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!

 

[jpiniero]

Since then, TSMC has shown the ability to provide way more capacity than initial projections from AMD. While AMD's financial model aims for 20% CAGR revenue growth, last year AMD started by forecasting 25%, adjusted down to 21% at the height of Covid-19 uncertainty, then up to 31%, then 41%, just to end at 45%, and they started this year at 37%, then adjusted up to 50%, and then further increased that to 60%. I am sure these forecasts are done knowing very well what TSMC can deliver. The rapid growth shows the substantial headroom AMD has had (so far, at least) in their supply agreement with TSMC.

Some of that might be price hikes...

 

[Vattila]

Some of that might be price hikes...

A little maybe. But mostly it is unit growth. Although increasing slowly, gross margin is still well below 50%.

 

[scineram]

A little maybe. But mostly it is unit growth. Although increasing slowly, gross margin is still well below 50%.

View attachment 49522

The gross margin was dragged down significantly by the hot new consoles.

 

[Vattila]

The gross margin was dragged down significantly by the hot new consoles.

I know what you mean — the lower gross margin in the semi-custom segment serving the consoles is counteracting the rising gross margin from the client and server segments — but AMD's overall gross margin has in fact risen every quarter since the introduction of Zen, although very slowly since the console ramp last year. So "held back", not "dragged down", is more correct to say.



AMD Gross Margin 2006-2021 | AMD | MacroTrends

However, the semi-custom segment has low operating expenses, since the clients pay for development and marketing, so a lot of the gross profit falls to the bottom line.

 

[Asterox]

Still an expected rumor, or up to 125W APU-s for AM5 socket.

https://twitter.com/x/status/1433039853525078019

 

[moinmoin]

"...at this point."

 

[eek2121]

Still an expected rumor, or up to 125W APU-s for AM5 socket.

https://twitter.com/x/status/1433039853525078019

The 170W rumor was an interesting one to say the least. I'm certain what such a processor would gain over simply using PBO.

 

[Saylick]

Wasn't the rumor that at some point AMD was going to release 24-core Ryzen parts but will launch only with 16-cores? 125W for 16-cores for now, then when they drop 24-core models, they'd have the 170W TDP?

 

[Thunder 57]

Wasn't the rumor that at some point AMD was going to release 24-core Ryzen parts but will launch only with 16-cores? 125W for 16-cores for now, then when they drop 24-core models, they'd have the 170W TDP?

Why eat into their TR line with a 24 core Ryzen?

 

[Saylick]

Why eat into their TR line with a 24 core Ryzen?

Because maybe it offers a more competitive SKU vs Intel, because TR still has the advantage of more IO and memory channels, and also because TR would then occupy a higher tier of product by having a 96-core option? Why does current 16-core Ryzen parts exist when there's also 16-core TRs?

 

[DrMrLordX]

TR still has the advantage of more IO and memory channels

In the case of TR Pro, a lot more memory channels. If yields are high enough and demand is high enough, though, AMD may just wind up selling most of its CCDs in EPYC and starve all the other parts of their product lineup.

 

[Thunder 57]

Because maybe it offers a more competitive SKU vs Intel, because TR still has the advantage of more IO and memory channels, and also because TR would then occupy a higher tier of product by having a 96-core option? Why does current 16-core Ryzen parts exist when there's also 16-core TRs?

I wonder how well they sell though, unless you really need the extra I/O, which is very much possible. But at that point you are investing a lot into the platform so why not go for a 24 core model instead? Depends on your workload for sure.

So I half assed it and just looked at Newegg, and they don't even offer 16 core TR. Seems they only bother with 24+. Not saying there isn't a market for 16 core TR, just doesn't seem to be very large.

Also, what if AMD plays the "Segmentation Game" a bit more like Intel has done forever, and doesn't offer a 96 core TR? Not to mention AMD seems to care about TR less and less as there is still no Zen 3 ones available. Not sure how it will all play out. At least we get to talk about CPU's again

 

[A///]

The 16c TR is a TR Pro. Those are about $1,100 online. There's a distinct segmentation between TR/TR Pro and regular Ryzen. The issue with TR Zen 3 coming out so late makes me want to say it was due to the pandemic and sky rocketing console APU sales. TR Zen 2 came out months after Ryzen Zen 2, and the 64 core TR Zen 2 came out in January or February of 2020, though it was more or less leaked months in advance much to the chagrin of people who thought it was BS.

There was a report not long ago that stated AMD had reduced the number of RDNA2 GPUs being made to make up for the excess demand of console APUs despite the fact having secured and delivered the ordered amount for Sony and Microsoft through March 2021, at least that was Sony's initial contract. I don't think either of the three companies realized just how bad things would get and how in-demand their products would be.

I'm only just seeing place cards for Xbox Series S and Series X as well as the PS5 digital/Disc at some stores. Costco had them advertised the other day. There was maybe 2 placecards left for the PS5 disc edition.

 

[Gideon]

The 170W rumor was an interesting one to say the least. I'm certain what such a processor would gain over simply using PBO.

It's not really a rumor, they have addet it in the motherboard requirements , so that they can add such a model in the lifecycle of the AM5 socket. It might be 4 years from now.

 

[Abwx]

It's not really a rumor, they have addet it in the motherboard requirements , so that they can add such a model in the lifecycle of the AM5 socket. It might be 4 years from now.

