I understand Cinebench makes use of AVX-512, if it's available. That means these figures are only likely to be accurate for software making similar use of AVX-512.
There isn't a use for Cinebench. Right now, it's only used as a marketing tool for AMD. The benchmark itself is not relevant for 99.99% of the people who buy these CPUs.
i doubt that you know what 99% of 5900x or 5950X user see as relevant. let´s say instead.... for the majority cinebench is not that relevant (or software compiling benchmarks or AI benchmarks). there are people out there who do more than gaming and surfing with their computers. for my line of work benchmarks like cinebench or the vray render benchmark are very valuable tools to measure CPU performance. maybe cinebench or software compiling benchmarks are not what "most" people see as relevant.... but they are relevant.
And? lol. If you're not doing rendering on it, it's not the chips' fault. lol. Go do some rendering! I have a 5900x. Gaming is good, but I really bought it to multitask AND game. Tired of all the other programs sucking up my gaming core use. Bring up the game, run discord, leave all anti-virus/malware web checkers, bring up you tube, bring up game support programs, bring up another for web searches based on comments, bring up more for general news while waiting for players to get it together to play or return from a break, ect. The best of all sides.
The original Zen was about as fast as or faster than Haswell and only marginally behind Skylake in single threading, quite an achievement for the first iteration of a new design, compared to Skylake, a fourth revision if we count from Conroe. If we go further back to the Pentium M, III, or Pro, I'll leave it at that.
" No one uses Cinebench " tell that to intel who has been touting it offand on for a few years now. " The only reason Cinebench matters is because Intel marketed the hell out of it when AMDs chips sucked at everything else before Zen 1 was released "
come on lemurbutton, most know you love apple with no end, but it seems you have yet to show how much better apple is over everything else, other then your highly biased and blind apple shilling.
That's nonsense. In fact AMD used Blender to showcase their first Ryzens. And I cannot remember a single scenario where my Ryzen 1600 sucked. Especially at content creation/conversion it was muuuuch better than my similar priced Skylake i5. I had like 50-100% performance in x264/x265 at very similar power consumption. Of course more cores and SMT helped a lot. But broken down to a single core the Ryzen was absolutely competitive. I didn't notice any disadvantage in daily used apps. Gaming performance also was comparable. You just sound like someone who never owned a Ryzen system. ^^ And btw, Intel is using Cinebench as "marketing tool" as well.
Dr. Ian Cutress: "CineBench R23 does not use AVX-512. Maxon has told me that the benefit of using AVX-512 is negated by the clock-down required to keep everything stable, so it isn't used. I'm told this is also why it is disabled in Intel's Embree, (which Cinema4D uses)."
For the doubters, see I told you. The companies don't behave the way you want them to, they behave in the manner that best interests them and of what is available.
I don't think it was you, but in some previous threads people didn't agree with my comment. I said Zen4 was a refresh of Zen3 on the new platform, and that AMD was pulling the brakes and dragging out their upgrade pathway because competition wasn't strong. Basically got called an Intel shill.
They didn't want to believe that Zen5 would be a new architecture, and a significant update, which they were not releasing yet. And that AMD is the cats pyjamas or bees knees.
Sucking out 8-10% IPC over Zen 3, which already has a high IPC, is nothing to sneeze at. Mixed in with the clock speed improvements we will be getting a fairly impressive gen over gen performance improvement. This will be more than what Intel was doing for gen over gen from Sandy Bridge until Rocket Lake. That said AMD probably knew from right when Zen 3 was designed that the DDR4 platform was going to be hindering its overall performance. However, in late 2020 DDR5 was not ready so they had no choice in the matter. Now that DDR5 is available they can make some improvements to Zen 3, make it Zen 4, and get some good added performance.
I think Zen4 improvements over Zen3 are fine, but don't measure performance improvements by gen - measure them by years. 8-10% IPC over 2 years is like 4% per year
That makes absolutely no sense to use year over year for performance gains. You have to use gen over gen since new CPUs aren't always released every 12 months. Especially in the current pandemic/supply chain issue in the world right now.
Zen 4 Raphael 8% to 10% IPC gain, 15% so said, means the architecture's CPU systems bus, fabric, is becoming saturated which means Zen 5 architecture improvement would address this traditional sign of any processors design age, typical of time. And AMD proof of saturation, Zen 5 "re-pipelined front end and wide issue". New inter chip communications Zen 4 + RDNA lll may be early but if CPU bus saturation is the issue large off CPU caches are one solution. By the times its RDNA lll and CDNA lll its now CXL? mb
What are you basing that on? Just the >15% single thread improvement and extrapolating from there? Notice in the greater than symbol in the slide? 15% will be the minimum single threaded improvement. Improved IPC, DDR5 and 5.5Ghz demonstrated across active cores (although not all cores) gives more than that alone... They're being conservative.
Bushlin, I don't doubt AMD is being conservation on my Zen 4 cost : price / margin estimates. TSMC + 20% price increase into second half 2022. Sumco + 30% wafer price increase. Zen 7 to 5 nm shrink + 40% dice per wafer. The performance increase has to be better than 15% to make up for the cost increase despite the good die increase I'm still working the estimates. mb
BushLin " What are you basing that on? " brazone is basing it on numbers he pulls out of thin air, so no one else can confirm what he comes up with. ignore him. he will NEVER post links to where he gets his info from, so other people can also come to the same results and conclusions. he is a fraud.
Qasar, You're welcome to read TSMC node to node power v frequency and density claims, industry news on fabrication CapEx and materials cost increases, perform a good dice estimate and post your results. mb
figures, STILL no direct links to your sources. why ? oh wait, cause you DONT HAVE ANY SOURCES ! searching for it, i could and up finding different numbers and such then you did. must be fun for you to post fud knowing no one can confirm or deny what you come up with.
face it your are just a fraud, and a couch potato " analyst ". any thing you post is made up, with numbers you pull out of thin air.
Zen 4 isn’t a refresh of Zen 3. Higher IPC different ram controller, way higher clocks, different instructions. Either you didn’t read the article or don’t understand much about CPUs in general.
> I said Zen4 was a refresh of Zen3 on the new platform
It's clearly more than a simple port of Zen3 to 5 nm. We already saw how little benefit the Ryzen 6000's gain from the 6 nm port, and if you go back to Zen+ (i.e. 2000-series non-APUs), you'll see that 12 nm also offered only a couple %.
So, whatever they did, it might not be a roots-and-branches overhaul, but it's also a lot more than a straight port + some AI instructions.
I don't get why you're all so upset, they spelt this out as their plan in 2019:
He also indicated that AMD's server CPU launches are set to rely on the "tick-tock" cadence that was once the hallmark of Intel CPU launches, with the launch of a CPU platform that relies on a new manufacturing process node but the same microarchitecture as the last platform (the "tick") followed by a platform that relies on a new microarchitecture but the same manufacturing process node (the "tock").
I don't think they need it, at least not yet. There was an article at Chips & Cheese that elucidated how Zen 2 (yes 2, not 3!) was extremely competitive in total energy use with Gracemont, and that Zen 2 on mobile blew completely past Gracemont on every power level.
It is little understood reality (and intentionally obfuscated by Intel) that E cores (Atoms) are not actually energy efficient - they're actually worse or similar to big P-cores. However, they're a lot smaller - they are die space Efficient.
The whole BIG.little on x86 is not there for energy efficiency like on ARM/mobile but exist so Intel can win on Cinebench and heavily multi-threaded workloads because P-cores can't compete with Zen otherwise. They can't compete because they've grown the P-core to ridiculous size in the hunt for IPC, and there was no way to back out of that (and losing process advantage to TSMC didn't help).
Intel needs the BIG.little to help reign in their power consumption. Their P cores when running all out on AVX workloads suck a lot of power. The 12900K with 8P+8E cores sucks down 272W in 16c/24t configuration and 239W in 8c/16t mode. The E cores are using about 4W/core but the P cores are using ~30W/core. If Intel wanted to do a 10P core chip it would be breaking 300W under full AVX. The Core uArch hasn't been really that power efficient since Broadwell. It wasn't designed to go over 4c/8t at first and over 4.0GHz its power curve looks bad. However, due to pressure from Zen, which was designed around efficiency and 8c/16t, Intel had to push Core to the limits.
>they're actually worse or similar to big P-cores The same Chips & Cheese article I alluded to showed Gracemont using half the energy of Golden Cove in similar workloads, so I think "worse than P-cores", at least, would be an exaggeration.
Gracemont certainly did do worse in AVX efficiency than otherwise, but I think that is to be expected, as energy use in AVX workloads is probably going to be dominated by the execution unit itself, and therefore isn't going to be significantly different just because it has been integrated into a different microarchitecture. I think the more pertinent picture is https://i0.wp.com/chipsandcheese.com/wp-content/up... which, being integer-bound, is more likely to show the energy use of the whole out-of-order engine and whatnot, rather than that of a specific execution unit.
That being said, if you look at the cumulative energy use of the libx264 execution, it looks more impressive than on the instantaneous power draw chart and shows that Gracemont is, at least, *more* efficient (if not by a huge amount): https://i0.wp.com/chipsandcheese.com/wp-content/up...
Thanks for sharing! Very interesting that Gracemont seems designed to clock only up to about 3.0 GHz, while the knee in Golden Cove's energy curve comes at about about 4.2 GHz.
It's also interesting to look at where those curves first start to climb. In Gracemont's case, it begins a gradual climb above 1.2 GHz, whereas Golden Cove doesn't markedly increase energy usage until about 2.0 GHz and 1.4 GHz. Those would be the peak-efficiency points (at least clock-speed wise - those graphs don't show actual performance).
One thing I didn't expect was for the slopes to be so similar, for so long. However, it's hard to read much into that, without knowing how actual performance scales with clocks. Ultimately, what we care about is how much perf/W the cores are delivering.
That's what the 4 and 4c designs will be, but mainstream users would probably not use all the 4c cores, so they are sticking to big cores (little cores on server for nT tasks).
More importantly, as another commenter said, zen 2 and (presumably, but less than zen 2) zen 3 are already pretty efficient.
Intel's big.little seem to have missed the mark in terms of efficiency, at least this generation. It allowed them to get solid 1T and nT performance, but the efficiency and hence laptop battery life is not great. Hopefully next generation will be more impressive in this regards.
I wonder if in the distant future, would AMD come up with something for portables.
Specifically for the 10in-18in portables, running anywhere from 7W-70W power. Hypothetically, they might be able to do a BIG.little design but with a mixture of x86 and ARM cores. For example; 15W for a chipset with 2x "Zen7" BIG Cores and 8x "Cortex-A750" little cores. And more "gaming" oriented Laptops (35W) could have a 4x Zen Cores with 8x ARM cores, for modern gaming engines which use more threads to work more appropriately. This all would run on the new Windows20 Operating System, where everything can run on the ARM cores with decent performance and low energy drain, but Programs needing to run Legacy Mode, would be shifted to the x86 cores. When maximum performance is required, the x86 cores are prioritised for that activity, and the ARM cores are pressed to handle the background tasks. This way we get the best of both worlds. Meanwhile, Full-sized Desktops would have an all-x86 chipset since they don't have the limitations of thermals, power, and portability.
