you know which you know is not necessarily I don't care I guess to a certain degree but like to them
48:08SPEAKERThey're just like hey the the more we can offload to the client side the better, you know
48:13SPEAKERand obviously if it's if it's more
48:15SPEAKEREfficient then and it's something that's that's going to be valuable
48:20SPEAKERBut also, I mean there's privacy concerns, too
48:22SPEAKERThere's a lot of people who really value their privacy and their data and especially if they're feeding in sensitive material into these data sets
48:30SPEAKERThat they want to say hey, you know what? I don't want this to be open to the public then
48:35SPEAKERYeah, you can run it locally. I mean another another thing that's not as serious as that is like face tagging
48:42SPEAKERRight, like in Google Photos is like hey, you know, like I'm gonna tag all the faces and organize everything but now there is
48:49SPEAKERGod, I can't remember Gordon got a demo of
48:51SPEAKERPiece software and now the name is escaping me where you can actually use
48:56SPEAKERLocal hardware to do all the the face tagging so it's just more private
49:02SPEAKERThat way so I mean there's benefits but also at the same time it's funny because on Thursday
49:07SPEAKERWhich was two days after innovation started
49:11SPEAKERIntel had I'm sorry Microsoft had their big event
49:15SPEAKERwith
49:16SPEAKERupdates to their surface lines
49:18SPEAKERBut then talking about you know, what's coming with Windows and whatnot and their big push was to put more things in the cloud
49:25SPEAKERBecause it's a subscription and service base. So like Intel two days prior was like hey
49:31SPEAKERWe want to bring so much more stuff local and then Microsoft was like, oh man, we want to do so much more on the cloud
49:37SPEAKERAll the battle lines are being rewritten. It's weird that Intel
49:40SPEAKERsells CPUs to consumers and the cloud
49:43SPEAKERProviders and a and Microsoft is is a big cloud service provider. Now, what a weird
49:49SPEAKERcoincidence that
49:51SPEAKERShould we talk about so this is the other thing about meteor lake. We're almost done talking about the next year's processors
49:58SPEAKERBut it's it's what they're calling the Intel for process
50:00SPEAKERWhich I think they previously called the seven nanometer process and this is a little in the weeds for us
50:05SPEAKERBut but basically the the story so far on Intel stuff is that over the last 50 years?
50:11SPEAKERThey've been better at process tech than and pretty much everybody else there and they've had a rough time the last few years
50:17SPEAKERFor for the first time in quite a while
50:20SPEAKERand this this seems to be them saying both this chip rearchitecture and
50:25SPEAKERThe new the next that is a seven nanometer process seems to be them
50:30SPEAKERAcknowledging the problems they've had and saying here. Here's what we're doing to make this work better
50:34SPEAKERIs that a do you think that's a fair read or is that oversimplification Adam? Yeah. Well once again like like
50:41SPEAKERYes, you could easily
50:43SPEAKERMake fun of Intel's problems in the past, you know and being able to deliver stuff
50:48SPEAKERAnd there's a lot of people out there who who look forward thinking and they're just like oh man. This is a
50:54SPEAKERYou know Intel they're they're they're talking a big game and who knows if they're gonna be able to deliver on this this roadmap
51:00SPEAKERBut Pat was on stage and he's like hey listen like the thing that I laid out last year is is in place
51:06SPEAKERWe're rolling forward
51:08SPEAKERAnd I'm sure I'm sure
51:10SPEAKERIn a different parallel universe they wanted meteor like desktop to be alongside here
51:17SPEAKERAt innovation and talk about it, so I'm sure I'm sure there's some sort of
51:22SPEAKERThings that are going on that that's not a core according to plan, but I mean that's business. I did it doesn't surprise me
51:29SPEAKERI guess what CPUs what CPU designs are what a five or ten year ten year development cycle at this point probably
51:35SPEAKERI I don't I have no idea how long that stuff takes at this point
51:38SPEAKERYeah, I mean there's so much forward-facing stuff like I said there
51:41SPEAKERThey they kind of talked about in terms of like four years like that usually that's kind of where they're talking about in that now
51:48SPEAKERthey're talking about the the kind of newer generational technology stuff like the glass core substrates and
51:55SPEAKERWhere the power via like different power delivery methods things like that so I mean that they are talking about it
52:01SPEAKERAnd they're looking forward
52:03SPEAKERAnd kind of being you know like hey check this out
52:06SPEAKERWe're doing awesome stuff, which I mean if it works, and it comes out on time cool. It is awesome stuff
52:13SPEAKERSo dude can we dig a little bit
52:15SPEAKERI know you I know you all Gordon's role always has been that anything that they're not shipping is is
52:23SPEAKERBullshit until they start shipping it kind of not not that's that's unfair
52:27SPEAKERBut he's been skeptical of things that are on paper and not on silicon. I guess is what I would say
52:32SPEAKERI mean, there's a whole cottage industry of
52:36SPEAKERPeople who like to look to the future and leaks and and roadmaps and try to you know make assumptions and things like that
52:43SPEAKERYeah, definitely tea leaves. Yeah on the PC world side. It's not just Gordon like I'm
52:50SPEAKERI don't really care too much about like leaker culture and things like that Brad
52:54SPEAKERBrad charcus who's our managing editor who kind of runs the day-to-day stuff on the the website side
53:00SPEAKERHe's kind of in the same, but we don't we don't really get in to a lot of rumors unless it's like okay
53:05SPEAKERWell this rumor is like oh hey this new GPU is probably coming out because it's one of the the AI bees
53:11SPEAKERAccidentally leaked it on their website like that's like yeah, okay, you know that that's something you can hang your head on, but yeah
53:16SPEAKERWe we typically like to just look at something when it comes out and say okay. What is it? How is it sure?
