WEBVTT 00:00:00.080 --> 00:00:07.600 Do you remember that one time when Intel launched the world's first quad core 00:00:03.919 --> 00:00:10.400 consumer CPU and then 10 years later was 00:00:07.600 --> 00:00:14.480 still doing the same thing? Okay. Yes, yes, I know there were plenty of other 00:00:12.400 --> 00:00:18.880 innovations along the way, but that doesn't change the fact that Intel's 00:00:16.240 --> 00:00:24.400 long-term stagnation did create an opening for AMD to not only continue 00:00:21.119 --> 00:00:26.960 existing, but against all odds to close 00:00:24.400 --> 00:00:31.039 the technology gap, taking serious market share from a company that up 00:00:28.960 --> 00:00:35.600 until then was functionally a monopoly in both the server [music] and the 00:00:32.880 --> 00:00:40.960 mobile space. Now, that level of stagnation, that's not where AMD is 00:00:37.840 --> 00:00:42.480 today. But if their new product 00:00:40.960 --> 00:00:47.680 announcements today are anything to go by, the warning signs are starting to 00:00:45.040 --> 00:00:52.160 show. In no particular order, we've got a darn near 1:1 rebrand of their last 00:00:50.000 --> 00:00:57.440 generation laptop chips. A couple new entries in the incredible but obscenely 00:00:55.120 --> 00:01:03.039 expensive Stricks Halo chip family, a better bend refresh of the 9800 X3D 00:01:00.320 --> 00:01:07.760 called the [music] 9850X3D, and some AI performance and usability 00:01:05.360 --> 00:01:12.479 improvements. I mean, none of these are bad things, but the complete [music] 00:01:10.240 --> 00:01:17.290 lack of any pricing information is a little ominous and raises another 00:01:14.799 --> 00:01:20.720 uncomfortable bit of news that AMD uh [clears throat] didn't [music] mention, 00:01:18.799 --> 00:01:28.320 the rumored price hikes that are coming on consumer GPUs. Let's talk through all 00:01:23.759 --> 00:01:31.040 of that after an uncomfortable segue 00:01:28.320 --> 00:01:35.119 to our sponsor. Grammarly is an AI writing partner that helps you go from 00:01:32.880 --> 00:01:39.119 draft to done with clear, polished writing that still sounds like you. It 00:01:37.439 --> 00:01:43.759 supports everything from emails to documents, improving clarity, tone, and 00:01:41.360 --> 00:01:48.320 flow while keeping your voice intact. Save 20% off Grammarly Pro by clicking 00:01:45.759 --> 00:01:52.640 the link in the video description. As a wise man once said, when you can't 00:01:50.399 --> 00:01:58.079 innovate in engineering, innovate in marketing. The unwashed masses who buy 00:01:55.040 --> 00:01:59.840 our won't know the difference anyway. 00:01:58.079 --> 00:02:04.719 Okay, that's [music] not entirely fair. To AMD's credit, there are some 00:02:02.000 --> 00:02:07.520 improvements in the new Ryzen AI 400 series. [music] 00:02:05.680 --> 00:02:13.120 Everything highlighted in gold here is new, but the part that they leave off is 00:02:10.560 --> 00:02:18.959 how much new they are. [music] Maximum boost clocks of 5.2 GHz on a mobile CPU. 00:02:16.800 --> 00:02:25.280 I mean, that's incredible. It's just that that's only.1 GHz more than the 00:02:22.239 --> 00:02:26.959 Ryzen AI 300 [music] series, which 00:02:25.280 --> 00:02:32.000 wouldn't be an issue. I mean, clock speed's not everything except that they 00:02:28.959 --> 00:02:35.280 are also reusing the same mix of Zen 5 00:02:32.000 --> 00:02:36.879 and Zen 5C cores. [music] So, when 00:02:35.280 --> 00:02:41.040 you're talking about mobile processors that are often limited by their cooling 00:02:38.640 --> 00:02:46.000 solutions and power delivery anyway, no amount of gold text is going to make a 00:02:43.360 --> 00:02:50.800 2% difference in boost clock noticeable day-to-day. The bump [music] in LPDDR5 00:02:48.480 --> 00:02:57.519 speed is welcome though, especially for applications that rely on the GPU core 00:02:53.040 --> 00:02:59.200 that is also faster, albeit just 7%. And 00:02:57.519 --> 00:03:04.480 it's not going to hurt to be able to run local AI slightly faster. With all that 00:03:02.480 --> 00:03:10.560 said, the devil's in the details. And while AMD's slide is technically true, 00:03:07.680 --> 00:03:15.360 it's only true for the top skew. And as we make our way down the lineup, the 00:03:12.400 --> 00:03:19.440 generational upgrades slowly fall away to make room for what appear in some 00:03:17.360 --> 00:03:25.519 cases to [music] be generational downgrades. The Ryzen AI7445, 00:03:23.120 --> 00:03:31.760 for instance, [music] gets lower clocks, less cache, and the same RAM speed 00:03:28.400 --> 00:03:33.519 compared to the Ryzen AI 5340. And this 00:03:31.760 --> 00:03:39.440 comes at the same time that Intel's Panther Lake is rumored to deliver GPU 00:03:36.000 --> 00:03:43.360 performance gains of like 50%. 00:03:39.440 --> 00:03:47.040 That's a yikes. Next up is the 9850 X3D, 00:03:43.360 --> 00:03:49.680 a truly impressive looking 5.6 GHz 00:03:47.040 --> 00:03:56.720 desktop chip that promises to once again steal AMD's gaming performance crown and 00:03:52.400 --> 00:04:00.080 give it to AMD. But again, probably not 00:03:56.720 --> 00:04:02.159 by as much as we'd hope or as much as 00:04:00.080 --> 00:04:11.280 was rumored. See, the internet's been ablaze with rumors of a 9950x3D2, 00:04:06.319 --> 00:04:14.480 a dual 3DV cache, 16 core, 200 W monster 00:04:11.280 --> 00:04:17.359 with an alleged 192 MGB of level three 00:04:14.480 --> 00:04:21.600 cache. But instead, all we got was a Speed [music] Bend 8 core. Once again, 00:04:19.519 --> 00:04:27.040 this looks like a really nice piece of kit. We get the same rated power as last 00:04:24.000 --> 00:04:29.199 gen along with up to 400 MHz faster 00:04:27.040 --> 00:04:33.520 clock speeds than the old King, which is nice. But um how does that [music] 00:04:31.919 --> 00:04:38.080 translate to real world gaming performance? Well, it's kind of hard to 00:04:35.680 --> 00:04:42.320 say. See, rather than comparing their own chips directly, AMD [music] instead 00:04:40.560 --> 00:04:49.199 compared their new hotness to Intel's Core Ultra 9285K, which is kind of like 00:04:45.440 --> 00:04:51.440 a real boxer fighting Jake Paul. Fair 00:04:49.199 --> 00:04:55.280 enough, AMD. It isn't really my place to tell you how to bully your competitors. 00:04:53.199 --> 00:04:58.028 And come to think of it, it's probably just as well that gamers will need to 00:04:56.720 --> 00:05:03.440 come to me for their benchmarks anyway. [music] Thanks. The Ryzen AIAX Plus 00:05:01.280 --> 00:05:08.240 announcement. Ooh, this one was shaping up to be a real rebranding snoozefest. I 00:05:06.560 --> 00:05:13.360 was sitting there thinking like, really? You guys are going to remind me that Stricks Halo can go in a desktop? I 00:05:11.600 --> 00:05:17.440 literally invested in a company that is doing that framework investment 00:05:15.280 --> 00:05:22.639 disclosure. But then AMD was like, "Boom. We heard that you love 00:05:20.000 --> 00:05:28.240 astonishingly fast PlayStation 5 level graphics in a mobile chip, but don't 00:05:25.120 --> 00:05:31.440 want to pay