Obviously this is just pie in the sky thinking at this point, but it might be a possibility someday.
Why? Introduce all that extra complexity for what? Currently their big cores are more power efficient that Intel's little cores. If you want extreme multithreaded performance then go with Zen 4c. But standard Zen4 will have excellent single threaded performance, and excellent multi threaded performance.
There are always going to be time insensitive or low performance tasks like file copying, network operations, and UI handling. If you can allocate those tasks to a 10W core instead of a 30W core you get an additional 20W, of both heat and energy, you can spend on a 30W core.
In other words if your system throttles after 3 seconds at max performance, that extra 20W budget may increase the time your 30W cores can run at full throttle for an additional 4 seconds. Or maybe it can run at 4/5 performance without ever throttling.
I imagine they will, and due to their chiplet design it shouldn’t have any extra complexity. They can allocate low resource tasks to efficiency cores and allocate the extra watts saved towards the performance cores.
Well, if they were, then you'd expect they would've announced an efficiency core by now.
As it stands, their answer to Intel's E-cores is the Zen 4c, and we have yet to see how its efficiency really compares to the normal Zen 4. It might not be a big enough gap to justify having a Big.Little configuration.
If we assume the 4C is their efficiency core we can compare their announced products. Their performance core maxes out at 96 cores in a pin compatible part vs 128 efficiency cores. The assumption is that if both parts consume the same amount of energy then their e cores consume 3/4 the energy and emit 3/4 the heat of their p cores.
If they further downclock the e cores slightly, they will then consume half, possibly even a third, of the energy.
So if they have a 1:7 e/p ratio we could end up with 12e/84p; because the e cores would be running OS tasks, networking and file copies and other time insensitive tasks, the energy previously wasted by running them on a p core can instead be allocated to one of the existing p cores. It’s possible those 84 p cores don’t throttle any more compared to a 96 p core part if there is enough surplus energy budget.
> If they further downclock the e cores slightly, they will then consume half, > possibly even a third, of the energy.
That downclocking should apply equally to the regular Zen 4 cores. So, we're back to your original figure of 3/4th the power, and we don't know how much they'd have *already* been downclocked, just to hit that target! So, 3/4ths is probably more of a best-case estimate than a starting point.
Then you’re wrong. Big little is a old design, if AMD wanted it they would’ve announced or made something at this point but they didn’t. They have also publicly said that they are not planning to do so, because they don’t need it, their Regular cores are efficient enough and not bloated like Intels. Modern tech not outdated 1637361th gen core arch.
Based on history, AMD’s APUs using their architectures are always slightly more advanced and delayed than their desktop counterparts. From this I gather that the Zen 4 APUs will likely be on 4nm. Zen 5 is likely on both nodes, either because TSMC N3 is facing delays or because the APUs will be a ‘tick’ on 3nm with the CPUs a ‘tock’ on 4nm.
Their map is so solid now, Intel on the other hand is bragging about Foveros and other things. And on the consumer side more of that Biglittle junk because they cannot fix Core uArch to scale past 8P anymore. Anyways Zen 4, 5 with V-Cache was not surprising but Genoa-X 1GB L3 was because I didn't expect them to reveal it before even Genoa launched.
No more big.little bs rumors, AMD is using 4c for Bergamo as we all know 5c will also have similar successor. Because ARM's Neoverse V1 and N2 are coming with high density cores. AMD want's to cull them before itself, good move. x86 needs to reign this space and consumer space else the insane BS fragmentation and software issues with compat will show up. And AMD doesn't need to break the CPUs as they provide full high performance x86 cores without the added "efficiency" baggage on a Desktop. Intel is meanwhile stuck until they find a replacement for the Core series uArch.
At this point not sure if their AM5 socket with 1718 pins will stick until 2024. I hope they do, that would make Intel pathetic given how they are going to axe LGA1700 / 1800 pins total for a new LGA1851 socket ripping off people again like LGA1511 disasters. Shame when ILM is busted on Alder Lake.
For consumer Zen 4 Raphael / Ryzen 7000, I'm waiting to see how the clocks will stay now. Also most importantly their IMC and Firmware. The IOD is now TSMC 6nm, hopefully all the dreaded garbage WHEA and USB nonsense is gone. Also AGESA 1.2.0.7 has it's share of issues apparently despite aiming to fix TPM and stutters. This is the main weakness of AMD, garbage firmware and poor tuning experience. CO is very much head ache given how the clock behavior is on AMD and the heat on Zen 3 scales heavily dependent on that unlike Intel which can stick to a given clock and poor IMC with annoying XMP issues vs Intel again.
Who cares about AMD vs Intel anymore? Wake up. M series is destroying both. I doubt AMD can catch up to the M1 until Zen6. By then, M5 will be out since Apple will do a yearly update.
The 15w M1 was trading blows with the already launched 15w Renoir (Zen 2) in 2020. https://www.anandtech.com/show/16252/mac-mini-appl... Fantastic claims require fantastic evidence and I've not seen anything that comes close to backing up any of what you're claiming.
lol, who cares about Apple? They have 9% market share on desktop and 15% in laptops. Everything else is x86. Would an M1 equipped Macbook be amazing if it could run all my stuff natively (including thousands of indie games) and I had 2000€ to spare? Sure. But for most people those are BIG ifs.
Who cares about Apple....? It is not Apple that is the issue for AMD and Intel, it is Qualcomm and what they can do with Nuvia. If that is competitive with M2 in terms of performance and efficiency when it launches, there is going to be a shift to Windows on ARM. My high powered x86 laptop is a total joke on battery, it can barely last a 2 hour meeting, let alone a conference. Ill be switching as soon as possible
Qualcomm is run by some greedy corporate types, and most of their talented designers have left the company a few years ago. The Nuvia cores did look promising and they were supposed to be released ages ago, until Qualcomm basically bought them out.
I just don't expect much to come from Qualcomm-Windows. You are better off sticking with Android instead for the Efficiency, UI, and Apps. Or if you need proper Windows, stick with the x86_64 platform instead of an emulated-environment.
Intel and AMD do, for one. Apple shows what a good architecture can do in terms of performance, performance-per-watt, dedicated logic and what trends are. Intel referenced Apple‘s M1 specifically in internal slides as what they are shooting for (some years later).
I expect that Apple‘s decision to dramatically increase memory bandwidth will have an impact on AMD‘s and Intel‘s decisions as well. Even the M2 (without modifier) has 100 GB/s memory bandwidth, which is close to what AnandTech measured in AMD’s Milan. (I realize I compare theoretical maximums with actually measured throughputs, but my point is more about where to put the emphasis in design.) And it stands to reason that the M2 Pro/Max/Ultra will at least match the M1 variants.
If the 100GB/s bandwidth figure is anything like what was marketed on the M1 Max then only a fraction of that is available to the CPU cores. M1 Max headline number from Apple is 400GB/s but the reality is 100GB/s single threaded and <250GB/s best case scenario with all cores loaded. https://www.anandtech.com/show/17024/apple-m1-max-... Not saying the performance numbers won't be impressive for a little SoC but far from the Jesus silicon some fantasize might about. Not aimed at the poster above but it helps to post links with actual results rather than slinging around far fetched claims.
Fair point, and I should have mentioned that. However, my point was two-fold: my comparison to measurements in Milan was to make the point that Apple decided to spend silicon to give the M2 — Apple’s smallest non-iPhone SoC — roughly the same type of memory bandwidth as a server-class x86 CPU — or way more in case of the Mx modifier variants. So rather than spending more transistors and power optimizing the CPU, Apple dedicates them to memory and caches.
And secondly, the large increase in bandwidth means that memory bandwidth is not a bottleneck for total performance (true at least for the M1 Max and M1 Ultra, because even when CPU, GPU and other specialized logic is running full tilt, the SoC still has enough memory bandwidth to saturate them all). The same cannot be said for many of AMD‘s and Intel‘s CPUs (at least the ones with higher core counts and/or integrated GPUs).
My prediction is that Intel and AMD will follow suit in the future and give more emphasis to memory bandwidth, i. e. Apple‘s SoCs definitely do influence what Intel and AMD are doing — which is what I was replying to. You can also see that Intel and AMD are trying to integrate more specialized circuitry for specialized tasks as well, also something that Apple has emphasized (in terms of transistors and power) on its M-series SoCs.
> roughly the same type of memory bandwidth as a server-class x86 CPU
Not at all. It's only workstation-class. EPYC Milan and last year's Ice Lake SP both have 8-channel DDR4, providing nominal bandwidth of 204.8 GB/s. Cut that in half and you get something like Threadripper (non-Pro) bandwidth numbers.
DDR5, with Sapphire Rapids shipping probably around the same time as M2-based systems, will move the bar yet again. With it, even desktop platforms are creeping into M2 territory. Alder Lake has a nominal bandwidth of 76.8 GB/s.
So, Apple is really only barely keeping pace, considering graphics needs a share of that bandwidth. For perspective, the PS4 that launched almost 9 years ago had 176 GB/s unified memory bandwidth.
“Not at all. It's only workstation-class.” Anandtech tested memory bandwidth of two Milan EPYC 7000-series processors, i. e. AMD's server processors, and they were tested at 103~113 GB/s peak throughput (https://www.anandtech.com/show/16529/amd-epyc-mila... Your figures are the theoretical max bandwidth. To be fair, I could not find figures for tested bandwidth for the plain M1. On the other hand, to be fair, we are comparing Apple's second-gen lowest end chips meant for its entry-level notebooks and desktops to AMD's current-gen highest-performing chips. Extrapolating from the the numbers of the M1 Max, the M1 Ultra should have about twice the sustained throughput due to the CPU complex only than AMD's current top-of-the-line server CPUs.
“So, Apple is really only barely keeping pace, considering graphics needs a share of that bandwidth. For perspective, the PS4 that launched almost 9 years ago had 176 GB/s unified memory bandwidth.” Barely keeping pace? Apple's highest-end SoC has a theoretical max memory bandwidth of 800 GB/s, which is comparable with a current top-of-the-line graphics card like nVidia's RTX 3080 (about 912 GB/s).
And why don't we look at the PS5 instead of the PS4? The PS5 has a theoretical max memory throughput of 448 GB/s, roughly on par with the M1 Max. So is Apple really behind? Doesn't look like it to me. Especially once you include efficiency in your considerations.
> Anandtech tested memory bandwidth of two Milan EPYC 7000-series processors, > i. e. AMD's server processors, and they were tested at 103~113 GB/s peak throughput
So, you're comparing Apple's theoretical bandwidth to AMD's actual bandwidth? Nice. Try again.
