Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[HurleyBird]

Currently, nobody outside of AMD besides close partners knows the performance of Zen 4. No leaked benchmarks, no leaked SKUs, no logs or even a hint of the SKUs are out.
That is for a part that has been sampling for almost a quarter.
Gigabyte Leak was a blemish on an otherwise stellar attempt at securing competitive advantage but still not much can be gleaned from it.

However, Zen 4 is not the big leap though.
Zen 5 is the next big Zen moment for AMD.

Zen 2 wasn't a new architecture either, but if anything ended up more significant than Zen 3.

There have been plenty of early whispers from different sources on Zen 4 (and RDNA3) that all point in the same direction: a very good increase in performance. Of course, there's error bars here, but the largest error bars tend to be when you have rumors in disagreement, or where leakers disagree on other particulars; not when everyone is singing a similar tune. Besides performance, everyone seems to agree on AVX-512 and 1MB L2 cache as well. It's not an environment that calls for a huge amount of skepticism. Neither does the the consensus of remaining at 16-cores while 24 were reportedly considered in the past invite skepticism into performance. That's not a decision you make unless you anticipate your 16-cores parts already give you the win. Can you point to a rumor where Zen 4 is anything less than a 15% IPC increase, or less than a 25% overall improvement? There may be something older, although I can't recollect anything off the top of my head.

Zen 5 will be a bigger change architecturally, obviously, and may also provide more uplift than Zen 4 does (or not), but doesn't get the same kind of process or platform uplifts that will make Zen 4 look better than it actually is. Not to mention, the FP improvements in Zen 4 are almost certainly going to be greater than in Zen 5, in the same way that Zen 2 had a greater FP uplift than Zen 3, and Zen 5 may or may not increase the amount of cache like the Zen 4 does.

Raptor Lake's reportedly meagre big core improvement won't be enough to beat Zen 4, and at the end of the day if there's any lingering doubt then the addition of V-cache will erase it. The larger uncertainty is with the small cores, and if there's enough improvement there to get Intel the win (or at least be competitive) in MT, and if so, whether AMD has a mechanism to release 24-core variants.

 

[DisEnchantment]

Can you point to a rumor where Zen 4 is anything less than a 15% IPC increase, or less than a 25% overall improvement?

Cautious optimism is what I wanted to emphasize, because there are no rumors from the reputable characters. Just to restrain unwarranted drum beating.
There has been no rumors from anyone with a track record (e.g. Execufix) saying it is some quantifiable performance gain.
If I am being open with my thoughts, I don't believe AMD will design some bloated core, it is not their current philosophy. They are going to get 2x more XTors with N5 CCD and I believe they will use them well.

But in general I would be inclined to agree with the rest of your post regarding Zen4/Zen5.

 

[Kedas]

With the separate IOD, I guess that they could fairly easily produce an AM4 version of the next gen chip just by packaging it with the old IOD. But that feels like a losing strategy- DDR5 is only going to get more mature and more readily available over the next year. It's time to let DDR4 go.

From the rumors I understand that the AM5 6nm IO die can handle both DDR4 and DDR5
But that AMD will probably only start high end with only DDR5 support.
Depending on prices they can then decide if they will introduce a low-end DDR4 motherboard also.

Move on and leave DDR4 behind is a good plan for high-end but they are ready to change back to DDR4 without problem if needed, but I'm pretty sure it will not be on AM4. Also not releaseing an AM5 with DDR4 avoids confusion if you don't have to do it..

 

[jpiniero]

Can you point to a rumor where Zen 4 is anything less than a 15% IPC increase, or less than a 25% overall improvement?

There's been some speculation that desktop frequencies might be no better than Zen 3 or even worse due to 5 nm heat density.

Move on and leave DDR4 behind is a good plan for high-end but they are ready to change back to DDR4 without problem if needed, but I'm pretty sure it will not be on AM4. Also not releaseing an AM5 with DDR4 avoids confusion if you don't have to do it..

I dunno, I haven't seen any speculation that Rembrandt supports DDR4.

 

[Kedas]

I dunno, I haven't seen any speculation that Rembrandt supports DDR4.

For the APU not sure if they provided the option but following the same reasoning they could.

