They're launching in just over a month, but no pricing info? And no technical details? Is there another event before launch where they're going to announce this, or is it just going to be put into a press release sometime?
Note that AMD hasn't used a trade show for a major product launch in years. Any trade show announcements have always been followed by an AMD event for the actual launch.
Most likely rather close 7000 series msrp + inflation. The 7000. Series is so much down from their starting prices that there is Room in original msrp price points, maybe little above those. That will make the 7000 series the best bang for money option and 9000 series the new flagship tier.
I don't see anywhere else reporting the base clocks, which are 200-800 MHz lower than their Zen 4 counterparts. Didn't see in slide deck. Where are these coming from?
I wouldn't worry too much about the drop in base clocks, since that's probably specified for AVX-512 "power virus" workloads. The other reason for the drop should be the reduction in TDP on all models except the 950X tier.
What's got me a little more worried is the lack of improvement in boost clocks, other than at the 700X and 600X tier, which each increased only 100 MHz. However, if these new CPUs actually have an easier time achieving and sustaining their boost clocks, perhaps it'll save AMD from some of the criticism it's gotten in the past for low-boosting CPUs.
They moved the CCD from N5 to an N4-class node, maybe N4P. There could be an architectural reason why they can't scale clocks up from Zen 4, or there is less benefit to hitting those higher clocks. So they get to focus on efficiency this time around. If it's doing 1.16x the work at 0.78x the power (N4P benefit), I guess it's a whopping 49% more efficient.
And I applaud this move. They obviously felt less need to chase Intel as they believe to have product that is superior enough without exploding on the TDP. TDP and heat were the least liked aspect of the previous series (well, maybe 2nd behind a huge drop in value proposition). I hope 65W Zen5 parts will bring back the expected balance in power/performance without having to undervolt or tweak power in any other way. Sadly I don't expect the same for the value. AMD is no longer the underdog and it shows on the price tag.
The perf/W scaling for the single-CCD Zen 4 CPUs is horrible, above 65 W TDP (88 W PPT). Maybe it's even worse, with Zen 5. Would you *really* want to burn more power for no benefit?
AMD processors use dual issue AVX512 and unlike Intel they do not need AVX offsets. There's no power virus aspect for AMD processors in this category from what I can recall.
Yes there's no boost clock improvement, but also regression in base clocks, perhaps the Zen 5 is just an overhaul of Zen 4 there's no new arch, its literally like Intel Core which they are using since 2006 and AMDs Zen is from 2017, iterative improvements rather than a clean slate design.
AMD processors had the higher base clocks unfortunate that they are not anymore. Perhaps they are targeting the stupid TDP nonsense. Shame really. Intel on the other hand goes overboard the power consumption and pushes their silicon to absolute limit at 400W and 300A and creates a total degradation and instability, AMD won't even try to push 50W more and get a better performance. Don't know if it's the TSMC node limit or silicon limit or the uArch limitation, as since Zen 4 these processors scale in performance by 94C Temp target.
Chipset side is also nothing big, its just a USB4 and Wifi M.2 upgrade, the base is same ASM/AMD dual chipset. Maybe Zen 6 ? Time will tell, since the SMT performance vs a 14900K is just nothing esp considering the fact how AMD always lead in SMT vs Intel.
It's possible the dropping of clock speeds went hand in hand with widening the core and increasing IPC. Perhaps the pipeline was shortened, though that seems doubtful. I think there won't be a drastic new architecture because the industry has seen, through trial and error, that high-performance CPU design falls along similar lines, whether it's x86 or ARM. Departures, like NetBurst and Bulldozer, have been painful.
It's really unclear to me why people care about base clocks. So far as I am concerned the lower the better. my CPU spends most of it's time sitting at 1-20% usage, only things like gaming and video/photo editing really spike it much. I'd rather use as little power as possible for day to day tasks so long as it can easily ramp up as soon as a workload demands it.
It's simple math, with an ipc increase of 15% more work gets done per cycle and if the perf/W improvement is less than the ipc increase, the base clock needs to go down to keep within the TDP.