Also, to interpret benchmarks, you need to know what you're looking at. The Stream Triad benchmark is meant to stress the cache subsystem. It assigns each thread to do a series of 2 reads + 1 write. In copy-back caches, the write actually generates an extra read. There are ways around this, but the benchmark is written in a naive fashion.
So, to achieve 113.1 GB/s on that benchmark, the underlying memory system is actually doing 150.8 GB/s. As mentioned in the article, Altra has a hardware optimization which detects the memory access pattern and eliminates the extra read. That's the main reason it's able to achieve 174.4 GB/s, using the same 8x DDR4-3200 memory configuration as the Epyc. Also, the Epyc is being run in the generally slower NPS1 topology, meaning traffic is crossing all of the NUMA sub-domains.
Now, as for the rest of the margin, you seem to have fallen in the trap of citing the launch review of Milan, which used a pre-production platform. Unfortunately, they didn't rerun the Stream.Triad benchmark for the updated benchmarks, but if you compare the SPECint results, multithreaded performance between the two articles jumped by 7.9%. It's hard to say what difference there'd have been in memory performance, but we can use that to get a rough idea.
Yes, because Intel mainstream desktop CPUs have a nominal bandwidth almost as high. Everyone gets a boost, moving to DDR5. They are not special. 100 vs. 76.8 is significant, but not the massive difference you tried to paint.
> Apple's highest-end SoC ...
Uh oh. Now you go and change the subject. Irrelevant.
This article is about the M2 and I was simply putting it into perspective. I take it you're offended by this or what it reveals, because otherwise I don't see why you'd drag completely different Apple chips into the discussion. Seems like a face-saving move, but ultimately just makes you look desperate.
And BTW, I'm even going to retract what I said about "workstation bandwidth", because a quad-channel DDR5 setup is going to spec out north of 150 GB/s. So, if we're comparing like-for-like, rather than against old DDR4-based products about to be replaced, then it doesn't even reach the workstation tier.
Apple is winning the premium laptop market and is growing significantly faster than x86 laptop makers. In a few years, Mac marketshare will look very different.
So, Apple and their fearsome M have destroyed both AMD and Intel. The tide has turned. Dear friends, who knew this day would come so swiftly? The only thing we can do now is take all our x86 computers and, with a sigh, throw them in the bin as fast as we can because they're useless now.
Inside of the MAC universe the M series is king. However, what happens when it has to run on an OS that isn't highly tuned like OSX or software that doesn't have the fine tuning? Probably the same thing that happens to every ARM uArch, it bogs down to 2015 x86 performance level.
That article first analyzes the penalty of running under Linux vs. MacOS, then compares Linux performance to a stack of x86 machines. Currently, it's delivering 10th or 11th gen i5 performance, under Linux. Not bad, for a community effort.
Most of the time the M1 and Ryzen 3 3300X are close in performance. That puts the M1, with 4+4, around the performance of the top low end CPU, 4c/8t, from AMD in April 2020. There are times when the M1 performance is higher but there are an equal amount when it is lower.
It won't. Apple has had the premium smartphone market for more than a decade and its share has remained quite similar since then. And Mac is not as popular as iPhones.
Apple's market share has been on the cusp of a dramatic explosion for decades now. To be blunt, Apple doesn't WANT to be mainstream. If everyone has a Mac, it isn't special and doesn't inspire cult-like devotion. Also, they'd have to give up profit margins to secure a toehold in more cost-sensitive market segments.
That's the whole point. Apple, quite consciously, cultivates this aura that they're better than everyone else, and that's partly why many folk buy their products. Now, if Apple were a humble firm, nobody would mind and they'd likely be applauded by the Win/x86 camp; but they're not, believe they're the best, and think there's nobody like them, with all their bought-off companies.
LOL. The iPhone proves you wrong. It's *very* mainstream, which proves they (and their investors) would happily settle for a position of market dominance over premium prestige.
Perf per watt is actually fairly similar for nT workloads between M1, AMD 6000 series and alder lake H. The difference is in 1T and in graphics and in software optimization. The push for higher clock speeds is not going to help AMDs efficiency with Zen 4, unfortunately.....there's a reason apple clocks it's CPUs at ~3-3.5 Ghz.
M1 has never beaten even 5900X in ST, much less 5950X/
I know this for sure, because I have a 5900X. In all published benchmarks where I have access to the benchmark programs, so I can run them, a 5900X is slightly faster in ST than any M1. e.g. in GB5 ST the score is around 1800 for 5900X (non-overclocked, non-PBO) vs. around 1750 for M1.
Obviously, it is true that a desktop AMD CPU must consume maybe 4 to 5 times more power than an M1. when running in ST, to be able to beat it, but this does not justify exaggerations like claiming that M1 is better in absolute performance.
Please could you provide some credible evidence for any of that, especially 0.5w package power while supposedly winning any kind of benchmark... The rest is only slightly less realistic and still hard to take seriously. Presumably you have some solid benchmark results with accurate power measurements including something like a Ryzen 6800U.
You are grasping straws. It is empirically more power-efficient than any Intel/AMD laptop. We can play that game all day about this benchmark and that benchmark.
But no sane person would think that their laptop, running under similar conditions, can last longer than an M1/M2 MBA.
Interesting take on someone challenging nonsense. When the M1 was released, it was impressive for the single thread performance, wide design and general efficiency; albeit with the double edged sword of closely integrated RAM and frankly anti-consumer non-upgradeable and proprietary SSD. Those are the facts; it was a momentous release for Apple and I've recommended the Macbook Air to people who don't have issues with the software library... However, this thread has someone claiming 5x better performance per watt!!! Be real. Objectivity is in short supply for some it seems. Also, the article is about Zen 4 ffs.. For what it's worth, AMD is sandbagging. >15% single thread improvement will be a worst case scenario and Zen 4 will absolutely stomp on Intel for multithread (I don't know about Apple's future plans). Raptor Lake is delayed so won't get a chance to take the crown before Zen 4 is released and it only gets worse from there. I'm not taking credit for those predictions; "Moore's Law is Dead" is rarely wrong, has contacts throughout the industry and is super confident about those events playing out. I actually think this could be bad for consumers because we now need Intel to step up to keep AMD's pricing competitive.
True. When M1 launched, it was good. All the M series processors that followed after that weren't that great. In fact, the new M2 is dismal. It's only going to get worse for apple when you see AMD & Intel roadmap.
M1 may be about 5x better perf/watt, but as the other poster has written, this is true only for single-thread benchmarks, where the Intel or AMD CPUs approach 5 GHz, having low energy efficiency.
As the other poster has said, in multi-threaded benchmarks, and if the power limits have not been raised above the nominal TDP, the clock frequency drops and also the power consumption is reduced to a few watts per core.
In this case, the Apple CPUs remain more efficient, but their advantage on perf/watt becomes significantly less than double, so very far from the "5x" claimed above.
Moreover, 5x is not true even for single-thread. In the Apple presentation from the beginning of the week, Apple has acknowledged that their new improved M2 has a perf/watt advantage over a 28-W Alder Lake P of only 3.6 times, i.e. much less than 5 times.
If the advantage is 3.6 times vs. Alder Lake P, than it is less than 3.6 times vs. Ryzen 6000 and it will also be less than 3.6 times vs. the future Intel Raptor Lake, which will be the competitor of Apple M2 for most of its lifetime.
For multithreaded applications the perf/watt advantage of M2 cannot exceed 1.5 to 2 times.
There is no doubt that Apple M2 will have better performance per watt than any Intel or AMD CPU, at least until the second half of 2023, when Intel will launch Meteor Lake and AMD will launch Zen 5.
This true fact does not justify claims of fantastic numbers for M1 or M2.
Also, M1 can use 0.5 W total package power on GB5 ST only if the benchmark runs on one of the small cores. Any large Apple core consumes alone at least 4 to 5 W (when in iPhones, more in laptops) when running a ST benchmark. Look at all the reviews ever published on Anandtech to see that.
Is 5x better performance per watt even a credible claim in the first place? Even if not pulled out of someone's rear, is that comparing a mobile CPU a high end desktop CPU which scales to way higher performance, in a very specific hand-picked scenario?
1T perf/watt is about 3 times higher, with core power consumption being about 2 times higher, while nT perf/watt is similar. Both in CB23, which probably disfavors apple by about 20% compared to Geekbench for performance, but that's not a big change.
Who cares about Apple anyway? Wake up. M series doesn't support dedicated graphic cards nor mainstream OS like Windows or Linux distros. I doubt Apple would open up M-series at all. by then, they're on their own. especially when Apple didn't do yearly update on M-series anyway — did 19 months after original release considered as yearly release?
Apple hasn't supported Windows or Linux distros ... for years now.
They seem to be doing pretty well for themselves.
Why do people continue to use the same argument over and over. They are successful DESPITE what you said. So clearly, those aren't their weaknsesses. ;-)
I refrained from the beta BIOSes for two years on my B450 Tomahawk + Raven Ridge, but about a fortnight ago upgraded to 1207, which, as far as I can see, has been working smoothly. I believe the 1206x branch had issues.
> the message from AMD is clear: they’re going to be doing some significant reworking > of their core CPU architecture in order to further improve their performance > as well as energy efficiency.
No, it means they *already* did some significant reworking. From the Oct 26th interview with Mike Clark, he indicated the architecture of Zen 5 was already done by that time:
"I have this annual architecture meeting where we go over everything that's going on, and at one of them (I won't say when) the team and I went through Zen 5. ... Coming out of that meeting, I just wanted to close my eyes, go to sleep, and then wake up and buy this thing."
That he wouldn't disclose *when* suggests the meeting was some months prior.
> on top of Zen 4’s new AI instructions, Zen 5 is integrating further AI and > machine learning optimizations.
I wonder to what extent this will continue to be relevant. I think they've felt pressure from their competitors and customers, but it's an open question how relevant CPUs will be for machine learning, by the time Zen 5 finally reaches the market. Everybody has AI accelerators, from chip vendors all the way to their biggest customers.
I think ARM did the right thing, here. They restricted their wide vector compute to their V-series cores, while their N-series - their bread-and-butter cloud-oriented cores - will have narrower implementations.
I expect AMD to use narrower vector compute in their c-series and APUs, while their chiplets have wider implementations.
On WikiChip it is claimed that Zen 4 will support exactly the same AVX-512 instruction subsets like Ice Lake, together with extra instructions introduced in Cooper Lake.
The (few) newer extensions of Tiger Lake and Sapphire Rapids are not supported.
In an interview, AMD has confirmed that the instructions for AI that have been mentioned are those included in the two AVX-512 subsets introduced by Cascade Lake and by Cooper Lake.