53:24SPEAKERMaybe it could have been different in a different way, but it doesn't matter
53:27SPEAKERThis is what you're gonna be buying off the shelf is this how you're gonna be using it and just evaluate it like that
53:33SPEAKERI mean just to be clear this isn't leaks
53:35SPEAKERThis is stuff that Intel stood up on a stage and said hey, here's what we're sure the next four or five years
53:41SPEAKERBut I mean it's it's usually it's paired with leaks of like hey
53:44SPEAKERYou know what like I I have some inside sources in their fabs, and they're they're running into problems here
53:48SPEAKERAnd they're you know so yeah the roadmap says this, but you know production is not all got it. You know any
53:55SPEAKERThe big thing to me out of the long-term stuff that I read about out of this event is the UCI e stuff
54:02SPEAKERWhich is their?
54:04SPEAKERuniversal
54:05SPEAKERHold on. It's universal chiplet interconnect Express, which is very conveniently close to PCIe
54:11SPEAKERI guess is the acronym. I'll also if I'm using to see the word chiplet snuck its way in there
54:17SPEAKERI thought something right or professional or formal sounding
54:20SPEAKERAnd not tile it could have been the UTI e but yeah
54:24SPEAKERNobody wants a UTI
54:26SPEAKERThank you
54:28SPEAKERSo I've got the the list. This is a full-on consortium. We love those around here other consortium. It's everybody
54:34SPEAKERIt's like will you were asking or actually sorry you were in the show notes is AMD on board with this
54:40SPEAKERThe members of this this group are AMD arm
54:43SPEAKERGoogle Intel meta Microsoft Qualcomm Samsung TSMC and a SE group, so I think it's oh
54:51SPEAKERYes, actually it's the Nvidia's the well of course. They don't make CPUs exactly well
54:55SPEAKERThey will try to and then the UK said hey, you can't buy arm. You can't make arm CP. You can't buy arm yes
55:03SPEAKERThis was this was one that Intel rolled out and just built and then just hand it off to the group right isn't that?
55:09SPEAKERIs that is that how this one worked? It seemed like I can't remember I?
55:14SPEAKERPersonally don't have like too much knowledge into this stuff
55:17SPEAKERBut yeah like it sounds like with a lot of a lot of different consortiums usually something major
55:24SPEAKERIs it is handed and said hey? This is what we came up with you know? What do we want to do with it?
55:29SPEAKERSo yeah, I think Intel was kind of leading the charge on that end this so this may be too in the weeds
55:34SPEAKERI mean maybe some hardware engineers could speak to this on our discord or something, but I like I wonder
55:38SPEAKERWhat is the purpose of a standardization in the space because it's not like these these companies are?
55:45SPEAKERContributing different chiplets to the same design like everybody everybody's
55:49SPEAKERShip exists unto itself right so like well TSMC and Intel obviously that's like that. They need to be working together
55:56SPEAKERI mean, they're working together here in meteor like I mean
55:59SPEAKERBut I assumed in that case that's Intel designing and providing specs to TSMC and just saying hey fab this not you know
56:06SPEAKERLike I wonder like how much interoperability does there actually need to be or standardization in something where each product is kind of bespoke
56:13SPEAKERfrom each company
56:15SPEAKERWell, I guess it's not an interoperability thing like I don't you know
56:19SPEAKERBut that's just like I think it's just more like oh hey. This is this is something that can benefit everybody
56:25SPEAKERAnd I think okay. Yeah, that may be the case. I may be missing some
56:30SPEAKERIt's it's if you look at what happened with arm they have a they have a similar like there's a there's a protocol by which you
56:37SPEAKERAre all arm CPUs can talk to the other bits of the of the SOC?