Ugh, you're right. It seems all of ASRockRack's, Supermicro's, and Tyan's AM5 motherboards are no bigger than micro-ATX. I checked an ASUS Prime Pro ATX motherboard that supports ECC memory, but it doesn't include EPYC 4xx4 in the list of supported CPUs.
If you are PCIe slot limited on the 4004 series you would probably be a good candidate for the 8004 series instead. Otherwise from a cost point looking at the 7003 series is another choice.
The AM5 CPUs and higher-end chipsets have enough lanes to easily populate a regular ATX motherboard. Sometimes, you just need one more x4 slot, in which case EPYC or TR would be way overkill.
"To begin, let’s break down AVX10.N/M: AVX10 is the new “foundational” SIMD/vector instruction set for x86_64. The “.N” denotes the version of AVX10 as a version modifier, allowing incremental updates. It is important to note, if you support “AVX10.N+3,” you must support all of AVX10.N, N+1 and N+2. In simpler words, users are guaranteed supersets of previous instruction sets.
What does the “/M” mean? It’s a reference to vector register implementation size of a given AVX10.N version. Specifically, it may be 512-bit, 256-bit, or the topic of this article, 128-bit wide."
LOL. Don't matter for client workloads, because Intel has telegraphed that they'll only be using 256-bit in client processors. Therefore, software developers using AVX10/512 (which Intel pejoratively terms "legacy") will be only those targeting servers that could run AVX-512, also.
Now, when AMD does support AVX10, they could revert C-series cores to using a 256-bit wide implementation of AVX-512, just like Zen 4 did, and thereby maintain backwards compatibility without being at a density disadvantage on AVX10/256 relative to Intel.
"LOL. Don't matter for client workloads, because Intel has telegraphed that they'll only be using 256-bit in client processors."
AVX512 is becoming more useful and common on the server side. Since the Zen cores is virtually identical between server/desktop CPUs, it makes sense to keep it on both.
> Since the Zen cores is virtually identical between server/desktop CPUs, > it makes sense to keep it on both.
I'm not saying AMD should remove it. I'm saying that Intel found a way to nullify that advantage by creating FUD for people like game devs around the future of AVX-512 so they don't go and write a bunch of code that runs faster on AMD client processors.
Eventually, Intel will role out AVX10/256 and that's *just* incompatible enough not to work on AMD CPUs without hardware modifications. And if AMD does implement support for it, depending on how they added AVX-512, half of their vector pipelines might be sitting idle, meanwhile Intel's will be fully utilized.
I'm glad to see the increase in encoding performance. With all the hype behind Zen 5, I think people were expecting more across the board, but an average of 16% is not half bad.
there's some weird geekbench cherry picking going on there. I look forward to seeing independent reviews. My first reaction, though, is that Zen 5 is not the revolution some commenters were predicting and that AMD faces a possible Core2Duo-esque scenario with Arrow Lake on Intel's A20 process.
Luckily for AMD, they will have free reign for a few months until Arrow Lake launches, and they have X3D in their pocket. Arrow Lake is also launching on a new socket.
Yes, but a few months out of a two year product cycle isn’t much — especially since AMD might never regain the process lead they’ve enjoyed these last several years.
"Expecting 20% to 40% IPC increase would be pretty nuts"
Who said anything about specifically IPC? Expecting 20% total performance increase is VERY VERY low expectations for 2 years of development. And given that frequencies has decreased, looks like total performance stays about the same.
> Expecting 20% total performance increase is VERY VERY low expectations for 2 years of development.
Whatever dude. You're definitely one of the spoiled ones.
We'll have to see how Intel managed with Arrow Lake. Since Raptor Lake is really just a tweak of Alder Lake and not even a new microarchitecture, they've effectively had *3* years and 2 major nodes since their last real update to the desktop platform!
Boost frequency limits stayed the same, except dropping 100 MHz on the 700X and 600X tiers. However, if the CPUs are better at maintaining boost, then we could see a *practical* increase in frequency!
You can't say it runs at lower frequencies until it's actually *tested*.
> looks like total performance stays about the same.
You're either very bad at math or a shameless troll.