Interesting, and it will be too funny when AMD has this stuff end of year in the desktop platform and intel hasn’t. You’re basically forced to buy AMD then for AVX512 or low core 11th gen or far outdated Hedt platform of intel. The choice is clear
> it will be too funny when AMD has this stuff end of year in the desktop platform
*if* they enable it on the desktop. I expect they will, but they could play games with market segmentation.
I'm hoping they don't nerf it, like Intel did by disabling one of the two FMAs per core, in the lower-end 14 nm CPUs that had it. That meant there were some cases where AVX-512 was a net performance penalty, because even using the single FMA would trigger clock throttling that slowed down the CPU by more than the benefit AVX-512 provided.
AMD has more or less already said it will be in upcoming Ryzen 7000, it’s a 99% thing. And for anything else, we will have to wait and see. So far, Ryzen was executed solidly so I don’t expect otherwise.
If AMD restricts it anywhere, I think it might be leaving it out of their APUs. But that would probably be about reducing die area, in which case the actual silicon wouldn't have it.
> Maybe they'll implement AVX-512 based on the pipes of two AVX-256 units.
One thing that makes AVX-512 annoying is the bigger vector registers. Also, there are twice as many of them. So, even if you duplex AVX2 instructions down half as many AVX-512 pipes, it still doesn't negate the die area impact.
Plus, AVX-512 adds different types of instructions than AVX/AVX2, which means additional dedicated logic with no other purpose.
It's not that simple. There are operations with AVX-512 that don't map cleanly on to AVX2. In some cases, it's due to structural changes in the instruction set. In others, it's new element data types. And there are things like scatter/gather that AVX2 simply doesn't have.
BTW, AVX-512 even added support for the new instructions at narrower width. So, you can use them on 256-bit vectors without the full thermal impact of 512-bit processing.
However, I don’t agree with the statement that if won’t be relevant in consumer space. Consumers can work with their PCs too, and the way normal desktop unfolded in the last years, more or less because of AMD, particularly the 16 core CPUs are for workstation not really for gaming, the least thing they are for is content creation. Intel is just a joke with their terrible efficiency, with Intel everything is just a complete joke, they make their top CPU a power hog which is unprofessional, it gives off a “gamey” vibe.
It's pretty hard to make the case for AVX-512, outside of deep learning and HPC apps, I think.
There are some AVX-512 optimized string processing libraries and it's useful in video compression. However, GPUs are better for video compression, when they support it (and you don't need the absolute best quality).
There aren't very many reviews of Milan-X anywhere outside of what is provided by places like MS Azure. However, Milan-X was so much faster than regular Milan in HPC workloads that MS is replacing there almost brand new Milan systems with Milan-X (HBv3 instance types). https://www.servethehome.com/amd-milan-x-scaling-t...
Well Apple was ahead of the curve, they are using lpddr5, nothing strange there and they are feeding the gpu with that bandwidth as well, they really need it.
I think a confluence of factors and bad decisions resulted in bulldozer. AMD seems to have changed their ways, but they're still somewhat at the mercy of circumstances.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
156 Comments
Back to Article
jamesindevon - Thursday, June 9, 2022 - link
I understand Cinebench makes use of AVX-512, if it's available. That means these figures are only likely to be accurate for software making similar use of AVX-512.LightningNZ - Thursday, June 9, 2022 - link
Well, what use is Cinebench now, except to compare against other Cinebench runs, and as a general guide for rendering performance.lemurbutton - Friday, June 10, 2022 - link
There isn't a use for Cinebench. Right now, it's only used as a marketing tool for AMD. The benchmark itself is not relevant for 99.99% of the people who buy these CPUs.Gothmoth - Friday, June 10, 2022 - link
i doubt that you know what 99% of 5900x or 5950X user see as relevant.let´s say instead.... for the majority cinebench is not that relevant (or software compiling benchmarks or AI benchmarks).
there are people out there who do more than gaming and surfing with their computers.
for my line of work benchmarks like cinebench or the vray render benchmark are very valuable tools to measure CPU performance. maybe cinebench or software compiling benchmarks are not what "most" people see as relevant.... but they are relevant.
Crazyeyeskillah - Friday, June 10, 2022 - link
i bought a 5950x as a gamer thinking i would do more rendering on it. . .All i've used it for is gaming and emails. Youtube also gets a big rotation.
GeoffreyA - Saturday, June 11, 2022 - link
True. Since Sandy Bridge or so, every CPU is fast for most purposes. Gone are those days of slow and fast computers.Rufnek99 - Friday, June 17, 2022 - link
And? lol. If you're not doing rendering on it, it's not the chips' fault. lol. Go do some rendering!I have a 5900x. Gaming is good, but I really bought it to multitask AND game. Tired of all the other programs sucking up my gaming core use. Bring up the game, run discord, leave all anti-virus/malware web checkers, bring up you tube, bring up game support programs, bring up another for web searches based on comments, bring up more for general news while waiting for players to get it together to play or return from a break, ect. The best of all sides.
lemurbutton - Saturday, June 11, 2022 - link
No one uses Cinebench. Even for rendering it's a niche product.The only reason Cinebench matters is because AMD marketed the hell out of it when its original Zen chips sucked at everything else.
GeoffreyA - Saturday, June 11, 2022 - link
The original Zen was about as fast as or faster than Haswell and only marginally behind Skylake in single threading, quite an achievement for the first iteration of a new design, compared to Skylake, a fourth revision if we count from Conroe. If we go further back to the Pentium M, III, or Pro, I'll leave it at that.Qasar - Saturday, June 11, 2022 - link
" No one uses Cinebench " tell that to intel who has been touting it offand on for a few years now." The only reason Cinebench matters is because Intel marketed the hell out of it when AMDs chips sucked at everything else before Zen 1 was released "
come on lemurbutton, most know you love apple with no end, but it seems you have yet to show how much better apple is over everything else, other then your highly biased and blind apple shilling.
gruffi - Sunday, June 12, 2022 - link
That's nonsense. In fact AMD used Blender to showcase their first Ryzens. And I cannot remember a single scenario where my Ryzen 1600 sucked. Especially at content creation/conversion it was muuuuch better than my similar priced Skylake i5. I had like 50-100% performance in x264/x265 at very similar power consumption. Of course more cores and SMT helped a lot. But broken down to a single core the Ryzen was absolutely competitive. I didn't notice any disadvantage in daily used apps. Gaming performance also was comparable. You just sound like someone who never owned a Ryzen system. ^^ And btw, Intel is using Cinebench as "marketing tool" as well.evilpaul666 - Monday, June 20, 2022 - link
People were benchmarking with Cinebench before AMD latched onto it when they had a core advantage (and non-trash chips) starting with Ryzen in 2017.inf_64 - Thursday, June 9, 2022 - link
Cinebench doesn't use AVX 512, if it did, some intel chips that support it would be blowing everything out of the waters by more than 2x delta.mode_13h - Friday, June 10, 2022 - link
Dr. Ian Cutress: "CineBench R23 does not use AVX-512. Maxon has told me that the benefit of using AVX-512 is negated by the clock-down required to keep everything stable, so it isn't used. I'm told this is also why it is disabled in Intel's Embree, (which Cinema4D uses)."https://twitter.com/iancutress/status/132735837337...
Nov. 23, 2020
mode_13h - Friday, June 10, 2022 - link
With the clock-speed impact lessened in more recent CPUs, I'd expect Maxon & others to revisit some of these decisions.Khanan - Friday, June 10, 2022 - link
I also think Ryzen will manage the clock speed loses better than Intel CPUs, but it’s just a guess at this point.Iketh - Monday, June 20, 2022 - link
uhh ya that's the case for every benchmarkKangal - Thursday, June 9, 2022 - link
For the doubters, see I told you.The companies don't behave the way you want them to, they behave in the manner that best interests them and of what is available.
mode_13h - Friday, June 10, 2022 - link
What are you talking about? What did you supposedly tell us?Kangal - Friday, June 10, 2022 - link
I don't think it was you, but in some previous threads people didn't agree with my comment. I said Zen4 was a refresh of Zen3 on the new platform, and that AMD was pulling the brakes and dragging out their upgrade pathway because competition wasn't strong. Basically got called an Intel shill.They didn't want to believe that Zen5 would be a new architecture, and a significant update, which they were not releasing yet. And that AMD is the cats pyjamas or bees knees.
schujj07 - Friday, June 10, 2022 - link
Sucking out 8-10% IPC over Zen 3, which already has a high IPC, is nothing to sneeze at. Mixed in with the clock speed improvements we will be getting a fairly impressive gen over gen performance improvement. This will be more than what Intel was doing for gen over gen from Sandy Bridge until Rocket Lake. That said AMD probably knew from right when Zen 3 was designed that the DDR4 platform was going to be hindering its overall performance. However, in late 2020 DDR5 was not ready so they had no choice in the matter. Now that DDR5 is available they can make some improvements to Zen 3, make it Zen 4, and get some good added performance.kwohlt - Friday, June 10, 2022 - link
I think Zen4 improvements over Zen3 are fine, but don't measure performance improvements by gen - measure them by years. 8-10% IPC over 2 years is like 4% per yearschujj07 - Friday, June 10, 2022 - link
That makes absolutely no sense to use year over year for performance gains. You have to use gen over gen since new CPUs aren't always released every 12 months. Especially in the current pandemic/supply chain issue in the world right now.Bruzzone - Sunday, June 12, 2022 - link
Zen 4 Raphael 8% to 10% IPC gain, 15% so said, means the architecture's CPU systems bus, fabric, is becoming saturated which means Zen 5 architecture improvement would address this traditional sign of any processors design age, typical of time. And AMD proof of saturation, Zen 5 "re-pipelined front end and wide issue". New inter chip communications Zen 4 + RDNA lll may be early but if CPU bus saturation is the issue large off CPU caches are one solution. By the times its RDNA lll and CDNA lll its now CXL? mbBushLin - Sunday, June 12, 2022 - link
What are you basing that on? Just the >15% single thread improvement and extrapolating from there?Notice in the greater than symbol in the slide? 15% will be the minimum single threaded improvement.
Improved IPC, DDR5 and 5.5Ghz demonstrated across active cores (although not all cores) gives more than that alone... They're being conservative.
Bruzzone - Tuesday, June 14, 2022 - link
Bushlin, I don't doubt AMD is being conservation on my Zen 4 cost : price / margin estimates. TSMC + 20% price increase into second half 2022. Sumco + 30% wafer price increase. Zen 7 to 5 nm shrink + 40% dice per wafer. The performance increase has to be better than 15% to make up for the cost increase despite the good die increase I'm still working the estimates. mbQasar - Wednesday, June 15, 2022 - link
BushLin " What are you basing that on? " brazone is basing it on numbers he pulls out of thin air, so no one else can confirm what he comes up with. ignore him. he will NEVER post links to where he gets his info from, so other people can also come to the same results and conclusions. he is a fraud.Bruzzone - Wednesday, June 15, 2022 - link
Qasar, You're welcome to read TSMC node to node power v frequency and density claims, industry news on fabrication CapEx and materials cost increases, perform a good dice estimate and post your results. mbQasar - Thursday, June 16, 2022 - link
figures, STILL no direct links to your sources. why ? oh wait, cause you DONT HAVE ANY SOURCES ! searching for it, i could and up finding different numbers and such then you did. must be fun for you to post fud knowing no one can confirm or deny what you come up with.face it your are just a fraud, and a couch potato " analyst ". any thing you post is made up, with numbers you pull out of thin air.