56:40SPEAKERWhich is why Qualcomm and Apple and Motorola and 15 other companies can make modems for for arm arm phones
56:47SPEAKERRight this this gives them this this fixes that SOC problem for x86
56:55SPEAKERIn a way that arm is it's just been table stakes for arms since the beginning of arm
57:00SPEAKERAnd I like me that seems really interesting. It's interesting that arms a signatory on this too, so
57:06SPEAKERQualcomm like there's definitely some some
57:09SPEAKERnews that the Qualcomm is excited to get out there about
57:12SPEAKERgetting
57:13SPEAKERLike better laptop x86 laptop and performance like Gordon's definitely been like somebody to kind of put down
57:22SPEAKERQualcomm chips on on Windows for a while
57:25SPEAKERI mean, it's it's it's a good story
57:27SPEAKERBut it's also one that they haven't executed on very well, so you know I I don't think they're getting out of it
57:33SPEAKERIt sounds like they really want to kind of push forward and and be a big player there, so I mean yeah
57:38SPEAKERI I should have asked this before
57:40SPEAKERBut does this mean that meteor lakes gonna have a v1 support if you have an arc see dark GPU on the die
57:46SPEAKERProbably right yeah
57:48SPEAKERThat's pretty exciting
57:50SPEAKERBecause nobody there's not a lot. I mean AMD does it our
57:54SPEAKERGPUs do our
57:55SPEAKERAcceleration I think but Nvidia doesn't right now so for people who work in in live video. That's that's thrilling
58:02SPEAKER40 series can encode AV one yeah
58:04SPEAKERThere's AV one on all of the the newer GPUs now. Yeah, yeah, yeah
58:10SPEAKERLet's see
58:11SPEAKERShould we dig into?
58:14SPEAKERWell, okay, so there's this fove rose thing which is is a little further off
58:18SPEAKERThat's lunar lake which is for people keeping track at home meteor lake is next year arrow lake is the one
58:25SPEAKERLunar lake is the one after that and then pant
58:27SPEAKERWhy did I don't know why they like lakes so much?
58:29SPEAKERBut Panther Lake is the is the fourth gen down the line from what's so that's five generations from what's shipping today basically
58:39SPEAKERPanther Lake is far enough off that when I google it it still comes up with the actual Panther Lake first
58:44SPEAKERThe first result is Panther Lake campground road. I mean honestly my favorite was coffee lake
58:49SPEAKERI think you guys can can dig that
58:52SPEAKERYeah, imagine a lake full of coffee a lake full of Panthers does not excite me no that's not bad
58:58SPEAKERYeah, I I just I assume that there's a team of about 40 people at Intel that are just driving around the West looking for
59:03SPEAKERNew lakes they can name stuff after with cool names
59:07SPEAKERYeah
59:08SPEAKERIt's funny. That's another knockoff of
59:10SPEAKERGordon is like like there's there's the code names right which we definitely should get into naming because that's a big topic
59:17SPEAKERBut the code name meteor lake is is you know a signifying sign
59:22SPEAKERBut the more general consumer they don't really care
59:25SPEAKERThey just look at like is the the number bigger which gen is it is a 12th Gen 13th Gen 14th Gen
59:31SPEAKERYou know so like like that's that's why usually on the package. They don't really talk about
59:36SPEAKERLakes kind of thing this is more of a internal kind of discussion well
59:40SPEAKERAnd I kind of wonder like typically when they've made a big change to the architecture like this
59:46SPEAKERThey often drop the old branding and go to new branding like the move from Pentium to core
59:52SPEAKERHappen when they when they went to multiple like that
59:55SPEAKERDid they announce branding for these news that for the meteor lake suffer is it they said it's gonna be 14 14 series
1:00:01SPEAKERNo, no, they all they've given us so far which there was actually a lot of talk around the naming stuff
1:00:07SPEAKERBut all they've given us so far is now. It's Intel core ultra processors
1:00:12SPEAKERYeah, so I think they're dropping the i9 and the i5 and i3
1:00:15SPEAKERSome of that came out in like June. I think they announced like they were dropping the I and everything but like this sound so
1:00:21SPEAKERThey're not putting generational markers on the product naming anymore. Is that correct so so there will be
1:00:28SPEAKERI just don't know exactly how it's gonna be there was talk of
1:00:32SPEAKERSeries nomenclature like like the Xbox like hey, here's core ultra series one. Here's core ultra series two
1:00:39SPEAKERBut in terms of numbering I'm sure they're gonna have some sort of number or they will have some sort of number and supposedly
1:00:46SPEAKERIt'll just start with a one so instead of a you know like a 13 900 K
1:00:51SPEAKERKind of thing like the 13 is is in reference to the generation now
1:00:56SPEAKERIt's just dropping back down to one so this will be first-gen. I guess
1:01:02SPEAKERYou know in a base 10 world
1:01:04SPEAKERThere's only so much you can do like in video also right starts has to start over every decade or so right yeah
1:01:10SPEAKERI mean Samsung phones right did you know it was 1 through 10, and then now it's just the 10 20 30 40
1:01:18SPEAKERKind of thing so I mean naming naming is hard and naming sucks
1:01:23SPEAKERI think that's that's the biggest question coming out of there is like
1:01:27SPEAKEROkay with so many different options and the fact that the generations of each tile can change
1:01:33SPEAKEROr I'm sorry the the process technology of each tile can change and the configurations can change
1:01:38SPEAKERHow the hell are you gonna market this to people I mean cuz core ultra is only one part of the naming
1:01:45SPEAKERLike you need some sort of signifier of what's gonna be in it. You know what what kind of?