"And according to AMD's product pages posted since the keynote, the Ryzen 9000 family will top out at JEDEC DDR5-5600 speeds for in-warranty configurations."
This sentence bothers me. Specifically, why create motherboards that support EXPO memory profiles if you don't plan on honoring the warranty for those motherboards if you use EXPO memory profiles?
We don't yet know enough to say, however I'm not hopeful because their slide specifically said "2x AI and AVX-512 throughput", rather than a broader statement that you'd expect to see if it also applied to 256-bit operands.
5900XT and 5800XT on AMD so said pricing is sufficient to push Vermeer channel holdings down in price at so said $359 and $249 now pulled by AMD in the moment. The channel might not have been happy with that regulating price move on how much R5K there is too clear from the channel. R5K channel available is up + 68% since March 9 when R5K was 68% of all R7K and today 98% of R7K available.
R7K desktop since March 9 channel supply volume available + 18%. R9K will minimally dribble out allowing R7K and R5K to clear? R9K might have to be priced up on specific SKUs to accomplish the same dribbling out objective allowing AMD back generation to clear?
Notably 3600 gains in the channel + 94% in the prior month. 3600X came back to secondary resale + 35%. 3700X is up + 15.8% that's all trade-in.
AMD might have to adjust R9K desktop top SKU and R5K desktop regulating SKUs not to interfere with the channel's ability to liquidate especially Vermeer from channel inventory holdings plus R7K SKUs that will follow in a first in first out channel sales system.
In summary, there is plenty of Intel and AMD product in the channel. The PC market remains in a downward deflationary price spiral until at least q1 2025 aimed to clear existing inventories for channel financial reclaim to buy next generation.
Subsequently there's this inventory bridge to traverse to Intel and AMD next generation products and through the summer into q4 it's never been a better time to buy a PC. I don't think desktop and mobile prices will be as low as they are heading into year end and for a long time following.
Base clocks are lower & boost clocks not changed because of wider FP/Vectors (from 256b -> 512b) & other widening. Workloads without AMX512 will allow the cores to clock higher. AMD are being more conservative to promote energy efficiency & reliability over the Intel default extreme overclocks. AMD IPC improvements may increase after additional software optimisations. It's disappointing to see limited improvement on release with existing software. Lack of advanced 3nm fab capacity ready in time (available maybe next year). Current 3nm aims at Apple & lower power mobile, desktop 3nm needs more time to optimise & build capacity. The other possibility is that PBO & user tweaking will allow higher clock boosts (over stock speed) than previous models. I would really love to be able to set voltage & frequency limits for number of cores etc so safety can be maintained. I don't want constant >1.3v but I also don't want sudden 1.45v extremes (for current & equivalent values for next gen). I can limit PPT but voltages still go extreme.
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.
Facebook
Twitter
RSS
62 Comments
Back to Article
Dolda2000 - Sunday, June 2, 2024 - link
They're launching in just over a month, but no pricing info? And no technical details? Is there another event before launch where they're going to announce this, or is it just going to be put into a press release sometime?Ryan Smith - Sunday, June 2, 2024 - link
Note that AMD hasn't used a trade show for a major product launch in years. Any trade show announcements have always been followed by an AMD event for the actual launch.ballsystemlord - Monday, June 3, 2024 - link
Thanks for clarifying. I thought the lack of an MSRP was odd as well.haukionkannel - Friday, June 14, 2024 - link
Most likely rather close 7000 series msrp + inflation.The 7000. Series is so much down from their starting prices that there is Room in original msrp price points, maybe little above those. That will make the 7000 series the best bang for money option and 9000 series the new flagship tier.
nandnandnand - Sunday, June 2, 2024 - link
I don't see anywhere else reporting the base clocks, which are 200-800 MHz lower than their Zen 4 counterparts. Didn't see in slide deck. Where are these coming from?bubblyboo - Monday, June 3, 2024 - link
https://ir.amd.com/news-events/press-releases/deta...mode_13h - Monday, June 3, 2024 - link
I wouldn't worry too much about the drop in base clocks, since that's probably specified for AVX-512 "power virus" workloads. The other reason for the drop should be the reduction in TDP on all models except the 950X tier.What's got me a little more worried is the lack of improvement in boost clocks, other than at the 700X and 600X tier, which each increased only 100 MHz. However, if these new CPUs actually have an easier time achieving and sustaining their boost clocks, perhaps it'll save AMD from some of the criticism it's gotten in the past for low-boosting CPUs.
