Discussing Lunar Lake, Zen 5, Blackwell, and RDNA 4 w/ High Yield! (Guest Telegrams)
Added 2024-06-21 23:57:17 +0000 UTCThe next episode of Broken Silicon will see us be joined by High Yield yet again! We will be heavily discussing Zen 5 vs Arrow Lake, Lunar Lake, Strix, and upcoming GPU launches But of course, you can ask us ANYTHING about recent or future hardware releases (whether PC of console). This is one of those episodes that will be heavily steered by what YOU want us to talk about!
Just make sure you use good grammar, are as concise as possible, and are respectful to have your submissions below considered. :)
Last time High Yield was on BS: https://youtu.be/ZL0AtNEkEL0?si=pG3WwVLMYTJMUkKx
Check out High Yield’s YouTube Channel: 
/ @highyield
Comments
Hello Tom and High yield, How much of a roll do you think software plays in the success of hardware? And how much of this is in their own hands. Can you think of any great hardware destroyed by bad software.
Time Unveiled
2024-06-24 16:38:57 +0000 UTCDo you think Lunar Lake realistically has a chance of perf/watt leadership vs Strix and X Elite (let alone Apple silicon)? I saw one benchmark suggesting efficiency was roughly on par with hawk point. And any early thoughts on what kind of PPA 18A will achieve vs existing N3E?
swarm2958
2024-06-23 18:37:40 +0000 UTCStrix Halo looks like a great product if it performs like rumors suggest (4060 to 4070 mobile for 40cu). Do you think it's going to have wide acceptance in the market if that's AMDs "gaming laptop" offering or do you think people will prefer Nvidia + something else (regular strix, fire range, arrow lake etc). Personally, I don't care about DLSS vs FSR and I think the ability of the AMD solution to use 12-16gb for qhd laptop gaming without having to spend $3-4000 could be compelling.
B. Fish
2024-06-23 14:36:55 +0000 UTCIntel has been rapidly changing their cache systems recently, going from a simple sliced ring L3 in the pre-Alder Lake days to a ring with clustered L2 E cores for Alder and Raptor lake, to bolting on a separate LPE cache domain in Meteor lake, and now with Lunar Lake they are adding a dedicated system-level cache and appear to be making the main L3 only be used by the P cores, rather than being for everything. Where do you think they will eventually settle?
qhfreddy
2024-06-23 04:10:30 +0000 UTCHave you seen anything tangible that can only be done on a system with a 40 TOPs NPU as per Microsoft's requirements or is TOPs just the newest dick measuring contest?
qhfreddy
2024-06-23 03:59:42 +0000 UTCLunar Lake looks great, Meteor Lake did too until it flopped? Any reason to believe Lunar Lake won't flop too?
qhfreddy
2024-06-23 03:59:18 +0000 UTCHow do you rate the chance that intel and AMD go heavily into providing x86 chiplets for third parties? Which one of them do you think is more likely to do it? How will the market receive it? AMD is already has some experience with the console semi-custom lines, but it doesn't feel like they've extended much beyond that. I feel like many third parties would still prefer AMD just because of going through TSMC rather than likely having to do IFS instead. However, if IFS does actually go somewhere then I would see Intel having everything under one roof being an advantage.
qhfreddy
2024-06-23 03:57:34 +0000 UTCWhat do you see happening with the built up advanced packaging capacity if the AI bubble were to burst dramatically? Are there any potentially interesting products you'd want to have made?
qhfreddy
2024-06-23 03:56:54 +0000 UTCAs a pure hardware design, how do you rate the X Elite? What kind of market share would Qualcomm need to achieve a year from now to be considered a success from your perspective? Do you think it has any chance of achieving that?
Chris Rijk
2024-06-22 22:26:39 +0000 UTCDo you get the sense that the rate of improvement in CPUs and GPUs is slowing down? How do you expect the rate of improvement to change in the years ahead?
Chris Rijk
2024-06-22 22:25:48 +0000 UTCWhat’s the biggest positive surprise you’ve seen from CPUs or GPUs launched in the last year or so? What’s been the biggest disappointment?