It comes down to what AMD decides to do but it will not be a die limitation.
Seems DDR5 and DDR4 hardware are always combo package, just like inlet's version.

https://www.cadence.com/content/dam/cadence-www/global/en_US/documents/tools/ip/design-ip/ddr-phy-ip-for-tsmc-5nm-ip-brochure.pdf

 

[jpiniero]

I think the other "issue" is that if AMD had planned on doing DDR4 for AM5 they would sort of be obligated to support DDR4 for the sockets lifespan.

 

[Hans Gruber]

I think the other "issue" is that if AMD had planned on doing DDR4 for AM5 they would sort of be obligated to support DDR4 for the sockets lifespan.

If they released a Zen 4 sku on the AM4 platform. Meaning one or two CPU models only. Then they could do everything on the AM5 platform. I have Hynix sticks that say on the packaging the memory is good up to at least 4500mhz speeds. That would extend the life of a motherboard as well as providing good options for upgrades once the DDR5 memory speeds increase or the price of DDR5 memory comes down inline with DDR4 and below.

 

[Ajay]

If they released a Zen 4 sku on the AM4 platform. Meaning one or two CPU models only. Then they could do everything on the AM5 platform. I have Hynix sticks that say on the packaging the memory is good up to at least 4500mhz speeds. That would extend the life of a motherboard as well as providing good options for upgrades once the DDR5 memory speeds increase or the price of DDR5 memory comes down inline with DDR4 and below.

That doesn't make any sense. As has been pointed out in this thread before, Zen 3D is the AM4 life extension product. We should be thankful for that, even though it was a business decision - not a 'gift' to consumers. AM4 + Zen 3D will offer 3-5 years of excellent performance for gaming, video editing, streaming, etc. Rembrandt/AM5 will be an expensive upgrade initially. Expensive RAM, expensive motherboards (higher quality signaling needed, new IOD is expected to be more expensive) and CPUs that will likely be unable to satisfy demand and hence have jacked prices. So, unless you are one of our resident bleeding edge DIYers with cash to spare, it's better to wait anyway.

 

[nicalandia]

As has been pointed out in this thread before, Zen 3D is the AM4 life extension product.

This

 

[soresu]

So, what you are saying is that they want to fade into irrelevance?

You seem to have taken my point as a lack of faith in AMD's drive to succeed.

All I did was make a purely logical deduction about the necessary time to recoup R&D costs on Zen4 and likely availability of fab capacity on N3 nodes.

Zen4 is unlikely to land before late Q3 2022 - more likely Q4 to prevent it cannabilising Zen3D sales.

Therefore the likelihood of any Zen5 chip launching in 2023 seems low to me.

I could be wrong of course - especially if Zen5 is in fact using some advanced variant of TSMC's 5nm based processes.

 

[nicalandia]

You seem to have taken my point as a lack of faith in AMD's drive to succeed.

All I did was make a purely logical deduction about the necessary time to recoup R&D costs on Zen4 and likely availability of fab capacity on N3 nodes.

Zen4 is unlikely to land before late Q3 2022 - more likely Q4 to prevent it cannabilising Zen3D sales.

Therefore the likelihood of any Zen5 chip launching in 2023 seems low to me.

I could be wrong of course - especially if Zen5 is in fact using some advanced variant of TSMC's 5nm based processes.

Plus Alder lake and Raptor Lake are not the Core2Duo comeback that some were expecting. Zen3D will even things out until the Zen 4 shows up to dominate. Zen4 will dominate gaming and general tasks

 

[scineram]

I definitely expect Zen 5 in 2023 or early 2024. It will use N5 or equivalent of course. Maybe it’s time to open a Zen 5 topic?

 

[soresu]

Zen4 will dominate gaming and general tasks

Depends on whether the added V cache SKUs come early or late I think.

Eeven without V cache it may well top Raptor Lake, but with it should be much more of a serious competitor if RL is another significant IPC jump from ADL - though clearly Intel has some ways to go in working out the perf/watt kinks at the moment.

Maybe it’s time to open a Zen 5 topic?

IMHO until we have a significant info dump on Zen4 that is still premature.

Or at the very least the anaemic info dump at CES.

 

[Ajay]

I definitely expect Zen 5 in 2023 or early 2024. It will use N5 or equivalent of course. Maybe it’s time to open a Zen 5 topic?

Zen4 is N5. Unless I missed something, Zen5 will be on one of the N3 nodes. I suppose that it could come out on N4 - but that wouldn't be optimal given the targets set for Zen5.