nandnandnand - Monday, June 3, 2024 - link
They moved the CCD from N5 to an N4-class node, maybe N4P. There could be an architectural reason why they can't scale clocks up from Zen 4, or there is less benefit to hitting those higher clocks. So they get to focus on efficiency this time around. If it's doing 1.16x the work at 0.78x the power (N4P benefit), I guess it's a whopping 49% more efficient.LuxZg - Monday, June 3, 2024 - link
And I applaud this move. They obviously felt less need to chase Intel as they believe to have product that is superior enough without exploding on the TDP. TDP and heat were the least liked aspect of the previous series (well, maybe 2nd behind a huge drop in value proposition). I hope 65W Zen5 parts will bring back the expected balance in power/performance without having to undervolt or tweak power in any other way. Sadly I don't expect the same for the value. AMD is no longer the underdog and it shows on the price tag.peevee - Tuesday, June 4, 2024 - link
Giving Ryzen 7 only 65W is just plain dumb. Especially if it is X and does not come with a cooler.mode_13h - Friday, June 7, 2024 - link
The perf/W scaling for the single-CCD Zen 4 CPUs is horrible, above 65 W TDP (88 W PPT). Maybe it's even worse, with Zen 5. Would you *really* want to burn more power for no benefit?mode_13h - Friday, June 7, 2024 - link
Besides, if there *is* any benefit, you'll probably be able to use PBO to unlock it.Zoolook13 - Saturday, June 15, 2024 - link
That it doesn't come with a cooler doesn't mean it should be used without a cooler, that would be seriously stupid.Samus - Tuesday, June 18, 2024 - link
Zen architecture historically scales down incredibly well with TDP. I suspect the 65w parts will be incredibly potent.Silver5urfer - Tuesday, June 4, 2024 - link
AMD processors use dual issue AVX512 and unlike Intel they do not need AVX offsets. There's no power virus aspect for AMD processors in this category from what I can recall.Yes there's no boost clock improvement, but also regression in base clocks, perhaps the Zen 5 is just an overhaul of Zen 4 there's no new arch, its literally like Intel Core which they are using since 2006 and AMDs Zen is from 2017, iterative improvements rather than a clean slate design.
AMD processors had the higher base clocks unfortunate that they are not anymore. Perhaps they are targeting the stupid TDP nonsense. Shame really. Intel on the other hand goes overboard the power consumption and pushes their silicon to absolute limit at 400W and 300A and creates a total degradation and instability, AMD won't even try to push 50W more and get a better performance. Don't know if it's the TSMC node limit or silicon limit or the uArch limitation, as since Zen 4 these processors scale in performance by 94C Temp target.
Chipset side is also nothing big, its just a USB4 and Wifi M.2 upgrade, the base is same ASM/AMD dual chipset. Maybe Zen 6 ? Time will tell, since the SMT performance vs a 14900K is just nothing esp considering the fact how AMD always lead in SMT vs Intel.
GeoffreyA - Tuesday, June 4, 2024 - link
It's possible the dropping of clock speeds went hand in hand with widening the core and increasing IPC. Perhaps the pipeline was shortened, though that seems doubtful. I think there won't be a drastic new architecture because the industry has seen, through trial and error, that high-performance CPU design falls along similar lines, whether it's x86 or ARM. Departures, like NetBurst and Bulldozer, have been painful.Reflex - Thursday, June 6, 2024 - link
It's really unclear to me why people care about base clocks. So far as I am concerned the lower the better. my CPU spends most of it's time sitting at 1-20% usage, only things like gaming and video/photo editing really spike it much. I'd rather use as little power as possible for day to day tasks so long as it can easily ramp up as soon as a workload demands it.mode_13h - Saturday, June 8, 2024 - link
If you don't think about it too hard, the drop in base clocks looks like a bad sign.BTW, it's not like Intel didn't lower base clocks, going from Kaby Lake to Comet Lake!