Beefing up MBs power deliveries first benefit is much improved MBs reliability with existing SKUs that range within 65 to 105W nominal, wich is good for reputation, AMD certainly have this in mind more than an eventual 170W SKU.

 

[A///]

Wasn't the power limits on AM4 the reason they couldn't push the Zen chips harder or was that bunk info? Someone mentioned it on an IRC to me a year or two ago, but I never looked into it. AM4 came out long before Ryzen was announced, no?

 

[A///]

I'm awfully curious how Zen 4 will pan out, though. Mainly in regard to platform stability and raw power. I'm not sure if I'll readily dump my 10th gen setup I recently built or how as luck would have it Zen 3 became readily available almost a week after I'd bought my 10th gen build parts. From the Gigabyte hack, Zen 4 will be launching with DDR5 speeds above Intel's for Alderlake. There was that rumor of the IOD being redesigned, but knowing AMD I hope they don't cripple the IF again.

 

[eek2121]

Why eat into their TR line with a 24 core Ryzen?

Because Intel is going to have an RPL-S SKU with 8 big cores and 16 small cores, which AMD will end up having to compete against.

 

[LightningZ71]

That would be an interesting fight:. AMD Zen 4 cores, 24 cores, 48 threads, all performance vs. Intel Cove next in an 8+16 p/e ratio with 32 threads. That's going to be quite interesting to see benchmarked.

 

[DrMrLordX]

Wasn't the power limits on AM4 the reason they couldn't push the Zen chips harder or was that bunk info? Someone mentioned it on an IRC to me a year or two ago, but I never looked into it. AM4 came out long before Ryzen was announced, no?

AM4 first launched with A320 and Bristol Ridge in very limited markets. There were some A320s sold in the German and Japanese markets (and possibly China) that were never really designed for Ryzen and, if I recall correctly, had UEFIs that wouldn't even boot Ryzen 1xxx chips. As for power limits? The socket itself could move impressive amounts of power even on early x370 boards, but VRMs were all over the place (especially on A320 and B350 boards).

 

[lobz]

That would be an interesting fight:. AMD Zen 4 cores, 24 cores, 48 threads, all performance vs. Intel Cove next in an 8+16 p/e ratio with 32 threads. That's going to be quite interesting to see benchmarked.

I'd say that would rather be a slaughterhouse. 16/32 Zen4 vs 8/16+16 RTL on the other hand? Now THAT will be really interesting to see, where the strengths and weaknesses of both designs really lie.

 

[DrMrLordX]

I'd say that would rather be a slaughterhouse. 16/32 Zen4 vs 8/16+16 RTL on the other hand? Now THAT will be really interesting to see, where the strengths and weaknesses of both designs really lie.

Raptor Lake may actually be in an 8P + 32E configuration at the top end. For consumers I don't see that as being much of a win, especially since AMD's 16c chips are still overkill for most desktop users. If AMD really can average +29% IPC on Zen4 while sustaining clocks, throwing out scads of E cores won't do Intel many favors. AMD has also made significant strides in interconnect bandwidth/latency to keep intercore communications between CCDs snappy enough that you don't really notice many problems when running software designed for monolithic CPUs. It's not clear at all how intercore latency will affect Alder Lake or Raptor Lake. Possibly not well at all.

 

[eek2121]

That would be an interesting fight:. AMD Zen 4 cores, 24 cores, 48 threads, all performance vs. Intel Cove next in an 8+16 p/e ratio with 32 threads. That's going to be quite interesting to see benchmarked.

I'd say that would rather be a slaughterhouse. 16/32 Zen4 vs 8/16+16 RTL on the other hand? Now THAT will be really interesting to see, where the strengths and weaknesses of both designs really lie.

An 8+16 part would beat a 5950x unless…

Raptor Lake may actually be in an 8P + 32E configuration at the top end. For consumers I don't see that as being much of a win, especially since AMD's 16c chips are still overkill for most desktop users. If AMD really can average +29% IPC on Zen4 while sustaining clocks, throwing out scads of E cores won't do Intel many favors. AMD has also made significant strides in interconnect bandwidth/latency to keep intercore communications between CCDs snappy enough that you don't really notice many problems when running software designed for monolithic CPUs. It's not clear at all how intercore latency will affect Alder Lake or Raptor Lake. Possibly not well at all.

Such a design would be probably be power limited in all core workloads. Gracemont may be efficient, but I doubt it is 32x as efficient.

 

[lobz]

An 8+16 part would beat a 5950x unless…


Such a design would be probably be power limited in all core workloads. Gracemont may be efficient, but I doubt it is 32x as efficient.

I don't understand you mentioning us here. The 5950 is not based on Zen 4.

 

[DrMrLordX]

An 8+16 part would beat a 5950x unless…

5950X is the least of Intel's worries.

Such a design would be probably be power limited in all core workloads. Gracemont may be efficient, but I doubt it is 32x as efficient.

Alder Lake is already going to be power-limited in all-core workloads. Adding more e cores instead of p cores won't change that. Intel's real problem is that they'll be under pressure to throw more die area at increasing performance without being able to grow beyond 10ESF. And that's why Zen4 is such a threat. AMD doesn't have that limitation.