Tams80 - Friday, June 10, 2022 - link
I haven't seen anyone calling Zen 4 a massive improvement and we all knew it wouldn't be until Zen 5 that we'd get a new architecture.Perhaps you found one or two morons.
But mainly, from your past posts, I think you just like sniffing your own farts.
Kangal - Friday, June 10, 2022 - link
Plenty of morons around. I found the posts, and realised it was on another site. But you are right about the farts.Khanan - Friday, June 10, 2022 - link
Zen 4 isn’t a refresh of Zen 3. Higher IPC different ram controller, way higher clocks, different instructions. Either you didn’t read the article or don’t understand much about CPUs in general.mode_13h - Saturday, June 11, 2022 - link
> I said Zen4 was a refresh of Zen3 on the new platformIt's clearly more than a simple port of Zen3 to 5 nm. We already saw how little benefit the Ryzen 6000's gain from the 6 nm port, and if you go back to Zen+ (i.e. 2000-series non-APUs), you'll see that 12 nm also offered only a couple %.
So, whatever they did, it might not be a roots-and-branches overhaul, but it's also a lot more than a straight port + some AI instructions.
GeoffreyA - Saturday, June 11, 2022 - link
Likely, the usual widening of the retire queue, more scheduler entries, improved branch prediction, and bigger micro-op cache.GeoffreyA - Saturday, June 11, 2022 - link
+ bigger register files, etc.Shorty_ - Saturday, June 11, 2022 - link
I don't get why you're all so upset, they spelt this out as their plan in 2019:He also indicated that AMD's server CPU launches are set to rely on the "tick-tock" cadence that was once the hallmark of Intel CPU launches, with the launch of a CPU platform that relies on a new manufacturing process node but the same microarchitecture as the last platform (the "tick") followed by a platform that relies on a new microarchitecture but the same manufacturing process node (the "tock").
(https://realmoney.thestreet.com/investing/technolo...
michael2k - Thursday, June 9, 2022 - link
I was hoping to see a BIG.little design from them.Dolda2000 - Thursday, June 9, 2022 - link
I don't think they need it, at least not yet. There was an article at Chips & Cheese that elucidated how Zen 2 (yes 2, not 3!) was extremely competitive in total energy use with Gracemont, and that Zen 2 on mobile blew completely past Gracemont on every power level.qwerty109 - Friday, June 10, 2022 - link
It is little understood reality (and intentionally obfuscated by Intel) that E cores (Atoms) are not actually energy efficient - they're actually worse or similar to big P-cores. However, they're a lot smaller - they are die space Efficient.The whole BIG.little on x86 is not there for energy efficiency like on ARM/mobile but exist so Intel can win on Cinebench and heavily multi-threaded workloads because P-cores can't compete with Zen otherwise. They can't compete because they've grown the P-core to ridiculous size in the hunt for IPC, and there was no way to back out of that (and losing process advantage to TSMC didn't help).
schujj07 - Friday, June 10, 2022 - link
Intel needs the BIG.little to help reign in their power consumption. Their P cores when running all out on AVX workloads suck a lot of power. The 12900K with 8P+8E cores sucks down 272W in 16c/24t configuration and 239W in 8c/16t mode. The E cores are using about 4W/core but the P cores are using ~30W/core. If Intel wanted to do a 10P core chip it would be breaking 300W under full AVX. The Core uArch hasn't been really that power efficient since Broadwell. It wasn't designed to go over 4c/8t at first and over 4.0GHz its power curve looks bad. However, due to pressure from Zen, which was designed around efficiency and 8c/16t, Intel had to push Core to the limits.Bruzzone - Wednesday, June 15, 2022 - link
mbDolda2000 - Friday, June 10, 2022 - link
>they're actually worse or similar to big P-coresThe same Chips & Cheese article I alluded to showed Gracemont using half the energy of Golden Cove in similar workloads, so I think "worse than P-cores", at least, would be an exaggeration.
Login - Friday, June 10, 2022 - link
It makes some sensehttps://i0.wp.com/chipsandcheese.com/wp-content/up...
Dolda2000 - Saturday, June 18, 2022 - link
Gracemont certainly did do worse in AVX efficiency than otherwise, but I think that is to be expected, as energy use in AVX workloads is probably going to be dominated by the execution unit itself, and therefore isn't going to be significantly different just because it has been integrated into a different microarchitecture. I think the more pertinent picture is https://i0.wp.com/chipsandcheese.com/wp-content/up... which, being integer-bound, is more likely to show the energy use of the whole out-of-order engine and whatnot, rather than that of a specific execution unit.That being said, if you look at the cumulative energy use of the libx264 execution, it looks more impressive than on the instantaneous power draw chart and shows that Gracemont is, at least, *more* efficient (if not by a huge amount): https://i0.wp.com/chipsandcheese.com/wp-content/up...
mode_13h - Monday, June 20, 2022 - link
Thanks for sharing! Very interesting that Gracemont seems designed to clock only up to about 3.0 GHz, while the knee in Golden Cove's energy curve comes at about about 4.2 GHz.It's also interesting to look at where those curves first start to climb. In Gracemont's case, it begins a gradual climb above 1.2 GHz, whereas Golden Cove doesn't markedly increase energy usage until about 2.0 GHz and 1.4 GHz. Those would be the peak-efficiency points (at least clock-speed wise - those graphs don't show actual performance).
One thing I didn't expect was for the slopes to be so similar, for so long. However, it's hard to read much into that, without knowing how actual performance scales with clocks. Ultimately, what we care about is how much perf/W the cores are delivering.
ian9298 - Friday, June 10, 2022 - link
Intel clocked E core way too high, to the point they were not efficient any moretechjunkie123 - Thursday, June 9, 2022 - link
That's what the 4 and 4c designs will be, but mainstream users would probably not use all the 4c cores, so they are sticking to big cores (little cores on server for nT tasks).More importantly, as another commenter said, zen 2 and (presumably, but less than zen 2) zen 3 are already pretty efficient.
Intel's big.little seem to have missed the mark in terms of efficiency, at least this generation. It allowed them to get solid 1T and nT performance, but the efficiency and hence laptop battery life is not great. Hopefully next generation will be more impressive in this regards.
Kangal - Sunday, June 12, 2022 - link
I wonder if in the distant future, would AMD come up with something for portables.Specifically for the 10in-18in portables, running anywhere from 7W-70W power. Hypothetically, they might be able to do a BIG.little design but with a mixture of x86 and ARM cores. For example; 15W for a chipset with 2x "Zen7" BIG Cores and 8x "Cortex-A750" little cores. And more "gaming" oriented Laptops (35W) could have a 4x Zen Cores with 8x ARM cores, for modern gaming engines which use more threads to work more appropriately. This all would run on the new Windows20 Operating System, where everything can run on the ARM cores with decent performance and low energy drain, but Programs needing to run Legacy Mode, would be shifted to the x86 cores. When maximum performance is required, the x86 cores are prioritised for that activity, and the ARM cores are pressed to handle the background tasks. This way we get the best of both worlds. Meanwhile, Full-sized Desktops would have an all-x86 chipset since they don't have the limitations of thermals, power, and portability.
Obviously this is just pie in the sky thinking at this point, but it might be a possibility someday.
michael2k - Sunday, June 12, 2022 - link
Since the 4C design is socket compatible it, naively, appears that the 4C design is designed to consume and shed 3/4 the energy/heat.DannyH246 - Friday, June 10, 2022 - link
Why? Introduce all that extra complexity for what? Currently their big cores are more power efficient that Intel's little cores. If you want extreme multithreaded performance then go with Zen 4c. But standard Zen4 will have excellent single threaded performance, and excellent multi threaded performance.michael2k - Sunday, June 12, 2022 - link
Fundamentally it’s for performance.There are always going to be time insensitive or low performance tasks like file copying, network operations, and UI handling. If you can allocate those tasks to a 10W core instead of a 30W core you get an additional 20W, of both heat and energy, you can spend on a 30W core.
In other words if your system throttles after 3 seconds at max performance, that extra 20W budget may increase the time your 30W cores can run at full throttle for an additional 4 seconds. Or maybe it can run at 4/5 performance without ever throttling.
Khanan - Friday, June 10, 2022 - link
No thanks. That shit will hopefully not be needed with AMD.michael2k - Sunday, June 12, 2022 - link
I imagine they will, and due to their chiplet design it shouldn’t have any extra complexity. They can allocate low resource tasks to efficiency cores and allocate the extra watts saved towards the performance cores.mode_13h - Sunday, June 12, 2022 - link
Well, if they were, then you'd expect they would've announced an efficiency core by now.As it stands, their answer to Intel's E-cores is the Zen 4c, and we have yet to see how its efficiency really compares to the normal Zen 4. It might not be a big enough gap to justify having a Big.Little configuration.
michael2k - Sunday, June 12, 2022 - link
If we assume the 4C is their efficiency core we can compare their announced products. Their performance core maxes out at 96 cores in a pin compatible part vs 128 efficiency cores. The assumption is that if both parts consume the same amount of energy then their e cores consume 3/4 the energy and emit 3/4 the heat of their p cores.If they further downclock the e cores slightly, they will then consume half, possibly even a third, of the energy.
So if they have a 1:7 e/p ratio we could end up with 12e/84p; because the e cores would be running OS tasks, networking and file copies and other time insensitive tasks, the energy previously wasted by running them on a p core can instead be allocated to one of the existing p cores. It’s possible those 84 p cores don’t throttle any more compared to a 96 p core part if there is enough surplus energy budget.
mode_13h - Monday, June 13, 2022 - link
> If they further downclock the e cores slightly, they will then consume half,> possibly even a third, of the energy.
That downclocking should apply equally to the regular Zen 4 cores. So, we're back to your original figure of 3/4th the power, and we don't know how much they'd have *already* been downclocked, just to hit that target! So, 3/4ths is probably more of a best-case estimate than a starting point.