Zoolook13 - Sunday, June 16, 2024 - link
It's simple math, with an ipc increase of 15% more work gets done per cycle and if the perf/W improvement is less than the ipc increase, the base clock needs to go down to keep within the TDP.Commenter1 - Monday, June 3, 2024 - link
I'm hoping we see some more Epyc 4004 motherboards at Computex.mATX is limiting the number of PCI slots too much for me.
mode_13h - Monday, June 3, 2024 - link
Ugh, you're right. It seems all of ASRockRack's, Supermicro's, and Tyan's AM5 motherboards are no bigger than micro-ATX. I checked an ASUS Prime Pro ATX motherboard that supports ECC memory, but it doesn't include EPYC 4xx4 in the list of supported CPUs.schujj07 - Monday, June 3, 2024 - link
If you are PCIe slot limited on the 4004 series you would probably be a good candidate for the 8004 series instead. Otherwise from a cost point looking at the 7003 series is another choice.mode_13h - Friday, June 7, 2024 - link
The AM5 CPUs and higher-end chipsets have enough lanes to easily populate a regular ATX motherboard. Sometimes, you just need one more x4 slot, in which case EPYC or TR would be way overkill.ballsystemlord - Monday, June 3, 2024 - link
Intel: Cuts AVX-512 support from their CPUs.AMD: Improves AVX-512 performance on their CPUs.
I'm laughing. This is too good!
mode_13h - Monday, June 3, 2024 - link
You missed the part where Intel flipped over the board by replacing AVX-512 with AVX10.schujj07 - Monday, June 3, 2024 - link
AMD could be using AVX10/512."To begin, let’s break down AVX10.N/M: AVX10 is the new “foundational” SIMD/vector instruction set for x86_64. The “.N” denotes the version of AVX10 as a version modifier, allowing incremental updates. It is important to note, if you support “AVX10.N+3,” you must support all of AVX10.N, N+1 and N+2. In simpler words, users are guaranteed supersets of previous instruction sets.
What does the “/M” mean? It’s a reference to vector register implementation size of a given AVX10.N version. Specifically, it may be 512-bit, 256-bit, or the topic of this article, 128-bit wide."
https://chipsandcheese.com/2023/10/11/avx10-128-is...
Therefore it could be very easy for it to be AVX10 as that includes AVX512 if need be.
mode_13h - Monday, June 3, 2024 - link
> AMD could be using AVX10/512.LOL. Don't matter for client workloads, because Intel has telegraphed that they'll only be using 256-bit in client processors. Therefore, software developers using AVX10/512 (which Intel pejoratively terms "legacy") will be only those targeting servers that could run AVX-512, also.
Now, when AMD does support AVX10, they could revert C-series cores to using a 256-bit wide implementation of AVX-512, just like Zen 4 did, and thereby maintain backwards compatibility without being at a density disadvantage on AVX10/256 relative to Intel.
schujj07 - Tuesday, June 4, 2024 - link
"LOL. Don't matter for client workloads, because Intel has telegraphed that they'll only be using 256-bit in client processors."AVX512 is becoming more useful and common on the server side. Since the Zen cores is virtually identical between server/desktop CPUs, it makes sense to keep it on both.
mode_13h - Friday, June 7, 2024 - link
> Since the Zen cores is virtually identical between server/desktop CPUs,> it makes sense to keep it on both.
I'm not saying AMD should remove it. I'm saying that Intel found a way to nullify that advantage by creating FUD for people like game devs around the future of AVX-512 so they don't go and write a bunch of code that runs faster on AMD client processors.