1:01:51SPEAKERGenerational technology because I mean at the end of the day people are gonna look at the number and be like okay
1:01:55SPEAKERWell this number is bigger and better
1:01:57SPEAKERSo AMD has kind of done this a little bit on the laptop side they at their event last year
1:02:03SPEAKERThey gave us literally a decoder wheel
1:02:04SPEAKERI wish I had it on me because it's it's a layer
1:02:07SPEAKERIt was like a decoder wheel where it's like hey the first digit means this the first digit meant a year
1:02:13SPEAKERSecond digit meant the the the architecture selects in three years and two you know so yeah
1:02:20SPEAKERThat might be that might be the most self-aware marketing tchotchke
1:02:23SPEAKERI have ever heard of but but also like this all comes out
1:02:27SPEAKERThis is like the last vestige the last best bad vestige from computer nerds starting out as car nerds
1:02:34SPEAKERWhere everybody's like man?
1:02:36SPEAKERMercedes really does this right you know that the three series is the cheap one and the six series is the expensive one and like
1:02:42SPEAKERThen the whether it's the two or the 30 or the 35 or what like oh god
1:02:47SPEAKERIt's just it's too complicated for that. Just just make a sane naming scheme, man
1:02:52SPEAKERYeah, I mean the the the least complicated thing. I think was to drop the I like like yeah
1:02:59SPEAKERIt signifies something for you know historical purposes, but also at the same time. Do you really need the I there?
1:03:04SPEAKERIt's not like there was a different letter before a number. I mean though
1:03:09SPEAKERYeah, cuz they're like and like the i-5 i-3 i-7 i-9
1:03:14SPEAKERUsed to like it when they first rolled that out that meant something specific in terms of like what was on that like the i-5s had
1:03:21SPEAKERgraphics processors and i-7s were unlocked and and like there were there were there were meanings to those things and then those gradually eroded away until
1:03:29SPEAKERIt just meant that the nine was the expensive one of the three was the cheap one
1:03:32SPEAKERYeah, the five and the seven were in between so now and now it's like what is it the?
1:03:38SPEAKERK's are unlocked the F's have no GPU is that rights yeah?
1:03:42SPEAKERWas it the I think is yes right and then but they're also s is like a spec bump
1:03:45SPEAKERBut right wait can the K's get F'd or is it is the case?
1:03:50SPEAKERYeah, okay, right yes, so I mean when you're just when you're just layering hierarchical naming nonsense on top of your thing
1:03:57SPEAKERMaybe it is time to start over. Yeah, but that's the thing like I like they're gonna start over and it's gonna be
1:04:04SPEAKERComplicated as hell. I don't I don't think
1:04:07SPEAKERNo, nobody there made me feel like
1:04:10SPEAKERLike oh, yeah, we've got it under control. It's gonna make crystal-clear sense so
1:04:16SPEAKERIt's just gonna be different
1:04:18SPEAKERWell, I mean it seems like it seems like with with that many variables on the package now
1:04:22SPEAKERIt don't is is on them to keep the roadmap like focused and sensible and understandable right like well
1:04:27SPEAKERI mean they were the way they were talking about it or at least Michelle Johnson hold house
1:04:32SPEAKERYou know I mean she's kind of that's one of her main jobs is to figure this out
1:04:37SPEAKERThe way she was talking about I was like hey, you know like you just buy it for your use case though
1:04:42SPEAKERIt sounds like maybe there'll be a line for gaming. Maybe there'll be a line for content creation
1:04:46SPEAKERMaybe there's a line for low power. You know I don't know she once again
1:04:49SPEAKERThey haven't like actually given anything other than core ultra, but maybe there's not an ultra
1:04:56SPEAKERyou know maybe there's like core is maybe a signifier of
1:05:00SPEAKERSomething maybe ultra is a signifier of something. I we just don't know exactly what that means yeah
1:05:05SPEAKERIt could be that like the core means the SOC and the ultra means the core one means the first gen SOC
1:05:13SPEAKERAnd then the ultra means it has a GPU and then extreme means it has the the gamer
1:05:18SPEAKERCPU the gamer compute tile and then plus plus means it has
1:05:23SPEAKERDouble IO for faster lanes so you what you have is a 23 digit long character
1:05:29SPEAKERSerial number and if you look at the four character blocks
1:05:31SPEAKERThen the hex will tell you exactly which one is inside it is simple and even a baby could do this can I can I?
1:05:37SPEAKEROkay, I just wanna make sure I got like monitors core one ultra extreme plus plus is that what we've got. That's that's my bet
1:05:43SPEAKERThat's my that's what I'm thinking. I like me. I like we know what but just hire us until it's worth
1:05:48SPEAKERIt's worth noting one plus is spelled out, and the other one is the care the plus sign aha
1:05:55SPEAKERI'm just amending for this podcast title alright podcast title now. Yeah, you're welcome
1:06:00SPEAKERSpeaking of stupid names can we talk about foveros for a minute?