Khanan - Monday, June 13, 2022 - link
Then you’re wrong. Big little is a old design, if AMD wanted it they would’ve announced or made something at this point but they didn’t. They have also publicly said that they are not planning to do so, because they don’t need it, their Regular cores are efficient enough and not bloated like Intels. Modern tech not outdated 1637361th gen core arch.Otritus - Thursday, June 9, 2022 - link
Based on history, AMD’s APUs using their architectures are always slightly more advanced and delayed than their desktop counterparts. From this I gather that the Zen 4 APUs will likely be on 4nm. Zen 5 is likely on both nodes, either because TSMC N3 is facing delays or because the APUs will be a ‘tick’ on 3nm with the CPUs a ‘tock’ on 4nm.Silver5urfer - Thursday, June 9, 2022 - link
AMD is now showing the muscle. No question.Their map is so solid now, Intel on the other hand is bragging about Foveros and other things. And on the consumer side more of that Biglittle junk because they cannot fix Core uArch to scale past 8P anymore. Anyways Zen 4, 5 with V-Cache was not surprising but Genoa-X 1GB L3 was because I didn't expect them to reveal it before even Genoa launched.
No more big.little bs rumors, AMD is using 4c for Bergamo as we all know 5c will also have similar successor. Because ARM's Neoverse V1 and N2 are coming with high density cores. AMD want's to cull them before itself, good move. x86 needs to reign this space and consumer space else the insane BS fragmentation and software issues with compat will show up. And AMD doesn't need to break the CPUs as they provide full high performance x86 cores without the added "efficiency" baggage on a Desktop. Intel is meanwhile stuck until they find a replacement for the Core series uArch.
At this point not sure if their AM5 socket with 1718 pins will stick until 2024. I hope they do, that would make Intel pathetic given how they are going to axe LGA1700 / 1800 pins total for a new LGA1851 socket ripping off people again like LGA1511 disasters. Shame when ILM is busted on Alder Lake.
For consumer Zen 4 Raphael / Ryzen 7000, I'm waiting to see how the clocks will stay now. Also most importantly their IMC and Firmware. The IOD is now TSMC 6nm, hopefully all the dreaded garbage WHEA and USB nonsense is gone. Also AGESA 1.2.0.7 has it's share of issues apparently despite aiming to fix TPM and stutters. This is the main weakness of AMD, garbage firmware and poor tuning experience. CO is very much head ache given how the clock behavior is on AMD and the heat on Zen 3 scales heavily dependent on that unlike Intel which can stick to a given clock and poor IMC with annoying XMP issues vs Intel again.
lemurbutton - Friday, June 10, 2022 - link
LOL @ this AMD fanboy.Who cares about AMD vs Intel anymore? Wake up. M series is destroying both. I doubt AMD can catch up to the M1 until Zen6. By then, M5 will be out since Apple will do a yearly update.
lemurbutton - Friday, June 10, 2022 - link
I expect AMD to finally match M1's perf/watt when it's Zen6 and using second-gen 3nm.By then, Apple will be on M5 using 2nm.
BushLin - Saturday, June 11, 2022 - link
The 15w M1 was trading blows with the already launched 15w Renoir (Zen 2) in 2020.https://www.anandtech.com/show/16252/mac-mini-appl...
Fantastic claims require fantastic evidence and I've not seen anything that comes close to backing up any of what you're claiming.
Qasar - Saturday, June 11, 2022 - link
BushLin, he cant, cause he knows his posts are bs, and just fanboy postsdr.denton - Friday, June 10, 2022 - link
lol, who cares about Apple? They have 9% market share on desktop and 15% in laptops. Everything else is x86.Would an M1 equipped Macbook be amazing if it could run all my stuff natively (including thousands of indie games) and I had 2000€ to spare? Sure. But for most people those are BIG ifs.
Speedfriend - Friday, June 10, 2022 - link
Who cares about Apple....? It is not Apple that is the issue for AMD and Intel, it is Qualcomm and what they can do with Nuvia. If that is competitive with M2 in terms of performance and efficiency when it launches, there is going to be a shift to Windows on ARM. My high powered x86 laptop is a total joke on battery, it can barely last a 2 hour meeting, let alone a conference. Ill be switching as soon as possibleKangal - Friday, June 10, 2022 - link
Qualcomm is run by some greedy corporate types, and most of their talented designers have left the company a few years ago. The Nuvia cores did look promising and they were supposed to be released ages ago, until Qualcomm basically bought them out.I just don't expect much to come from Qualcomm-Windows. You are better off sticking with Android instead for the Efficiency, UI, and Apps. Or if you need proper Windows, stick with the x86_64 platform instead of an emulated-environment.
OreoCookie - Friday, June 10, 2022 - link
Intel and AMD do, for one. Apple shows what a good architecture can do in terms of performance, performance-per-watt, dedicated logic and what trends are. Intel referenced Apple‘s M1 specifically in internal slides as what they are shooting for (some years later).I expect that Apple‘s decision to dramatically increase memory bandwidth will have an impact on AMD‘s and Intel‘s decisions as well. Even the M2 (without modifier) has 100 GB/s memory bandwidth, which is close to what AnandTech measured in AMD’s Milan. (I realize I compare theoretical maximums with actually measured throughputs, but my point is more about where to put the emphasis in design.) And it stands to reason that the M2 Pro/Max/Ultra will at least match the M1 variants.
BushLin - Saturday, June 11, 2022 - link
If the 100GB/s bandwidth figure is anything like what was marketed on the M1 Max then only a fraction of that is available to the CPU cores.M1 Max headline number from Apple is 400GB/s but the reality is 100GB/s single threaded and <250GB/s best case scenario with all cores loaded.
https://www.anandtech.com/show/17024/apple-m1-max-...
Not saying the performance numbers won't be impressive for a little SoC but far from the Jesus silicon some fantasize might about.
Not aimed at the poster above but it helps to post links with actual results rather than slinging around far fetched claims.
OreoCookie - Saturday, June 11, 2022 - link
Fair point, and I should have mentioned that.However, my point was two-fold: my comparison to measurements in Milan was to make the point that Apple decided to spend silicon to give the M2 — Apple’s smallest non-iPhone SoC — roughly the same type of memory bandwidth as a server-class x86 CPU — or way more in case of the Mx modifier variants. So rather than spending more transistors and power optimizing the CPU, Apple dedicates them to memory and caches.
And secondly, the large increase in bandwidth means that memory bandwidth is not a bottleneck for total performance (true at least for the M1 Max and M1 Ultra, because even when CPU, GPU and other specialized logic is running full tilt, the SoC still has enough memory bandwidth to saturate them all). The same cannot be said for many of AMD‘s and Intel‘s CPUs (at least the ones with higher core counts and/or integrated GPUs).
My prediction is that Intel and AMD will follow suit in the future and give more emphasis to memory bandwidth, i. e. Apple‘s SoCs definitely do influence what Intel and AMD are doing — which is what I was replying to. You can also see that Intel and AMD are trying to integrate more specialized circuitry for specialized tasks as well, also something that Apple has emphasized (in terms of transistors and power) on its M-series SoCs.
mode_13h - Sunday, June 12, 2022 - link
> roughly the same type of memory bandwidth as a server-class x86 CPUNot at all. It's only workstation-class. EPYC Milan and last year's Ice Lake SP both have 8-channel DDR4, providing nominal bandwidth of 204.8 GB/s. Cut that in half and you get something like Threadripper (non-Pro) bandwidth numbers.
DDR5, with Sapphire Rapids shipping probably around the same time as M2-based systems, will move the bar yet again. With it, even desktop platforms are creeping into M2 territory. Alder Lake has a nominal bandwidth of 76.8 GB/s.
So, Apple is really only barely keeping pace, considering graphics needs a share of that bandwidth. For perspective, the PS4 that launched almost 9 years ago had 176 GB/s unified memory bandwidth.
You need to get with the times.
schujj07 - Sunday, June 12, 2022 - link
Zen 4 Epyc has even more bandwidth as it will have 12 channel DDR5.OreoCookie - Wednesday, June 15, 2022 - link
“Not at all. It's only workstation-class.”Anandtech tested memory bandwidth of two Milan EPYC 7000-series processors, i. e. AMD's server processors, and they were tested at 103~113 GB/s peak throughput (https://www.anandtech.com/show/16529/amd-epyc-mila... Your figures are the theoretical max bandwidth. To be fair, I could not find figures for tested bandwidth for the plain M1. On the other hand, to be fair, we are comparing Apple's second-gen lowest end chips meant for its entry-level notebooks and desktops to AMD's current-gen highest-performing chips. Extrapolating from the the numbers of the M1 Max, the M1 Ultra should have about twice the sustained throughput due to the CPU complex only than AMD's current top-of-the-line server CPUs.
“So, Apple is really only barely keeping pace, considering graphics needs a share of that bandwidth. For perspective, the PS4 that launched almost 9 years ago had 176 GB/s unified memory bandwidth.”
Barely keeping pace? Apple's highest-end SoC has a theoretical max memory bandwidth of 800 GB/s, which is comparable with a current top-of-the-line graphics card like nVidia's RTX 3080 (about 912 GB/s).
And why don't we look at the PS5 instead of the PS4? The PS5 has a theoretical max memory throughput of 448 GB/s, roughly on par with the M1 Max. So is Apple really behind? Doesn't look like it to me. Especially once you include efficiency in your considerations.
mode_13h - Wednesday, June 15, 2022 - link
> Anandtech tested memory bandwidth of two Milan EPYC 7000-series processors,> i. e. AMD's server processors, and they were tested at 103~113 GB/s peak throughput
So, you're comparing Apple's theoretical bandwidth to AMD's actual bandwidth? Nice. Try again.
Also, to interpret benchmarks, you need to know what you're looking at. The Stream Triad benchmark is meant to stress the cache subsystem. It assigns each thread to do a series of 2 reads + 1 write. In copy-back caches, the write actually generates an extra read. There are ways around this, but the benchmark is written in a naive fashion.
So, to achieve 113.1 GB/s on that benchmark, the underlying memory system is actually doing 150.8 GB/s. As mentioned in the article, Altra has a hardware optimization which detects the memory access pattern and eliminates the extra read. That's the main reason it's able to achieve 174.4 GB/s, using the same 8x DDR4-3200 memory configuration as the Epyc. Also, the Epyc is being run in the generally slower NPS1 topology, meaning traffic is crossing all of the NUMA sub-domains.
Now, as for the rest of the margin, you seem to have fallen in the trap of citing the launch review of Milan, which used a pre-production platform. Unfortunately, they didn't rerun the Stream.Triad benchmark for the updated benchmarks, but if you compare the SPECint results, multithreaded performance between the two articles jumped by 7.9%. It's hard to say what difference there'd have been in memory performance, but we can use that to get a rough idea.
https://www.anandtech.com/show/16778/amd-epyc-mila...
> Barely keeping pace?
Yes, because Intel mainstream desktop CPUs have a nominal bandwidth almost as high. Everyone gets a boost, moving to DDR5. They are not special. 100 vs. 76.8 is significant, but not the massive difference you tried to paint.
> Apple's highest-end SoC ...
Uh oh. Now you go and change the subject. Irrelevant.