 

[nicalandia]

Depends on whether the added V cache SKUs come early or late I think.

I expect Zen 4 to be without V Cache at launch and still beat Milan-X in gaming and general tasks, Zen3 3D V chache will give a 15% average IPC boost in games, Zen 4 will be a 25% over Zen3 but with better feature set(PCI5, DDR5)

 

[LightningZ71]

It sounds a lot like AMD's cadence is turning into Zen3-> Zen3 w/ cache -> zen4 -> zen4 w/ cache -> Zen5 -> zen5 w/cache etc... it's not necessarily unreasonable to use that approach as it can give longer life to base architectures. I just hope that it works out to keep them competitive.

 

[igor_kavinski]

https://www.electronicdesign.com/te...yc-server-cpus-to-another-level-with-3d-cache

Fascinating that Epyc with 3D cache is helping to make CPU architecture simulations run up to 66% faster. Who knows? Even Intel might be using it in-house!

 

[Joe NYC]

You seem to have taken my point as a lack of faith in AMD's drive to succeed.

All I did was make a purely logical deduction about the necessary time to recoup R&D costs on Zen4 and likely availability of fab capacity on N3 nodes.

Zen4 is unlikely to land before late Q3 2022 - more likely Q4 to prevent it cannabilising Zen3D sales.

Therefore the likelihood of any Zen5 chip launching in 2023 seems low to me.

I could be wrong of course - especially if Zen5 is in fact using some advanced variant of TSMC's 5nm based processes.

I kind of agree, but for different reasons:

AMD priority is:
1. Server
2. Notebook
3. Desktop

So Server grabbed all of the early Zen 3D capacity, desktop will end up with something when AMD has something left over.

The same will likely happen with Zen 4. Genoa will get all of the early capacity, Raphael will get some leftovers in H2 2022.

As far as Zen 5, if it is 3nm product, then it will most likely be 2024 product, just because of the status of TSMC N3. Which, from the latest reports appears to be Q1 2023 for the first wafers out, for Apple and Intel, and AMD likely getting its first wafers late 2023 / early 2024.

 

[Joe NYC]

Anyway, I am not sure HEDT SPR will be launched right off the bat in 2022.
But if they are then that is good, maybe we have a chance to see 96 Core 100Billion+ Xtor HEDT CPUs as a response at some point if not immediately.
But probably 8K a pop at least

From various sources, it seems SPR may be delayed further.

HEDT? When Intel is bleeding market share and can't ship SPR to server customers, HEDT would seem like vanity.

 

[Joe NYC]

Well I did whilst stating an extremely conservative estimate. I think it's more likely we'll see Genoa launch in Q3, but again Q4 is very much a conservative estimate assuming it'll be 2Q after initial shipments.

Ultimately that's based off of Milan, even though Milan could have been launched far earlier if AMD wanted to (I remember STH Patrick confirming this in an article a bit ahead of launch - that AMD could have launched Milan much closer to when Vermeer dropped and the platform would have been as ready as Rome was on it's own launch, but they decided against it).

Ian, from Anadtech commented in his video that AMD is publicly announcing server products some 6 months after shipments started to hyperscalers.

So Milan shipped from early Q4 2020, but was announced late Q1 2021.

Same with Milan X. Shipping since early Q4 2021, but official announcement, with SKUs and all may not come until late Q1 2022.

OTOH, Intel is scrambling, and the announcement will likely come as soon as first production silicon makes it out of the fab. (according to Ian).

 

[MadRat]

It seems like AMD always aims at market trends. With Apple putting memory into their CPU package to win benchmarks it is unimaginable that Intel or AMD won't take it up a notch to one-up Apple. We already are seeing Apple had to cherry-pick benchmarks that matched their strengths, but when doing general tasks they score all over. If AMD has their fabric technology, what stops them from embedding high performance memory chips directly onto their CPU packaging using the lowest possible latency?

With the move towards hardware feeding the OS their finer details, the low latency memory could be fine-tuned to benefit tasks that don't fit into cache. For server-based chips this might make a lot of sense to pare down parts necessary on entry level servers. Plus you get to cherry-pick memory timings. Isn't that what their fabric was more or less promised to do? And if you're worried about volatility of that memory being some kind of drawback compared to Apple's, spec in a battery to keep it from flushing if you kill power. Servers at these prices shouldn't be running without industrial grade UPS backups IMHO. Having that memory on the packaging also dissuades competitors from playing benchmark games using slow memory settings for your products.