1:06:03SPEAKERThis is another thing on their long-term roadmap that was looked neat
1:06:06SPEAKERAnd I watched a two-minute video and I was like oh that makes sense and then as soon as the video stopped
1:06:10SPEAKERI was like I don't have any idea what I just saw
1:06:12SPEAKERBut I think that it means that the chips are stacked on top of each other now
1:06:16SPEAKERYeah, yeah, I mean the like the ability
1:06:19SPEAKEROkay, so there's a lot of talk a lot of people love to keep talking about all Moore's law isn't dead
1:06:24SPEAKERAnd then some executive will come on and either be like yeah Moore's law is dead, or it's not I mean
1:06:30SPEAKERThere's they're just trying to figure out different ways to package this stuff shrink
1:06:35SPEAKERthe the actual processes and have it be stable and
1:06:39SPEAKERYeah, there's a lot going on foveros is one of those power via which is a power delivery method is another thing like I said
1:06:46SPEAKERThe glass substrate is something they just kind of started talking about
1:06:50SPEAKERSo like there's different different ways to kind of package this stuff up that is very forward-thinking
1:06:56SPEAKERoverall says a name I mean, yeah, whatever like I think
1:07:01SPEAKERYeah, yeah panther like seems like a crappier name to me
1:07:05SPEAKERBut you know it's whatever it's a different thing over. Oh seems like it should be a visual learning
1:07:11SPEAKERAI company or something anyway, yeah, so okay, so we don't need to worry about too many. That's for the future, right?
1:07:17SPEAKERIt's for the future. Yeah, it's the way things are gonna be
1:07:20SPEAKERput together really like they need to figure out and then I mean same thing with the was it uh the
1:07:27SPEAKEREV you or whatever the look oh, thanks violet
1:07:32SPEAKERUV extreme extreme UV. I'm sorry
1:07:35SPEAKERThe the idea of hey instead of instead of blasting ultraviolet is four times now
1:07:41SPEAKERWe only have to blast it once there's a different you know and that's gonna have implications the to cut down on
1:07:49SPEAKERLike you know things that go wrong and and things like that so it's very complicated supposedly that actually has to happen in a vacuum
1:07:56SPEAKERTo like and there's only one company that's making these these kind of things
1:08:01SPEAKERSo like they got to see that at in Malaysia
1:08:04SPEAKERI so I don't really have a ton of knowledge into that, but that seems like all those crazy shit
1:08:12SPEAKERYeah, anytime anytime they're talking about like hey
1:08:15SPEAKERWe have to we have to come up with a new way to do this process like this
1:08:20SPEAKERWe're inventing new technology for this process for this process to work. That's when you start thinking
1:08:25SPEAKEROh, this is one that's either gonna be hard to get at launch, or they're gonna. Maybe push this account
1:08:29SPEAKERI guess by the time they start talking about it
1:08:31SPEAKERThey're pretty committed to it, and they're not gonna change, but other than that. It's often. It's often like wild stuff
1:08:37SPEAKERThat means it's gonna be hard to get the new these CPUs when they land so I mean
1:08:41SPEAKERI started to I started I started to making like a sneering comment about the angstrom era
1:08:47SPEAKERBut now that I'm looking at their 18a process
1:08:50SPEAKERIt's like probably we're approaching the sub nanometer era like legitimately right so like
1:08:55SPEAKERTechnology is probably going to have to get insane to accomplish that right like they're the the Intel 18a process is 1.8
1:09:01SPEAKERnanometers apparently I mean that that is the crazy stuff like for me like somebody who doesn't have a background in electric
1:09:08SPEAKERYou know engineering anything like you know all this stuff is like way over my head
1:09:12SPEAKERBut for me like the fact from the outside seeing that so many people are trying to figure out different ways to like continue to
1:09:20SPEAKERPush technology forward like it just seems exciting if anything like to even if I don't fully understand it
1:09:27SPEAKERI can look at it and be like oh my god like people seem to really be trying to figure this shit out
1:09:31SPEAKERWell, I mean it's it's it's worth remembering that Intel
1:09:35SPEAKERBuilt a you know a multi-billion dollar company on being real good at process technology for 50 years, right and and
1:09:43SPEAKERLike they that's why they they always invested a lot in process technology TSMC and other companies obviously do too
1:09:49SPEAKERBut but like the company that figures this stuff out first
1:09:53SPEAKERWins in an enormous way because at the end of the day the performance is what matters to the vast majority of the market so
1:10:02SPEAKERThat's right. It's really fun. Thanks. Thanks for coming by and talking about this stuff Adam. We've
1:10:07SPEAKERI've definitely learned a lot
1:10:09SPEAKERNo, I know. Thank you. I once again. I if anything. I'm just I'm just excited this just feels like a different
1:10:16SPEAKERwe're stepping into a new age of computing and even though they
1:10:20SPEAKERThe we can't really see we're kind of looking off in the distance and being like okay. What does this mean?