Eventually, Intel will role out AVX10/256 and that's *just* incompatible enough not to work on AMD CPUs without hardware modifications. And if AMD does implement support for it, depending on how they added AVX-512, half of their vector pipelines might be sitting idle, meanwhile Intel's will be fully utilized.
ballsystemlord - Saturday, June 8, 2024 - link
Oh, no, Intel would never do something like that! (sarcasm)systemBuilder33 - Monday, June 3, 2024 - link
I cant see anything new and useful? What a meh release!systemBuilder33 - Monday, June 3, 2024 - link
Nothing new in x870 that is ...mode_13h - Monday, June 3, 2024 - link
If I'm not mistaken, there's no cache increase at the L2 or L3 levels. Maybe they increased the amount of L1D, which has been 32 kiB since Zen 1.In Zen 4, they L2 doubled to 1 MiB per core. In Zen 3, there was no improvement (not counting X3D). Zen 2 doubled the amount of L3 cache per core.
GeoffreyA - Monday, June 3, 2024 - link
I'm glad to see the increase in encoding performance. With all the hype behind Zen 5, I think people were expecting more across the board, but an average of 16% is not half bad.haplo602 - Monday, June 3, 2024 - link
Are you sure about the RDNA2 GPU arch ?peevee - Tuesday, June 4, 2024 - link
Looks like they use literally the same IOD chip.Dante Verizon - Monday, June 3, 2024 - link
Meh, they didn't use the 4090, the improvement in gaming should be more extensive. AMD wants to leave a surprise for the reviews?Blastdoor - Monday, June 3, 2024 - link
there's some weird geekbench cherry picking going on there. I look forward to seeing independent reviews. My first reaction, though, is that Zen 5 is not the revolution some commenters were predicting and that AMD faces a possible Core2Duo-esque scenario with Arrow Lake on Intel's A20 process.nandnandnand - Monday, June 3, 2024 - link
Luckily for AMD, they will have free reign for a few months until Arrow Lake launches, and they have X3D in their pocket. Arrow Lake is also launching on a new socket.Blastdoor - Monday, June 3, 2024 - link
Yes, but a few months out of a two year product cycle isn’t much — especially since AMD might never regain the process lead they’ve enjoyed these last several years.schujj07 - Monday, June 3, 2024 - link
Arrow Lake isn't expected until Q4 2024 and that might be little more than a paper launch with volume sometime in 2025.haukionkannel - Monday, June 3, 2024 - link
16% ipc increase is rather good. But if you allready have 7000 series… these are not for you! Unles you just want to have the fastest you can get.Terry_Craig - Monday, June 3, 2024 - link
I think the "AI people" will lead Zen5 to shortage. AVX512 enables 2x improvements in AI.schujj07 - Monday, June 3, 2024 - link
I'm hoping that isn't the case as I'm planning on finally upgrading my 2013 4770k desktop.mode_13h - Monday, June 3, 2024 - link
Anyone who's really serious about AI won't be doing it on the CPU.peevee - Tuesday, June 4, 2024 - link
"AVX512 enables 2x improvements in AI"Not if it is done mostly on GPUs or AI accelerators with VASTLY superior memory bandwidths.
Blastdoor - Monday, June 3, 2024 - link
It’s good if it’s real. I think it’s really weird that they used two geekbench subscores rather than the full benchmark.But even if it’s real, it’s not the 20-40 percent AMD fans were expecting (though now they’ll say they were never expecting that).
mode_13h - Monday, June 3, 2024 - link
Expecting 20% to 40% IPC increase would be pretty nuts, IMO. It's not like AMD is starting from a low baseline, like they did with Zen 1.peevee - Tuesday, June 4, 2024 - link
"Expecting 20% to 40% IPC increase would be pretty nuts"Who said anything about specifically IPC?
Expecting 20% total performance increase is VERY VERY low expectations for 2 years of development. And given that frequencies has decreased, looks like total performance stays about the same.
mode_13h - Friday, June 7, 2024 - link
> Expecting 20% total performance increase is VERY VERY low expectations for 2 years of development.Whatever dude. You're definitely one of the spoiled ones.
We'll have to see how Intel managed with Arrow Lake. Since Raptor Lake is really just a tweak of Alder Lake and not even a new microarchitecture, they've effectively had *3* years and 2 major nodes since their last real update to the desktop platform!
mode_13h - Friday, June 7, 2024 - link
> given that frequencies has decreased,Boost frequency limits stayed the same, except dropping 100 MHz on the 700X and 600X tiers. However, if the CPUs are better at maintaining boost, then we could see a *practical* increase in frequency!