Chris Rijk
2024-06-22 22:25:40 +0000 UTCHi, Max. I loved your video on backside power delivery and found it incredible that Intel got 30% power reduction for the same architecture with it, which brings me to my questions. Is this production method competitive when it comes to pricing, and will this be the game-changer that brings Intel out of the hole they've dug themselves?
coladict
2024-06-22 16:44:12 +0000 UTCSo, Nvidia is now the highest-valued company in the US. How long do you think this will last? I can't think of a single example of a company in US history that has shot up in value so quickly like this and not crashed back down within a couple years. I don't think Nvidia will ever go below their 2022 values ever again, but there are people that think Nvidia is just doing to keep growing and growing indefinitely, when my gut is telling me that it's just not likely.
IndyTheGreat
2024-06-22 14:46:52 +0000 UTCWhat do you all think of Nvidia's recent trend to load up on performance increases at the high end (90 class) while as you go down the lineup, the increases are less and less inspiring. I know crypto mining killed Ampere and RDNA 2 for gamers, but the launch lineups were great. 3070 as strong as 2080ti for less than half the price, thats awesome. Fast forward, 4070 is $100 more expensive that the 3070 and loses to the 3080. Do you see this trend getting any better with blackwell? (I don't).
B. Fish
2024-06-22 12:58:34 +0000 UTCAMD seems to have a great product portfolio (especially once they release strix halo), but why do you think we're seeing so many compromise products be introduced? X elite looks ok for battery life, lousy for gaming, Lunar lake looks decent but too weak to be a halo product, Intel's other products seem to be all in on "more multicore and more watts is more better." Do these companies not know what consumers want? If strix really can reach 3050 performance, that seems to me what the heart of the market wants: a do everything laptop that can be in a slim package and have decent batter life (and uh AI because even though consumers probably don't care, wall street is incentivizing AI all the things).
B. Fish
2024-06-22 12:46:39 +0000 UTCBack in 2021, Jim at AdoredTV leaked Arrow lake mobile specs 6 big, 8 little, 320 EU graphics with on die cache and I remember thinking "This is really cool and exactly what I'd want." This is before we heard about strix Halo. Now it doesn't remotely resemble that, but I really think a "Big brother" to lunar lake could have been successful. Why do you think it went off the rails and no longer resembles the leaked config?
B. Fish
2024-06-22 12:40:10 +0000 UTCHey guys, It seemed like the majority of the CAMM2 desktop motherboards at Computex were Intel systems. Do you think we'll see at least one Arrow Lake mobo featuring CAMM2 released on the market?
Cleansweep
2024-06-22 12:35:55 +0000 UTCHi guys, Strix looks really good despite the branding hiccup. 2 questions: do you think the volume of strix being with Asus is signalling a return of the "AMD Advantage" Laptops (I think strix + a 100 watt RDNA 4 GPU could be really compelling if it has more than 8gb VRAM), What do you think segmentation will be with Strix Halo, will there be an 8 + 40CU version for gamers? I'd probably buy that config with 32gb of RAM but I really don't need 16 cores on a personal machine.
B. Fish
2024-06-22 12:26:23 +0000 UTCHey guys, I'd be most interested in what the future of Zen for mobile holds. The upcoming release of Strix uses RDNA 3.5., but what about Strix Halo? Will it be using RDNA 4? Or will we go directly to RDNA 5 for upcoming ZEN SoCs? What are your predictions/thoughs? Danke!
Dave Scholze
2024-06-22 08:23:25 +0000 UTCWhat is meant by "High Yield"? Is it TSMC "high yield", or Samsung Foundry "high yield"? Those could be quite far apart.
KarbinCry
2024-06-22 08:00:26 +0000 UTCHello guys do you think AMD can keep up with Nvidia with their CDNA products? I know they revealed they will have yearly cadence of releases but won't the products become more or less small iterative improvements?
QuickJumper
2024-06-22 07:35:12 +0000 UTCThis question/opinion comes a little late, but I feel like it hasn't been sufficiently addressed by the press/media, plus I feel like this may be a good guest to talk about this question: Whenever I hear "Ray Tracing Cores" and "Optical Flow Accelerators", what I really hear is "Blast Processing"... Sure it may be a thing but it's likely not as big of a thing as corpo marketing wants you to believe it is. As someone who was fascinated by John Carmack's talks about ray tracing in 2008 and someone who wrote their own toy ray tracing engines over the years, I feel like it is much more about software than hardware. While AMD's upscaling is no match for DLSS, its frame generation seems to quite competitive without any "Optical Flow Accelerators". Nividia has been catering to the professional market for years. IMO it wouldn't sense to create bespoke RT/Flow Acceleration Hardware. Is seems likelier that Nvidia cleverly repurposes Tensor insructions. What do you think? How much is it about Hardware vs Software? Should Nvidia be under more scrutiny for its marketing claims?