1:10:25SPEAKERHow is this gonna change things like I really think like things are gonna look different like computing
1:10:31SPEAKERFive years from now is gonna look way different than now. It's it's yeah, it's exciting
1:10:36SPEAKERIt feels like there have been multiple parallel paradigm shifts taking place both at the both of the micro like in CPUs
1:10:42SPEAKERyou know you've you've had big and little cores and also the modular chiplet stuff happening kind of simultaneously and then like I'm like
1:10:49SPEAKERbigger picture or like you've got the machine learning acceleration coming in like there's still like
1:10:53SPEAKERhybridizing between local and cloud computer going on like I really
1:10:57SPEAKERWonder what the hell computing looks like 10 or 15 years from now in terms of like use cases that nobody has even thought of yet
1:11:03SPEAKERYeah, and it makes it makes our job or I should say Gordon and I's job even harder to be like, okay
1:11:09SPEAKERWell, how do you how do you even benchmark this stuff?
1:11:12SPEAKERright, I mean like one of the things that Gordon I've been talking about for a while is like I
1:11:16SPEAKERMean who even I mean the amount of people who even care about cinebench performance to like pound every core is like very minuscule
1:11:23SPEAKERRight, like I mean most of the people using computers
1:11:26SPEAKERIt just it just works and they don't need to a ton of a ton of performance kind of stuff
1:11:31SPEAKERSo like okay, how do we how do we figure this out? How do we benchmark it? I mean, we can't even figure out reliable
1:11:40SPEAKERApplication benchmarks we're still stuck here, you know for better or worse. We're still stuck in like hey you you literally configure it
1:11:48SPEAKERSo you're just benchmarking one application at a time and in showing the performance of that one application
1:11:54SPEAKERThat's not how people use computers anymore. And now we're gonna be even further behind because now there's different
1:12:00SPEAKERdifferent, you know a new processing unit in there that's happening at the same time and like there's there's no standard benchmark around that either like
1:12:07SPEAKERThere's no cinebench for NPU performance. Really? Well, yeah, it's interesting because as I've been doing more game even like game development stuff
1:12:14SPEAKERWhich is pretty light. I mean, I
1:12:17SPEAKERMean, it's as light as anything could be that you have an unreal editor open I guess but like I'll have I'll have Photoshop in
1:12:23SPEAKERAudition and wise and and unreal ed all open at the same time and like none of those are light
1:12:31SPEAKERApplications, I mean Photoshop is the lightest of those at this point probably which is which is kind of remarkable
1:12:38SPEAKERand and
1:12:39SPEAKERLike you it's to the point now that when I run that stuff on my old machine on the i9 99 her K
1:12:44SPEAKERWhich is not a not an insubstantial computer from like four or five years ago
1:12:49SPEAKERIt's it feels slow. Like it is a it is a there's a noticeable difference. So there is benefit to upgrading to newer machines
1:12:56SPEAKERIt's just I I don't envy you all having to figure out how to benchmark stuff with a 32 cores
1:13:00SPEAKERIt seems like a nightmare
1:13:02SPEAKERwell, I mean and it used to be I mean for me on the video side like I would I would keep a video to to
1:13:08SPEAKEREncode and then I'd be like, nope, don't touch it. Don't touch it. Just let it do its thing
1:13:12SPEAKERI'm gonna go off do something else now
1:13:14SPEAKERI'm like, oh, yeah, I can I can make a thumbnail while I'm encoding while I'm also, you know
1:13:18SPEAKERListening to a podcast and playing a YouTube video like like people can do so many
1:13:24SPEAKERmultitasking stuff that like I still feel like benchmarking has
1:13:29SPEAKERIs still kind of in like the like a different era like of when?
1:13:34SPEAKERYeah, everything could hit hit your system at once and he didn't really want to mess with it now. It's like
1:13:39SPEAKERYeah, well, well and we we always at maximum PC
1:13:43SPEAKERWe always tended to prefer real-world benchmarks, which is why like cinebench was a useful things
1:13:48SPEAKERIt would actually measured, you know video edit and encode nonlinear video edit and encode and like color grading and all those kinds of things
1:13:57SPEAKERWhereas like phone phone benchmarking has which is has a similar series of problems, right? It's mostly about experience
1:14:04SPEAKERIt's it's all single application at a time
1:14:06SPEAKERBut you're comparing multiple different architectures and multiple different OS's across all sorts of different
1:14:13SPEAKERThere's too many variables to do apples to apples benchmarking. So you end up in a situation that doesn't stop people from trying though
1:14:21SPEAKERSay geek benches the benchmark for Twitter, I mean, that's that's how you end up with stuff like geekbench
1:14:28SPEAKERThat is like it give it spits out a number and sometimes one number is bigger than the other
1:14:34SPEAKERWhat does it actually mean? I have no idea man. It's it's like it's like what is a 3d mark?