an extra $500 for the 16 00:05:28.240 --> 00:05:34.880 core model. So, we put the best graphics 00:05:31.440 --> 00:05:36.800 on an 8 core and a 12 core. Man, I've 00:05:34.880 --> 00:05:39.840 been using the GPD Win 5 for the last little while for gaming. This thing 00:05:38.800 --> 00:05:45.039 absolutely [music] shreds Clare Obscure at like epic 00:05:42.240 --> 00:05:50.160 details. No upscaling tom foolery. Like if I could save hundreds of dollars on 00:05:46.960 --> 00:05:51.360 CPU cores that I don't need. Absolutely 00:05:50.160 --> 00:05:55.360 killer. [music] Unfortunately, Stricks Halo is killer 00:05:53.520 --> 00:06:00.000 for your wallet, too, since even the entry level is crazy expensive. But hey, 00:05:57.520 --> 00:06:04.080 cool new product is a cool new product. AMD then talked for a while about 00:06:01.759 --> 00:06:08.240 running near cloud quality local AI models that save you from either the 00:06:05.680 --> 00:06:13.840 limitations of free AI chat bots or the cost of a subscription. The partner they 00:06:10.720 --> 00:06:15.600 showcased was Liquid AI, who has scary 00:06:13.840 --> 00:06:19.520 call to talk about pricing buttons on their corporate licenses, but at this 00:06:17.520 --> 00:06:23.919 time does provide enduser access to their models for free. So that's 00:06:21.759 --> 00:06:28.880 actually pretty neat. But then things got totally off the rails. I'm writing 00:06:26.720 --> 00:06:32.560 this based on my pre-briefing, so it is possible that the public presentation is 00:06:30.800 --> 00:06:37.360 a little bit different in the details. But after the local models, AMD shifted 00:06:35.360 --> 00:06:43.440 to talking about the idea of using iterate.ai AI to feed an AI agent your 00:06:40.400 --> 00:06:45.120 payubs, bank information, tax returns, 00:06:43.440 --> 00:06:50.400 and all that kind of stuff, and then have it generate powerful financial 00:06:47.759 --> 00:06:55.919 advice, saving you the cost of hiring a financial planner. Direct quote 00:06:52.639 --> 00:06:58.000 incoming. You can do it yourself. Now, 00:06:55.919 --> 00:07:02.479 look, I'm not saying you should spend a bunch of money on a financial planner, 00:06:59.759 --> 00:07:08.639 and there are clear privacy benefits to running workloads like this locally. 00:07:05.280 --> 00:07:11.520 But, and to be clear, this is financial 00:07:08.639 --> 00:07:17.919 advice. Do not trust an AI to give you financial advice without also having it 00:07:14.240 --> 00:07:19.840 thoroughly vetted. At least not for now. 00:07:17.919 --> 00:07:23.280 AMD then talked for a while about the strides that they've made recently on 00:07:21.440 --> 00:07:27.199 Rockom with respect to performance improvements and Windows compatibility, 00:07:25.599 --> 00:07:31.919 and they announced a handful of new features, [music] including Oh, okay. 00:07:30.000 --> 00:07:35.840 So, it's just integration with Comfy UI and a simplified install method. To be 00:07:34.080 --> 00:07:39.919 clear, I'm not mad. Those are both good things. It's just don't make me read 00:07:37.599 --> 00:07:44.880 through four enhancements when two of them are just we changed the software 00:07:42.080 --> 00:07:50.479 version number by 0.1 and we have support for the freshly rebranded 00:07:46.880 --> 00:07:53.199 version of last year's AI processor. Why 00:07:50.479 --> 00:07:58.155 is everything called AI? I know I'm just a lowly tech YouTuber, but like 00:07:56.560 --> 00:08:02.160 is this stuff really working? I mean, [music] AMD says it must be working. 