 

[igor_kavinski]

I like the embedded memory idea but unlike Apple, I hope Intel or AMD will also offer the ability to expand the on-chip memory with additional memory slots so people are not constrained by just 8/16/32/64 GB SKUs. If Apple offered a 128GB MBP for $4000, I would have bought it instantly. But now with DDR5, I'm thinking it would be better to have 256GB. More RAM can never hurt. Heck, I was thinking about getting a used server with 1TB RAM but then it crossed my mind that the server chassis is going to be extremely heavy. I hate hardware that I have trouble lifting. I gave away my Core 2 Quad in return for 50 Blurays coz it was in a Gigabyte case weighing 10KG. I hated it!

 

[Hans Gruber]

That doesn't make any sense. As has been pointed out in this thread before, Zen 3D is the AM4 life extension product. We should be thankful for that, even though it was a business decision - not a 'gift' to consumers. AM4 + Zen 3D will offer 3-5 years of excellent performance for gaming, video editing, streaming, etc. Rembrandt/AM5 will be an expensive upgrade initially. Expensive RAM, expensive motherboards (higher quality signaling needed, new IOD is expected to be more expensive) and CPUs that will likely be unable to satisfy demand and hence have jacked prices. So, unless you are one of our resident bleeding edge DIYers with cash to spare, it's better to wait anyway.

Zen3D is the holy redacted response to Intel. Look at my old posts. I warned everybody this is intel we are talking about. Remember when AMD had the upper hand because they had better value motherboards and a price/performance price point on CPU's. The memory market has always been a cartel of price fixing. The memory manufacturers were baffled when they lowered the price of memory and consumers bought double the amount of memory they would have bought. It's as if they do not understand supply and demand. When prices are low people buy in bulk.

AMD would be well served with a 20-24 core AM4 Ryzen 3d CPU. Alder Lake wiped the floor with everything except the 5950x. That held it's own against alder lake. I can't hate Intel. Since Core2 every intel processor I ever had lasted 8 years before becoming obsolete.

I will add one more note. AMD promised a core war with Intel. By Zen 3, they should have had at least a 24 core if not a 32 core on the Ryzen line. Go back to the original Zen introduction in 2017. AMD's excuse has been there is threadripper for more cores. But that is more of a server chip than a gaming CPU.

We have a zero tolerance policy for profanity in the tech sub-forums.
Don't do it again.

Iron Woode
Super Moderator

 

[DrMrLordX]

@Hans Gruber

That's overstating things a bit. AMD is not in any trouble. Their primary focus is on servers/workstations where Milan is already top dog and where Genoa is taking control. It really won't rattle them much that Alder Lake manages to win some benchmarks against lower core-count Vermeer products. Zen3D is not a "holy" anything response. It's a way to extend AM4's shelf life a little so they have flexibility on when to release Raphael. Plus with Milan-X in production, why not make use of some of the cast-off dice for a consumer Vermeer-X product?

As far as core wars go, software just won't scale infinitely. Adding more and more cores will only help a smaller and smaller segment of the buying population. There probably isn't any need for more than 12c on the consumer desktop right now, with 16c being a rather small-but-lucrative niche where AMD can drive higher ASPs by forcing performance users who want the best bins and best boost clocks to buy more cores than the normally need. 24c and 32c AM4 products would be, frankly, idiotic. We may eventually see higher core counts with Zen5, but Raphael will probably not go past 16c. Remember Intel can not now or in the forseeable future produce desktop CPUs with more than 8 P cores. The E cores are frankly irrelevant.

 

[soresu]

24c and 32c AM4 products would be, frankly, idiotic. We may eventually see higher core counts with Zen5, but Raphael will probably not go past 16c.

Depends - so far they have kept AM4 Ryzen to max exactly 1/4 the core count of Threadripper, 8/32 for Zen1/+ and 16/64 for Zen2 (3 still TBA).

Bare in mind that the 16 core R9 3950X SKU took 5 1/2 months longer to release and came out in step with Zen2 Threadripper.

Likewise I would expect a 24C maximum SKU for Zen4 AM5 in lockstep with the Zen4 Threadripper scaling to 96C, which could take even longer than Zen2 TR did if fab capacity is still problematic at the N5 nodes.

Anyone care to guess on whether the V cache density will increase for Zen4 SKUs that use it?