1:14:39SPEAKERWell a 3d mark 13 is this much faster than a 3d mark 14? I know that but beyond the like
1:14:45SPEAKERit's a hard situation and it's not gonna get easier unfortunately, especially as like
1:14:50SPEAKERLike how do you compare this to M to apples M silicon on their notebooks?
1:14:55SPEAKERand how do you benchmark an Intel laptop versus an AMD laptop versus a Mac when you can't run the same OS on them and
1:15:02SPEAKERGood luck. Adam is I guess what I'm saying. You guys are doing great over there
1:15:06SPEAKERYeah, well, thanks. And I mean a lot of people you have a lot of people who want to
1:15:11SPEAKERGet really deep into the benchmarks and figure out like okay
1:15:14SPEAKERWhat's what's my best bang for my buck or you know?
1:15:16SPEAKERLike so because of these costs the cost of all these parts are so high people are
1:15:21SPEAKERTrying to get more out of them and they want to they want to feel like they're they're getting the best that they can
1:15:26SPEAKERBut also at the same time like people who are just building i3 systems are like, oh, well, this is great. I don't
1:15:33SPEAKERI'm having no problem
1:15:35SPEAKERthe one thing that is true and has remained true for the last decade at this point is that
1:15:41SPEAKERGenerally if you buy a newer piece of hardware at the same price point as your old piece of hardware
1:15:46SPEAKERIt the new one's going to be faster and better
1:15:49SPEAKERPretty pretty consistently across the board
1:15:51SPEAKERLike if you buy if you have a mid-range i5 from five years ago and you replace it with a mid-range i5 from right now
1:15:57SPEAKERYou're gonna have a much better machine than you had before
1:16:00SPEAKERFor probably around the same money maybe less if you count and if you if inflation matters to you, so all right, I don't know
1:16:09SPEAKERfun times yeah
1:16:11SPEAKERAdam where can people find you PC world calm youtube.com forward slash PC world. We have a weekly podcast called the full nerd
1:16:17SPEAKERI am stepping in for hosting duties
1:16:22SPEAKERBrad was on recently. Will we need to get she on again?
1:16:27SPEAKERLove chatting with you and Brad. Yeah
1:16:30SPEAKERWe were over on the YouTube channel. I mean we just I'd like to have fun. Honestly, we we do a lot of
1:16:37SPEAKERDumb experimentation stuff. We do some serious benchmarking as well. I was a couple live builds coming up
1:16:43SPEAKERI like just hanging out and building stuff live. I'll say
1:16:47SPEAKERWhen I built my first system, which I mean this this will be an indicator of how new I am compared to you guys
1:16:53SPEAKERbut I built my first PC in
1:16:56SPEAKER2011 probably
1:16:58SPEAKERAnd when I was building it, I just sat down and watched
1:17:02SPEAKERThere was a tested video
1:17:04SPEAKERThere's actually a giant bomb video as well where you just you know recorded the whole thing
1:17:08SPEAKERjust just going through and and like that's what I like doing is like hey, I just like hanging out and doing live builds cuz
1:17:15SPEAKERSome people just use it to hang out and have fun
1:17:18SPEAKERbut then some people actually do use it as a way to build in real time and and kind of
1:17:22SPEAKEREven though it's not like a how-to a tutorial anything like that
1:17:25SPEAKERIt's still fun to just hang out and kind of see well the thing that I like about your live builds and just to be
1:17:30SPEAKERClear Brad and I both have built computers on your stream on your YouTube channel at this point
1:17:35SPEAKERIt's fun time. I'm gonna do it again. Well, yeah, I mean god, I don't not for a while
1:17:41SPEAKERBut you like it's good to show the mistakes right like everyone YouTube gets really angry when you show mistakes, but also
1:17:50SPEAKEReverybody makes mistakes even people who are really good at this and do it all the time and
1:17:54SPEAKERAnd if you if when you don't show the mistakes on YouTube
1:17:58SPEAKERThen it makes it makes it harder for new people to get into this because they think hey
1:18:04SPEAKERI've messed this up. I just spent a thousand dollars on this thing and now it's not working. I'm doomed but literally I
1:18:10SPEAKERMean, okay, so it doesn't travel all that but but most people when they turn when they power up the machine the first time
1:18:16SPEAKERIt's not gonna boot right? There's gonna be so you didn't plug something in right?