This article is about the M2 and I was simply putting it into perspective. I take it you're offended by this or what it reveals, because otherwise I don't see why you'd drag completely different Apple chips into the discussion. Seems like a face-saving move, but ultimately just makes you look desperate.
mode_13h - Wednesday, June 15, 2022 - link
And BTW, I'm even going to retract what I said about "workstation bandwidth", because a quad-channel DDR5 setup is going to spec out north of 150 GB/s. So, if we're comparing like-for-like, rather than against old DDR4-based products about to be replaced, then it doesn't even reach the workstation tier.lemurbutton - Friday, June 10, 2022 - link
Apple is winning the premium laptop market and is growing significantly faster than x86 laptop makers. In a few years, Mac marketshare will look very different.GeoffreyA - Friday, June 10, 2022 - link
So, Apple and their fearsome M have destroyed both AMD and Intel. The tide has turned. Dear friends, who knew this day would come so swiftly? The only thing we can do now is take all our x86 computers and, with a sigh, throw them in the bin as fast as we can because they're useless now.schujj07 - Friday, June 10, 2022 - link
Inside of the MAC universe the M series is king. However, what happens when it has to run on an OS that isn't highly tuned like OSX or software that doesn't have the fine tuning? Probably the same thing that happens to every ARM uArch, it bogs down to 2015 x86 performance level.web2dot0 - Saturday, June 11, 2022 - link
You can't prove a negative buddy. Where did you learn to reason?No amount of tuning can you get away with a 400GB/s memory bandwidth DDR5, ANE, and Media Engines.
Performance per watt is undisputed.
If you are gonna criticize Apple, pick something more tangible than a bunch of hypotheticals.
Oxford Guy - Saturday, June 11, 2022 - link
Sally: '8 - 5 = 3.'Amy: 'Sally, you're silly. Everyone knows you can't prove a negative!'
Oxford Guy - Saturday, June 11, 2022 - link
Stripping away features isn't 'fine tuning'.Look at the disaster that is the Music program, created by Apple to replace iTunes — without users having any choice.
mode_13h - Sunday, June 12, 2022 - link
> what happens when it has to run on an OS that isn't highly tuned like OSX ... ?Luckily, someone tested exactly that: https://www.phoronix.com/scan.php?page=article&...
That article first analyzes the penalty of running under Linux vs. MacOS, then compares Linux performance to a stack of x86 machines. Currently, it's delivering 10th or 11th gen i5 performance, under Linux. Not bad, for a community effort.
schujj07 - Sunday, June 12, 2022 - link
Most of the time the M1 and Ryzen 3 3300X are close in performance. That puts the M1, with 4+4, around the performance of the top low end CPU, 4c/8t, from AMD in April 2020. There are times when the M1 performance is higher but there are an equal amount when it is lower.Bruzzone - Wednesday, June 15, 2022 - link
3300X target in the channel to take sales from another Intel trading grade SKU as a package dud R9 save is Ivy EE. mbdudedud - Friday, June 10, 2022 - link
It won't.Apple has had the premium smartphone market for more than a decade and its share has remained quite similar since then.
And Mac is not as popular as iPhones.
schujj07 - Friday, June 10, 2022 - link
Don't forget that outside of the USA the iPhone market share is minimal and so is Mac.Lord of the Bored - Saturday, June 11, 2022 - link
Apple's market share has been on the cusp of a dramatic explosion for decades now.To be blunt, Apple doesn't WANT to be mainstream. If everyone has a Mac, it isn't special and doesn't inspire cult-like devotion. Also, they'd have to give up profit margins to secure a toehold in more cost-sensitive market segments.
web2dot0 - Saturday, June 11, 2022 - link
That's the definition of the "PREMIUM" market. How many people can afford premium products? 😂Their whole schtick is the high-profit margin products. Even their "low-profit margin" products have a higher profit margins than their competitors.
You make it sound like it's a problem.
GeoffreyA - Saturday, June 11, 2022 - link
That's the whole point. Apple, quite consciously, cultivates this aura that they're better than everyone else, and that's partly why many folk buy their products. Now, if Apple were a humble firm, nobody would mind and they'd likely be applauded by the Win/x86 camp; but they're not, believe they're the best, and think there's nobody like them, with all their bought-off companies.mode_13h - Sunday, June 12, 2022 - link
> Apple doesn't WANT to be mainstream.LOL. The iPhone proves you wrong. It's *very* mainstream, which proves they (and their investors) would happily settle for a position of market dominance over premium prestige.
SiliconFly - Tuesday, July 5, 2022 - link
Keep on dreaming. M2 is crap compared to the competitiontechjunkie123 - Friday, June 10, 2022 - link
Perf per watt is actually fairly similar for nT workloads between M1, AMD 6000 series and alder lake H. The difference is in 1T and in graphics and in software optimization. The push for higher clock speeds is not going to help AMDs efficiency with Zen 4, unfortunately.....there's a reason apple clocks it's CPUs at ~3-3.5 Ghz.lemurbutton - Friday, June 10, 2022 - link
You're joking right?M1 is about 5x better perf/watt than Zen3 mobile chips.
Don't use max TDP to compare perf/watt. Use total package power used. When running Geekbench5 ST, M1 uses ~0.5w package power for most of the test.
lemurbutton - Friday, June 10, 2022 - link
Not only this, M1 beats 5950x in ST while running fanless!No way AMD can match this kind of efficiency until Zen6.
AdrianBc - Friday, June 10, 2022 - link
This is complete BS.M1 has never beaten even 5900X in ST, much less 5950X/
I know this for sure, because I have a 5900X. In all published benchmarks where I have access to the benchmark programs, so I can run them, a 5900X is slightly faster in ST than any M1. e.g. in GB5 ST the score is around 1800 for 5900X (non-overclocked, non-PBO) vs. around 1750 for M1.
Obviously, it is true that a desktop AMD CPU must consume maybe 4 to 5 times more power than an M1. when running in ST, to be able to beat it, but this does not justify exaggerations like claiming that M1 is better in absolute performance.
lemurbutton - Friday, June 10, 2022 - link
Oh wow. Look into actual boost TDP of AMD mobile CPUs, let alone desktop CPUs. They boost way beyond their stated TDP.Anyways, you should be looking at total power used by the package anyways, not TDP.
M1 is 5x more efficient. Let's not even talk about M2.
BushLin - Saturday, June 11, 2022 - link
Please could you provide some credible evidence for any of that, especially 0.5w package power while supposedly winning any kind of benchmark... The rest is only slightly less realistic and still hard to take seriously. Presumably you have some solid benchmark results with accurate power measurements including something like a Ryzen 6800U.web2dot0 - Saturday, June 11, 2022 - link
You are grasping straws. It is empirically more power-efficient than any Intel/AMD laptop. We can play that game all day about this benchmark and that benchmark.But no sane person would think that their laptop, running under similar conditions, can last longer than an M1/M2 MBA.
It's a pointless debate.
BushLin - Saturday, June 11, 2022 - link
Interesting take on someone challenging nonsense.When the M1 was released, it was impressive for the single thread performance, wide design and general efficiency; albeit with the double edged sword of closely integrated RAM and frankly anti-consumer non-upgradeable and proprietary SSD.
Those are the facts; it was a momentous release for Apple and I've recommended the Macbook Air to people who don't have issues with the software library...
However, this thread has someone claiming 5x better performance per watt!!! Be real. Objectivity is in short supply for some it seems.
Also, the article is about Zen 4 ffs.. For what it's worth, AMD is sandbagging.
>15% single thread improvement will be a worst case scenario and Zen 4 will absolutely stomp on Intel for multithread (I don't know about Apple's future plans). Raptor Lake is delayed so won't get a chance to take the crown before Zen 4 is released and it only gets worse from there. I'm not taking credit for those predictions; "Moore's Law is Dead" is rarely wrong, has contacts throughout the industry and is super confident about those events playing out. I actually think this could be bad for consumers because we now need Intel to step up to keep AMD's pricing competitive.
SiliconFly - Tuesday, July 5, 2022 - link
True. When M1 launched, it was good. All the M series processors that followed after that weren't that great. In fact, the new M2 is dismal. It's only going to get worse for apple when you see AMD & Intel roadmap.biostud - Friday, June 10, 2022 - link
Where do I install my video card on the Apple computer?AdrianBc - Friday, June 10, 2022 - link
M1 may be about 5x better perf/watt, but as the other poster has written, this is true only for single-thread benchmarks, where the Intel or AMD CPUs approach 5 GHz, having low energy efficiency.As the other poster has said, in multi-threaded benchmarks, and if the power limits have not been raised above the nominal TDP, the clock frequency drops and also the power consumption is reduced to a few watts per core.
In this case, the Apple CPUs remain more efficient, but their advantage on perf/watt becomes significantly less than double, so very far from the "5x" claimed above.
Moreover, 5x is not true even for single-thread. In the Apple presentation from the beginning of the week, Apple has acknowledged that their new improved M2 has a perf/watt advantage over a 28-W Alder Lake P of only 3.6 times, i.e. much less than 5 times.
If the advantage is 3.6 times vs. Alder Lake P, than it is less than 3.6 times vs. Ryzen 6000 and it will also be less than 3.6 times vs. the future Intel Raptor Lake, which will be the competitor of Apple M2 for most of its lifetime.
For multithreaded applications the perf/watt advantage of M2 cannot exceed 1.5 to 2 times.
There is no doubt that Apple M2 will have better performance per watt than any Intel or AMD CPU, at least until the second half of 2023, when Intel will launch Meteor Lake and AMD will launch Zen 5.
This true fact does not justify claims of fantastic numbers for M1 or M2.
Also, M1 can use 0.5 W total package power on GB5 ST only if the benchmark runs on one of the small cores. Any large Apple core consumes alone at least 4 to 5 W (when in iPhones, more in laptops) when running a ST benchmark. Look at all the reviews ever published on Anandtech to see that.
BushLin - Saturday, June 11, 2022 - link
Is 5x better performance per watt even a credible claim in the first place? Even if not pulled out of someone's rear, is that comparing a mobile CPU a high end desktop CPU which scales to way higher performance, in a very specific hand-picked scenario?techjunkie123 - Friday, June 10, 2022 - link
See AdrianBC's response below, and read this article: https://www.notebookcheck.net/AMD-Ryzen-7-6800U-Ef...1T perf/watt is about 3 times higher, with core power consumption being about 2 times higher, while nT perf/watt is similar. Both in CB23, which probably disfavors apple by about 20% compared to Geekbench for performance, but that's not a big change.
lemurbutton - Friday, June 10, 2022 - link
Cinebench is terrible.Use Geekbench or SPEC.
schujj07 - Friday, June 10, 2022 - link
Both of them are terrible as well.techjunkie123 - Friday, June 10, 2022 - link
Doesn't change what I said. M1 Geekbench performance will only be 20% or so higher relative to cinebench.Tams80 - Friday, June 10, 2022 - link
Watch who you call a fanboy...Is there something medically wrong with you?