00:08:00.080 --> 00:08:07.599 They've seen 2 and 1/2x growth year-over-year for their Ryzen AI 00:08:04.240 --> 00:08:09.199 family. Except that, guys, I'm just not 00:08:07.599 --> 00:08:14.160 convinced that that's why anyone is flocking to AMD mobile chips right now. 00:08:11.280 --> 00:08:19.599 To be clear, I love my Stricks Halo Elite Book 8. This thing is [music] 00:08:16.240 --> 00:08:22.080 freaking awesome. But the co-pilot 00:08:19.599 --> 00:08:26.319 button has nothing to do with why I like it. I like the performance and I like 00:08:24.080 --> 00:08:31.280 the battery life. The AI branding thing has gotten so over the top that at this 00:08:28.800 --> 00:08:35.599 point it's hard for me to even tell what of it is a begrudging, well, we've got 00:08:33.839 --> 00:08:40.320 to shoehorn this into everything or our shareholders will abandon us and what is 00:08:38.320 --> 00:08:44.880 a genuine misunderstanding of what is driving customers to purchase your 00:08:42.320 --> 00:08:49.040 products. To be clear, I don't think anyone's going to be upset in the coming 00:08:46.640 --> 00:08:54.560 years about having a decent NPU so they can generate weird hybrid animals or 00:08:52.240 --> 00:08:58.720 13-fingered French mimes or whatever it is they're into. But I promise you that 00:08:56.640 --> 00:09:02.560 the other parents in my kids classes, they don't care if their PC is AI or 00:09:01.040 --> 00:09:06.880 not, and they don't even know what it means. Then again, they don't use LT to 00:09:05.200 --> 00:09:11.680 keep up with the latest tech, so they wouldn't even know about our sponsor, 00:09:09.120 --> 00:09:16.080 Grammarly. At LMG, writing is a big part of our day, especially mine. From emails 00:09:14.080 --> 00:09:19.680 to large projects, the hardest part is clearly getting ideas down when you're 00:09:17.839 --> 00:09:23.200 on a deadline. That's where you can find us, using Grammarly. It's an AI writing 00:09:21.760 --> 00:09:27.680 partner that helps professionals with all of their writing needs from draft to 00:09:25.279 --> 00:09:32.399 done with highquality personalization so you don't lose your voice. Heading over 00:09:29.519 --> 00:09:36.560 to app.grarly.com lets us use their AI chat to get a first draft on the page. 00:09:34.720 --> 00:09:41.440 It removes the friction of starting and as a muchneeded bonus, it saves us time. 00:09:39.279 --> 00:09:45.440 Once the draft is there, humanizer helps everything sound natural and on tone, 00:09:43.680 --> 00:09:48.880 which cuts down on rewriting. If something needs a tone adjustment, 00:09:46.959 --> 00:09:52.000 Paraphraser can quickly change how it sounds without altering the message. 00:09:50.480 --> 00:09:55.120 With Grammarly, we can finish writing faster, letting us focus on what 00:09:53.519 --> 00:09:58.720 actually matters, like making sure I can catch the next thing that Linus drops. 00:09:56.800 --> 00:10:02.880 And you get to drop 20% off the price of Grammarly Pro subscription by using our 00:10:00.720 --> 00:10:07.360 link in the video description. You guys enjoyed this video? Hey, shout out 00:10:04.720 --> 00:10:11.200 HyperX Arena for clearing aside enough of these boxes that we were able to use 00:10:09.040 --> 00:10:17.360 their shooting space. And uh hey, maybe go check out the ShortCircuit that I 00:10:12.720 --> 00:10:19.440 did on the GPD Win 5. It's maybe not the 00:10:17.360 --> 00:10:26.240 uh most practical application of Stricks Halo, but it is one of the coolest ones. 00:10:22.399 --> 00:10:26.240 This is an amazing little handheld.