1:18:19SPEAKERYou don't have your memory pushed in all the way whatever it is. It's not gonna boot
1:18:23SPEAKERIt's not a big deal you turn the power off you take stuff in and out you plug it back in you
1:18:28SPEAKERMake sure all your cables are in the buttons hooked up and all that and then the second time usually
1:18:33SPEAKERYou get a little more love
1:18:34SPEAKERIt's not the end of the world if it doesn't work and and just knowing that even people who do it all the time
1:18:39SPEAKERMake mistakes is comforting and makes it makes it more approachable more accessible for everyone
1:18:44SPEAKERI think so. I love what you all do over there. I made mistakes all the time
1:18:47SPEAKERWe made mistakes with your builds. Well, yeah, we're like, oh crap the radiators not gonna fit
1:18:54SPEAKERYeah, we made it fit we made it
1:18:56SPEAKERYeah, we got it in there
1:18:58SPEAKERYeah, so it's it's a it's a you guys do really good stuff over there and people should check it out
1:19:04SPEAKERThank you. And we really appreciate you taking the time to come by and and
1:19:08SPEAKERI hope we'll have to have you on again sometime soon Adam. This has been great. It's an honor
1:19:11SPEAKERYeah, I love what you guys do and and listen to almost every podcast that comes out on the feet
1:19:16SPEAKERSo keep doing what you're doing. Thank you so much Adam
1:19:21SPEAKERAdam thank you so much for coming on the show. It's wonderful to have you and again
1:19:24SPEAKERYou can find him at PC world comm or youtube.com slash PC world. Yes
1:19:28SPEAKERI'll I'll put some some links to some of their innovation videos in the show notes for this episode
1:19:33SPEAKERYeah, there were they had they had a lot of good stuff. I've been it's funny
1:19:36SPEAKERI've been dipping into PC world videos recently to looking at
1:19:41SPEAKERInterviews with some of their memory vendors from copy text earlier this year for reasons that we might talk about in the not-too-distant future
1:19:49SPEAKERKeen eyed keen eared listeners
1:19:51SPEAKERI guess might realize that this was the last Sunday of the month which we didn't realize until about five minutes ago
1:19:56SPEAKERSo we'll do Q&A next week fear not you can send your questions into tech bod at content on town or post them in the questions
1:20:02SPEAKERSeeking answers channel. Yes the discord we discovered next month or maybe this month. No, no next month next month
1:20:09SPEAKERSorry next month. I forgot what day it was next month is a five Sunday month
1:20:12SPEAKERMeaning it will now be a two email month double double Q's means twice the A's
1:20:19SPEAKERBut as always we reach the end of the show where we thank our patrons for supporting the podcast without the patrons
1:20:25SPEAKERThis podcast wouldn't be possible. So
1:20:27SPEAKERAs always we want to start by thanking everyone. Thanks patrons. We appreciate you. Thank you patrons special
1:20:34SPEAKERThank you goes out to our executive producer to your patrons including Andrew Slosky bunny fiend
1:20:39SPEAKERpaddle Creek games makers of fractured veil David Allen James kamek Joel Krauska
1:20:45SPEAKERJordan Lippett just wedge Nick Johnston twinkle Twinkie and
1:20:50SPEAKERPantheon makers of the HS 3 high-speed 3d printer. Thank you all so so much
1:20:55SPEAKERYes, thank you. And since it is the last episode of the month, we also thank our associate producer to your patrons including Arthur Geese
1:21:03SPEAKERAlejandro Navarro Andre M Burke then Talman Dave Julian Eric Eric Frey
1:21:09SPEAKEREric Klein still only three Eric's I'm hoping for a fourth Eric one day, but mate we'll get there
1:21:14SPEAKERFelix Kramer Graham Banks Jad Rita Jason Rooney the ref Matt Walker parentheses Walkman 8080 close parentheses
1:21:23SPEAKERNathan Phelps Sanchik Kumar Steve Lynn Thomas Shay and Tom Hilton. Thank you all so much. Thank you everyone
1:21:30SPEAKERWe got to increase our Eric rate
1:21:32SPEAKERWe look if we aren't at least I look I want at least 10% Eric's next month in the associate producer tier
1:21:38SPEAKERThat's where Eric's we need you and I've got to boost the Eric stepping around here
1:21:43SPEAKERBut we do appreciate everybody who supports the show
1:21:45SPEAKERif you want to find out how to do that you can go to patreon.com slash tech bod and for five bucks a month you
1:21:50SPEAKERGet access to the fabulous patron episode where we're going to reminisce on
1:21:54SPEAKER200 and change episodes of the pod that's right this month as well as I assume talk about well
1:22:00SPEAKERWe would say we're gonna talk about brands to be I think we might blow that out to a whole episode or part of a whole
1:22:04SPEAKEREpisode. Yeah, maybe not a whole episode
1:22:05SPEAKERBut it will we'll definitely talk about it in a in a public episode and not put it behind the the patron wall because we have
1:22:10SPEAKERDefinitely had some some requests from people to talk about whether I like it or not Brad likes TV
1:22:16SPEAKERWe'll find out soon. Can't wait to tell you about the boot loop that it got stuck in yesterday
1:22:20SPEAKEROh, I love a boot loop. That sounds fun. It's it is it did you get a Canadian TV? No, it's a Google TV
1:22:26SPEAKEROh, actually my TV runs Android. Oh, that's I don't know about that
1:22:31SPEAKERBut we will be back next week with another episode of the tech pod. Thanks everybody for listening. Thanks Adam for coming by
1:22:36SPEAKERWe would really appreciate it and we will see you all next time on Brad will made a tech pod