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!

 

[inf64]

Just to give more context to gaming performance, here is how things stand right now (on average @ 720P; Computerbase data ):





For 7000 series ( 8C/16T and up) to beat Alder Lake KS (~5.5Ghz boost part), they need to perform roughly 10% better than 5950X/5900X/5800X.
I find it very hard to believe that Ryzen 7000, with IPC and massive all core boost clocks cannot achieve ~10% better gaming performance on average Vs Ryzen 5000 series. It should easily beat 5000 series by at least 15% in games, basically beating the 5800X3D as well. This all goes for vanilla version of Zen4 without the Vcache.

 

[biostud]

So they will probably release RDNA3 and 7800X3D around the same time for the ultimate gaming experience.

 

[biostud]

Where the 3D cache improved some games massively and some games saw no benefit over the regular zen3, the zen4 will probably see a more even distribution of performance increase over zen3, and therefore in some games loose to 5800X3D.

Originally I had planned to get a regular zen4, but if they launch the 3D cach version this fall I'll wait for it.

 

[Racan]

If you’re a gamer mainly and not strapped for cash why would you even buy Zen 4 chips without 3D cache?

 

[RnR_au]

The only key question here will be: Why release Raphael-X so soon if vanilla Raphael can take on RPL-S, especially, on its own?

Because AMD wants to sell silicon at a premium? Take normal cpu. Add slivers of cheap silicon, sell for +$150. Cha-ching!

There will be plenty of buyers willing to pony up to get an awesome gaming cpu that don't require crazy watts.

If you’re a gamer mainly and not strapped for cash why would you even buy Zen 4 chips without 3D cache?

Because some just run Fortnite at 60hz... like me

 

[tamz_msc]

Just to give more context to gaming performance, here is how things stand right now (on average @ 720P; Computerbase data ):


View attachment 65347


For 7000 series ( 8C/16T and up) to beat Alder Lake KS (~5.5Ghz boost part), they need to perform roughly 10% better than 5950X/5900X/5800X.
I find it very hard to believe that Ryzen 7000, with IPC and massive all core boost clocks cannot achieve ~10% better gaming performance on average Vs Ryzen 5000 series. It should easily beat 5000 series by at least 15% in games, basically beating the 5800X3D as well. This all goes for vanilla version of Zen4 without the Vcache.

That graph is showing how things stand with official memory speeds and JEDEC memory timings. Computerbase did a comparison with max. OC memory on both platforms, including both DDR4 ad DDR5, and it shows a clear lead for Alder Lake with DDR5:

Ryzen 7 5800X3D vs. Core i7-12700K & i9-12900K im Benchmark: Testsysteme, Testergebnisse und Fazit

RAM-OC-Benchmarks: Testsysteme, Testergebnisse und Fazit / Die drei Testsysteme und ihre Einstellungen / Benchmarks: Übersicht und Analyse

www.computerbase.de

 

[Karnak]

That graph is showing how things stand with official memory speeds and JEDEC memory timings. Computerbase did a comparison with max. OC memory on both platforms, including both DDR4 ad DDR5, and it shows a clear lead for Alder Lake with DDR5:

It shows flawed results in CP2077 (CPU bug with Ryzen CPUs if you dont fix it manually) and SotTR. There are literally no other Reviews out there where the difference is >50% in these titles.

And btw. this is a user review from a forum member, not from CB itself.

 

[Hans de Vries]

From Hassan Mujtaba. Might be correct if his source is from a Mobo manufacturer.

 

[inf64]

Here is techspot's 5800X3D running DDR 3800 CL16 vs 12900K running DDR5 6400 CL32, tested in 40(!) games with 3090ti:

Ryzen 7 5800X3D vs. Core i9-12900K in 40 Games

A massive gaming benchmark comparison between the Ryzen 7 5800X3D and Core i9-12900K, pitting the two head to head across 40 games at 1080p, 1440p and 4K,...

www.techspot.com

5800X3D is still faster, so I'm not sure how computerbase got their results.

 

[tamz_msc]

It shows flawed results in CP2077 (CPU bug with Ryzen CPUs if you dont fix it manually) and SotTR. There are literally no other Reviews out there where the difference is >50% in these titles.

And btw. this is a user review from a forum member, not from CB itself.

Nothing you said invalidates the data itself. Besides there are other reviews out there with fewer game samples and unless the reviewer tests a large swathe of games like HWUB does, the 12900K is usually ahead of the 5800X3D.

 

[Karnak]

Nothing you said invalidates the data itself.

It "invalidates" especially the CP and SotTR P1 results where the 12900K is >50% ahead which affects the overall results.

It's not about "ahead" itself rather than your "clear lead". Anything single digits is close enough to me which would be the case with those two games performing how they should.

 

[Zucker2k]

Because AMD wants to sell silicon at a premium? Take normal cpu. Add slivers of cheap silicon, sell for +$150. Cha-ching!

There will be plenty of buyers willing to pony up to get an awesome gaming cpu that don't require crazy watts.

Because some just run Fortnite at 60hz... like me

The trend, including that of AMD with both the XT and X3D releases, has been to maintain superiority with a flagship product until the next upgrade cycle, or whenever the flagship was toppled by the competition.
I'm saying that AMD's plan to release a Zen 4 X3D right after releasing the vanilla version, may actually be indicative of the fact that the vanilla version is not going to be the dominant chip some people think it would be. This, of course, does not rule out your suggestion. We will see.

 

[tamz_msc]

It "invalidates" especially the CP and SotTR P1 results where the 12900K is >50% ahead which affects the overall results.

It's not about "ahead" itself rather than your "clear lead". Anything single digits is close enough to me which would be the case with those two games performing how they should.