You can't say it runs at lower frequencies until it's actually *tested*.
> looks like total performance stays about the same.
You're either very bad at math or a shameless troll.
James5mith - Tuesday, June 4, 2024 - link
"And according to AMD's product pages posted since the keynote, the Ryzen 9000 family will top out at JEDEC DDR5-5600 speeds for in-warranty configurations."This sentence bothers me. Specifically, why create motherboards that support EXPO memory profiles if you don't plan on honoring the warranty for those motherboards if you use EXPO memory profiles?
peevee - Tuesday, June 4, 2024 - link
So all the remains of Moore law are officially dead even for AMD, no improvements in 2 years.GeoffreyA - Wednesday, June 5, 2024 - link
The computer industry needs a breakthrough anyhow, beyond semiconductor and electrons.mode_13h - Friday, June 7, 2024 - link
peevee is just trolling. N4P is a quantifiably better node than N5 was.mode_13h - Friday, June 7, 2024 - link
TSMC claims 11% more performance, 22% higher power efficiency, and 6% higher transistor density over the N5 node.Some of that should translate into higher *practical* frequencies.
peevee - Tuesday, June 4, 2024 - link
Does expanding the vector ALU also means ~doubling of AVX256 code performance/throughput?GeoffreyA - Wednesday, June 5, 2024 - link
It should, if the number of floating-point units remain the same or increase.mode_13h - Friday, June 7, 2024 - link
We don't yet know enough to say, however I'm not hopeful because their slide specifically said "2x AI and AVX-512 throughput", rather than a broader statement that you'd expect to see if it also applied to 256-bit operands.Bruzzone - Wednesday, June 5, 2024 - link
5900XT and 5800XT on AMD so said pricing is sufficient to push Vermeer channel holdings down in price at so said $359 and $249 now pulled by AMD in the moment. The channel might not have been happy with that regulating price move on how much R5K there is too clear from the channel. R5K channel available is up + 68% since March 9 when R5K was 68% of all R7K and today 98% of R7K available.R7K desktop since March 9 channel supply volume available + 18%. R9K will minimally dribble out allowing R7K and R5K to clear? R9K might have to be priced up on specific SKUs to accomplish the same dribbling out objective allowing AMD back generation to clear?
Notably 3600 gains in the channel + 94% in the prior month.
3600X came back to secondary resale + 35%.
3700X is up + 15.8% that's all trade-in.
AMD might have to adjust R9K desktop top SKU and R5K desktop regulating SKUs not to interfere with the channel's ability to liquidate especially Vermeer from channel inventory holdings plus R7K SKUs that will follow in a first in first out channel sales system.
In summary, there is plenty of Intel and AMD product in the channel. The PC market remains in a downward deflationary price spiral until at least q1 2025 aimed to clear existing inventories for channel financial reclaim to buy next generation.
Subsequently there's this inventory bridge to traverse to Intel and AMD next generation products and through the summer into q4 it's never been a better time to buy a PC. I don't think desktop and mobile prices will be as low as they are heading into year end and for a long time following.
mb
tygrus - Thursday, June 6, 2024 - link
Base clocks are lower & boost clocks not changed because of wider FP/Vectors (from 256b -> 512b) & other widening. Workloads without AMX512 will allow the cores to clock higher. AMD are being more conservative to promote energy efficiency & reliability over the Intel default extreme overclocks.AMD IPC improvements may increase after additional software optimisations. It's disappointing to see limited improvement on release with existing software. Lack of advanced 3nm fab capacity ready in time (available maybe next year). Current 3nm aims at Apple & lower power mobile, desktop 3nm needs more time to optimise & build capacity.
The other possibility is that PBO & user tweaking will allow higher clock boosts (over stock speed) than previous models.
I would really love to be able to set voltage & frequency limits for number of cores etc so safety can be maintained. I don't want constant >1.3v but I also don't want sudden 1.45v extremes (for current & equivalent values for next gen). I can limit PPT but voltages still go extreme.
tygrus - Thursday, June 6, 2024 - link
Stupid web browser loses form contents (the comment)...