Molly the Bully
2024-06-22 07:05:37 +0000 UTCHi Tom and Max… RDNA4 seems like AMD “quiet quitting” gaming GPUs. Despite your hate for Intel, are they our only hope from escaping Nvidia monopoly in the gaming DIY market? Sarcasm aside… I’m worried we are heading to a midrange and above monopoly from a single company.
JJ Golden
2024-06-22 05:05:32 +0000 UTCWhat are Snapdragon and ARM chips really supposed to do for you that x86 can't? More competition is always better but I'm surprised there was any expectations at all for these snapdragon laptops to sell. I don't think most consumers know or care.
Eric Sizzlicks
2024-06-22 03:39:48 +0000 UTCIf the snapdragon line of chips don’t deliver all their promises. Do you think Microsoft will shift to AMD’s arm iteration of cpus to fund heavily?
XTX 999
2024-06-22 02:48:23 +0000 UTCHi Tom and Guest, are we hyped for a kingpin and PNY collab!
XTX 999
2024-06-22 02:44:38 +0000 UTCWith Meteor Lake Intel has introduced a third type of core with their low-power-island thing, alongside their existing P & E cores. Meanwhile AMD still just has their normal and size optimized, but substantially the same, cores. And those haven't yet both been on the same CPU afaik. Do you see AMD going into these hybrid designs as well or are they going to stick with the more conventional layout?
Lucas
2024-06-22 02:18:57 +0000 UTCEverything I have heard about Snapdragon X Elite reminds me of Windows RT. Is this a software (Microsoft) issue or a hardware (Qualcomm) issue? Or both? Each time Apple changes processors they manage to pull off Rosetta.
Ivalenz
2024-06-22 01:31:45 +0000 UTCGiven the larger performance increases of GPUs over CPUs each generation, is the gaming industry at risk of becoming entirely CPU bound, therefore making AMD the de facto leader with less driver overhead? What would be the direction of Nvidia and AMD if or when we hit that point?
FloridaMan
2024-06-22 01:10:21 +0000 UTCWelcome back! What do you think of the messaging that is coming out AMD? With all the keynotes that happened recently, I got the sense that AMD was the only one with strong confidence in their product launches. In some ways it almost was like a deceleration of war on their competitors. Phrases like "you will see Zen 5 everywhere' and mention of how many product stacks in various industries they have launched, are leading me to believe they are entering a new phase in their sales strategy.
MelodicWarrior
2024-06-22 01:03:14 +0000 UTCIs arrow lake likely to use, in part, Intel's 20A process. TSMC N3 is seems likely used as some reports suggest, but so far as I can tell 20A is set only to be use in arrow lake.
Michael Gordon
2024-06-22 00:49:29 +0000 UTCThis is a long one, first two paragraphs are basically background. When discussing Intel's hybrid design, typically (both in online discussions and Intel's presentations) workloads are categorised into "light" and "heavy", where light tasks are ones that can in theory be completed only on E cores because they do not have substantial processing requirements, whereas heavy ones have more processing requirements and generally justify spinning up less power efficient P cores. I would argue that this categorisation is inadequate at properly justifying why certain workloads should be done on a certain core type. One example where it doesn't work is "heavy" workloads like cinebench or video processing and encoding, where the highly parallel nature means that the E cores can efficiently provide a lot of compute bandwidth and in the end be more optimal at processing this workload. Intel specifically references this parallel "bandwidth" that the E cores provide in their briefings. Furthermore, games tend to not saturate cores in terms of throughput, and are therefore arguably not "heavy" workloads in the way cinebench is, yet these workloads need to be done on P cores because of the higher available single thread performance those provide. I would propose that a more useful categorisation can be made based on the available Instruction Level Parallelism (ILP) in the workload. Low ILP workloads like game threads tend to be highly serialised, whereas high ILP workloads tend to be parallel such as cinebench rendering or video encoding. These low ILP workloads should be the target for P cores because a larger core is needed to extract the ILP and maximise the single thread performance of these workloads, whereas more simple E cores can adequately extract the ILP out of highly parallel workloads and as such provide that "compute bandwidth" while being lean and efficient since they don't need to extract difficult to find ILP.
qhfreddy
2024-06-22 00:17:36 +0000 UTC