Just because you personally don't like the data, it doesn't mean that the data is invalid. Besides, like I said before, there are other data which show the same thing, i.e. 12900K + OC DDR5 is faster by a meaningful margin than 5800X3D with OC DDR4.

See this review for example, at 21:22

 

[Vope45]

Just because you personally don't like the data, it doesn't mean that the data is invalid. Besides, like I said before, there are other data which show the same thing, i.e. 12900K + OC DDR5 is faster by a meaningful margin than 5800X3D with OC DDR4.

See this review for example, at 21:22

That benchmark is rather weird.

 

[coercitiv]

From Hassan Mujtaba. Might be correct if his source is from a Mobo manufacturer.

If that pricing is the correct ballpark for every tier, then we're missing a piece of the puzzle around $449.

 

[exquisitechar]

Zen 4: Cores, boost frequency L2/L3 cache.
So with PBO an extra 150 MHz to 5.85 GHz?

View attachment 65346

https://twitter.com/x/status/1555082501630361600

So fmax for the 7950X is likely 5.85GHz. Angstronomics was right again, not that >5.5GHz wasn’t obvious based on what AMD showed and said.

 

[Karnak]

If that pricing is the correct ballpark for every tier, then we're missing a piece of the puzzle around $449.

Probably 7800X3D by the end of the year. Don't think we'll see another 8C above the 7700X without V-Cache.

 

[inf64]

Probably 7800X3D by the end of the year. Don't think we'll see another 8C above the 7700X without V-Cache.

Maybe there is a chance they'll introduce a lower clocked 12 core part since the performance gap between 8C and 12C is rather large. Something like 105W 12/24T part with 5.5Ghz ST boost and much lower all core/base clocks.

 

[Vope45]

Zen 4, a month before launch is still a mystery to me:

1/ All the extra die space for 8-10% ipc uplift? Why? Can't be all of that for just avx512 right?

2/ Why do amd focus heavily on frequency now? You can't have high ipc and clock obviously but still, why now? Is it because of tsmc limited capacity for N5?

 

[Thibsie]

Zen 4, a month before launch is still a mystery for me:

1/ All the extra die space for 8-10% ipc uplift? Why? Can't be all for avx512 right?

2/ Why do amd focus on frequency now?

Yeah, they should have done as Intel.
At least Intel bundled a couple hundred more watts in the package. Shame to AMD.

 

[inf64]

Zen 4, a month before launch is still a mystery to me:

1/ All the extra die space for 8-10% ipc uplift? Why? Can't be all of that for just avx512 right?

2/ Why do amd focus heavily on frequency now? You can't have high ipc and clock obviously but still, why now? Is it because of tsmc limited capacity for N5?

Does that really matter? What matters is performance and performance/watt. If we get a bigger performance jump going from Zen 3 to Zen 4 than what we had from Zen 2 ->Zen 3, it's a major win for AMD.

 

[Vope45]

Yeah, they should have done as Intel.
At least Intel bundled a couple hundred more watts in the package. Shame to AMD.

No need to be sarcastic, I like amd a lot, just more critical of them than the average forum poster.

Does that really matter? What matters is performance and performance/watt. If we get a bigger performance jump going from Zen 3 to Zen 4 than what we had from Zen 2 ->Zen 3, it's a major win for AMD.

It matters when Amd hit the ipc plateau which will happen eventually or when tsmc lose their edge against intel foundry. It seems to me intel is trying to go lean bulk which takes a long time. To compete in the next 5+ years I think amd need their own fab for sure.

 

[inf64]

It matters when Amd hit the ipc plateau which will happen eventually or when tsmc lose their edge against intel foundry. It seems to me intel is trying to go lean bulk which takes a long time. To compete in the next 5+ years I think amd need their own fab for sure.

I don't see any signs of them hitting IPC wall. They even said that Zen 5 will be the next big thing, a wider core with much greater IPC target (I think the target is >20% vs Zen 4).

Regarding fabs, it's actually detrimental for companies like AMD or Apple to have their own fabs - the cost for continually investing in the new process nodes is overwhelming and can be justified only if the capacities are used to the full (like tsmc is doing now). Intel is in big trouble with their fabs and this will only get more pronounced in the future.

 

[Asterox]

Zen 4, a month before launch is still a mystery to me:

1/ All the extra die space for 8-10% ipc uplift? Why? Can't be all of that for just avx512 right?

2/ Why do amd focus heavily on frequency now? You can't have high ipc and clock obviously but still, why now? Is it because of tsmc limited capacity for N5?

Double L2 Cache size is not small diference, and yes red takes up quite a lot of space.

Hm as far i remember, last AMD CPU(without being too old) with 1mb L2 Cache was for example 45nm Athlon II X2 240.

https://www.cpu-world.com/CPUs/K10/AMD-Athlon%20II%20X2%20240%20-%20ADX240OCK23GQ%20(ADX240OCGQBOX).html

 

[Timorous]

Zen 4, a month before launch is still a mystery to me:

1/ All the extra die space for 8-10% ipc uplift? Why? Can't be all of that for just avx512 right?

2/ Why do amd focus heavily on frequency now? You can't have high ipc and clock obviously but still, why now? Is it because of tsmc limited capacity for N5?

1) They have also added a lot of clock speed and it does not look like it blows the power budget. The extra TDP for the dual CCD parts will allow them to stretch their legs in MT workloads like you can with 5000 series if you exceed the 142W max.

2) Frequency lifts all boats and maybe with the current design there is not as much IPC fruit to pick. This is where Zen 5 comes in to increase the IPC ceiling and provide a few more generations of uplift.