Fulljack - Friday, June 10, 2022 - link
LOL @ this Apple fanboyWho cares about Apple anyway? Wake up. M series doesn't support dedicated graphic cards nor mainstream OS like Windows or Linux distros. I doubt Apple would open up M-series at all. by then, they're on their own. especially when Apple didn't do yearly update on M-series anyway — did 19 months after original release considered as yearly release?
you're just a fanboy as much as others.
web2dot0 - Saturday, June 11, 2022 - link
Apple hasn't supported Windows or Linux distros ... for years now.They seem to be doing pretty well for themselves.
Why do people continue to use the same argument over and over. They are successful DESPITE what you said. So clearly, those aren't their weaknsesses. ;-)
mode_13h - Sunday, June 12, 2022 - link
> M series doesn't support ... mainstream OS like Windows or Linux distros.They support it running in VMs.
Khanan - Friday, June 10, 2022 - link
Lemurbutton you’re one of the biggest trolls in the tech world. Please talk less, you have no clue about tech.SiliconFly - Tuesday, July 5, 2022 - link
Apple M series is a joke. Compare the dismal M2 to the the current AMD & Intel offerings and it's substandard.GeoffreyA - Friday, June 10, 2022 - link
"Also AGESA 1.2.0.7 has it's share of issues"I refrained from the beta BIOSes for two years on my B450 Tomahawk + Raven Ridge, but about a fortnight ago upgraded to 1207, which, as far as I can see, has been working smoothly. I believe the 1206x branch had issues.
Khanan - Friday, June 10, 2022 - link
Agreed, btw I’m running Agesa 1.2.0.7 right now and so far it’s fine.JayNor - Thursday, June 9, 2022 - link
Why do the slides omit 3D on the zen4 v-cache designs?techjunkie123 - Friday, June 10, 2022 - link
Probably implied. It's been done before at this point so the idea is not new.del42sa - Friday, June 10, 2022 - link
8% IPC in desktophttps://www.cnews.cz/wp-content/uploads/2022/06/Pr...
aritex - Friday, June 10, 2022 - link
Zen4 : 12 cores per chiplet. Die size is nearly the same. 25% more bandwidth PER CORE. 96 core Epyc = 8 * 12schujj07 - Friday, June 10, 2022 - link
I believe they said it is staying at 8c per chiplet. The 96c will be 12 * 8c which isn't surprising since AMD said it will be 12x DDR5 RAM channels.techjunkie123 - Friday, June 10, 2022 - link
Exactly as comment above, staying at 8c per chiplet. Zen 4c will be 16c per chiplet.mode_13h - Friday, June 10, 2022 - link
> the message from AMD is clear: they’re going to be doing some significant reworking> of their core CPU architecture in order to further improve their performance
> as well as energy efficiency.
No, it means they *already* did some significant reworking. From the Oct 26th interview with Mike Clark, he indicated the architecture of Zen 5 was already done by that time:
"I have this annual architecture meeting where we go over everything that's going on, and at one of them (I won't say when) the team and I went through Zen 5. ... Coming out of that meeting, I just wanted to close my eyes, go to sleep, and then wake up and buy this thing."
That he wouldn't disclose *when* suggests the meeting was some months prior.
https://www.anandtech.com/show/17031/anandtech-int...
mode_13h - Friday, June 10, 2022 - link
Oh, and I forgot to include this part:Ian Cutress: "You mentioned in a publicity video in April 2018 for AMD that you were working on Zen 5."
So, whatever it's going to be is already baked-in.
Khanan - Friday, June 10, 2022 - link
It’s pretty amazing if you think about it.Khanan - Friday, June 10, 2022 - link
*whenmode_13h - Friday, June 10, 2022 - link
> on top of Zen 4’s new AI instructions, Zen 5 is integrating further AI and> machine learning optimizations.
I wonder to what extent this will continue to be relevant. I think they've felt pressure from their competitors and customers, but it's an open question how relevant CPUs will be for machine learning, by the time Zen 5 finally reaches the market. Everybody has AI accelerators, from chip vendors all the way to their biggest customers.
I think ARM did the right thing, here. They restricted their wide vector compute to their V-series cores, while their N-series - their bread-and-butter cloud-oriented cores - will have narrower implementations.
I expect AMD to use narrower vector compute in their c-series and APUs, while their chiplets have wider implementations.
marcovtje - Friday, June 10, 2022 - link
Maybe they'll implement AVX-512 based on the pipes of two AVX-256 units. Still there are an awful lot of instructions in avx512 (and which one?)AdrianBc - Friday, June 10, 2022 - link
On WikiChip it is claimed that Zen 4 will support exactly the same AVX-512 instruction subsets like Ice Lake, together with extra instructions introduced in Cooper Lake.The (few) newer extensions of Tiger Lake and Sapphire Rapids are not supported.
In an interview, AMD has confirmed that the instructions for AI that have been mentioned are those included in the two AVX-512 subsets introduced by Cascade Lake and by Cooper Lake.
Khanan - Friday, June 10, 2022 - link
Interesting, and it will be too funny when AMD has this stuff end of year in the desktop platform and intel hasn’t. You’re basically forced to buy AMD then for AVX512 or low core 11th gen or far outdated Hedt platform of intel. The choice is clearmode_13h - Saturday, June 11, 2022 - link
> it will be too funny when AMD has this stuff end of year in the desktop platform*if* they enable it on the desktop. I expect they will, but they could play games with market segmentation.
I'm hoping they don't nerf it, like Intel did by disabling one of the two FMAs per core, in the lower-end 14 nm CPUs that had it. That meant there were some cases where AVX-512 was a net performance penalty, because even using the single FMA would trigger clock throttling that slowed down the CPU by more than the benefit AVX-512 provided.
Khanan - Sunday, June 12, 2022 - link
AMD has more or less already said it will be in upcoming Ryzen 7000, it’s a 99% thing. And for anything else, we will have to wait and see. So far, Ryzen was executed solidly so I don’t expect otherwise.mode_13h - Saturday, June 11, 2022 - link
If AMD restricts it anywhere, I think it might be leaving it out of their APUs. But that would probably be about reducing die area, in which case the actual silicon wouldn't have it.Khanan - Sunday, June 12, 2022 - link
That makes more sense, yes. But I wouldn’t do it, it’s a leg up against Intel and a funny way to mock them.mode_13h - Saturday, June 11, 2022 - link
> Maybe they'll implement AVX-512 based on the pipes of two AVX-256 units.One thing that makes AVX-512 annoying is the bigger vector registers. Also, there are twice as many of them. So, even if you duplex AVX2 instructions down half as many AVX-512 pipes, it still doesn't negate the die area impact.
Plus, AVX-512 adds different types of instructions than AVX/AVX2, which means additional dedicated logic with no other purpose.
Fulljack - Friday, June 10, 2022 - link
I don't know how relevant AVX-512 on server or workstation space, but I certainly doubt it will be relevant in consumer space.still, I understand where AMD is coming from, though, as apparently they designed those chiplet to be used from Ryzen, Threadripper, and Epyc.
Khanan - Friday, June 10, 2022 - link
It’s no problem since it’s 2x256 fused together instead of one big 512 bit unit.mode_13h - Saturday, June 11, 2022 - link
It's not that simple. There are operations with AVX-512 that don't map cleanly on to AVX2. In some cases, it's due to structural changes in the instruction set. In others, it's new element data types. And there are things like scatter/gather that AVX2 simply doesn't have.BTW, AVX-512 even added support for the new instructions at narrower width. So, you can use them on 256-bit vectors without the full thermal impact of 512-bit processing.
Khanan - Sunday, June 12, 2022 - link
However, I don’t agree with the statement that if won’t be relevant in consumer space. Consumers can work with their PCs too, and the way normal desktop unfolded in the last years, more or less because of AMD, particularly the 16 core CPUs are for workstation not really for gaming, the least thing they are for is content creation. Intel is just a joke with their terrible efficiency, with Intel everything is just a complete joke, they make their top CPU a power hog which is unprofessional, it gives off a “gamey” vibe.mode_13h - Saturday, June 11, 2022 - link
It's pretty hard to make the case for AVX-512, outside of deep learning and HPC apps, I think.There are some AVX-512 optimized string processing libraries and it's useful in video compression. However, GPUs are better for video compression, when they support it (and you don't need the absolute best quality).
shabby - Friday, June 10, 2022 - link
No one's asking the real questions here... will we see a review of it on AnandTech?Khanan - Friday, June 10, 2022 - link
I wouldn’t bet on itmode_13h - Saturday, June 11, 2022 - link
I would expect so. Gavin Bonshor did the i7-12700K and i5-12600K review. So, it seems the reins have been successfully passed from Ian.https://www.anandtech.com/show/17267/the-intel-cor...
Khanan - Sunday, June 12, 2022 - link
Granted if he meant any Zen 4 CPUs, you won’t see Epyc reviews here.mode_13h - Sunday, June 12, 2022 - link
Yeah, I don't see a Milan-X review on here. We'll have to see if they review Sapphire Rapids. If not, then they probably won't review Genoa, either.schujj07 - Thursday, June 16, 2022 - link
There aren't very many reviews of Milan-X anywhere outside of what is provided by places like MS Azure. However, Milan-X was so much faster than regular Milan in HPC workloads that MS is replacing there almost brand new Milan systems with Milan-X (HBv3 instance types). https://www.servethehome.com/amd-milan-x-scaling-t...mode_13h - Friday, June 17, 2022 - link
Phoronix reviewed Milan-X.https://www.phoronix.com/scan.php?page=article&...
https://www.phoronix.com/scan.php?page=article&...
UNLK A6 - Friday, June 10, 2022 - link
More hilarious bar charts (sheesh).Khanan - Sunday, June 12, 2022 - link
It doesn’t matter so long they are true, with which AMD, is expected.Zoolook - Monday, June 13, 2022 - link
Well Apple was ahead of the curve, they are using lpddr5, nothing strange there and they are feeding the gpu with that bandwidth as well, they really need it.Khanan - Monday, June 13, 2022 - link
Apple is really good, but they aren’t exactly ahead in anything. AMD has far stronger CPUs. Even Intel and other companies that use ARM for server.dicobalt - Tuesday, June 14, 2022 - link
If they do another Bulldozer I'm going to have Will Smith give them all a slapping.mode_13h - Wednesday, June 15, 2022 - link
I think a confluence of factors and bad decisions resulted in bulldozer. AMD seems to have changed their ways, but they're still somewhat at the mercy of circumstances.Foeketijn - Saturday, June 18, 2022 - link
Please AMD, don't do an other bulldozer. Make something nice with ZEN 5.We now have a healthy competition. Lets leave it that way.