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!

 

[Geddagod]

Considering how the 7950x absolutely kills the 5950x

Well ye, but the 7950x is still ~10% slower than the 13900k

and double the cache from the 5800x3d

I don't think the extra V-cache on the second chiplet is going to be well used if at all by the other chiplet. It's the same reason why the 5950x is barely any faster than the 5800x barring the extra cores and slightly higher clocks- the L3 on the other chiplet, to my knowledge, isn't used by the first chiplet.
Also didn't AMD claim at one point that there is no gaming benefit from an additional stacked CCD?

Look at Milan-x ????

No one is using or tested Milan-x for gaming.... what does Milan X have to do anything with supposed gains from the 7950x 3D?

 

[Geddagod]

Again, you only speak for some people, not the DC community. The upcoming project (tomorrow) would be killer if both ccds' had 96m cache. See the DC forum for details, but stop speaking for everyone.

Why does it matter? When speaking about performance, we use averages because that's what applies to MOST people. He isn't speaking for everyone or anything lmao, idk what's up with this "white knighting" for the DC community.
Also he already mentioned this to the post you replied too, but I'm also pretty sure this is true, but having an additional CCD with extra cache won't cause a performance boost I'm pretty sure.

 

[Geddagod]

These numbers aren't stock comparisons of 5800X vs. 3D - they are at normalised 4,4Ghz frequency, increasing the relative advantage. The reasoning being that the frequency gap will be much less this time.

The problem here being that the 5800x3d boosts less than 5% less than the 5800x. The average gaming performance isn't going from ~15% on average to 25% from that.

 

[Geddagod]

The question is, as I see it, which types of games see the most uplift. Personally I don’t find it very important that games like CS:GO, and other older games that are really not that bottlenecked don’t see any improvement. On the other hand when a cpu bound modern game see a uplift 30-40% then it is pretty significant, even though the average over several games is only 15%.

Isn't CSGO like ONLY CPU bottlenecked? Like new CPUs still get pretty good gains on CSGO, the 7950x gained nearly 40% over the 5950x, and the 13900k gained like 10%.
Also games released from 2018 and below actually gained more performance from the 5800x3d vs the 5800x compared to games released in 2021 and 2022. Being modern actually seems to hurt the 5800x3d gaming performance relative to the 5800x (techpowerup).

 

[biostud]

Isn't CSGO like ONLY CPU bottlenecked? Like new CPUs still get pretty good gains on CSGO, the 7950x gained nearly 40% over the 5950x, and the 13900k gained like 10%.
Also games released from 2018 and below actually gained more performance from the 5800x3d vs the 5800x compared to games released in 2021 and 2022. Being modern actually seems to hurt the 5800x3d gaming performance relative to the 5800x (techpowerup).

My point was, that some games which runs 200+ fps are not bottlenecked at all, except for benchmarking purposes. So these games are to me irrelevant when deciding which cpu to buy, because all CPUs released within the last 5 years can play them without any problems.

 

[Markfw]

Well ye, but the 7950x is still ~10% slower than the 13900k

I don't think the extra V-cache on the second chiplet is going to be well used if at all by the other chiplet. It's the same reason why the 5950x is barely any faster than the 5800x barring the extra cores and slightly higher clocks- the L3 on the other chiplet, to my knowledge, isn't used by the first chiplet.
Also didn't AMD claim at one point that there is no gaming benefit from an additional stacked CCD?

No one is using or tested Milan-x for gaming.... what does Milan X have to do anything with supposed gains from the 7950x 3D?

and where did I mention gaming ? The world does not revolve around gaming, in fact the opposite. The process we are now running in DC needs affinity set so 8 cores can be on the same chiplet to use the L3 cache, as this really page a difference in speed. And with 2 chiplets, we need extra cache on both.

And NO, the 7950x is not 10% slower than a 13900k, maybe 2%. Show me the 10% (and not on benchmark outlier)

 

[Carfax83]

Being modern actually seems to hurt the 5800x3d gaming performance relative to the 5800x (techpowerup).

I think that's because modern games tend to prefer raw computational power and memory bandwidth. Latency is still important though to be sure, but as games become larger and more detailed, bandwidth is starting to become a bigger deal.

A good example of that are the Spider-Man games. The addition of ray tracing means that the initial bvh setup and continued maintenance relies on both CPU compute and memory bandwidth for performance as the data sets are too large to fit in cache.

 

[nicalandia]

and where did I mention gaming ?

Don't bother with that type of posters, The 7600X is already beating the 13900K in gaming with good RAM Kits. The 7800X3D will just treat the 13900K like a redheaded stepchild.

 

[Geddagod]

and where did I mention gaming ? The world does not revolve around gaming, in fact the opposite. The process we are now running in DC needs affinity set so 8 cores can be on the same chiplet to use the L3 cache, as this really page a difference in speed. And with 2 chiplets, we need extra cache on both.

And NO, the 7950x is not 10% slower than a 13900k, maybe 2%. Show me the 10% (and not on benchmark outlier)

The article is about gaming.
The person you replied to was about gaming.
You talked about gaming perf when you said "Considering how the 7950x absolutely kills the 5950x, and double the cache from the 5800x3d, I would believe anything from this processor", as you said as a reply to the person talking about gaming.

The 10% figure is from the 3DCenter meta review data.

 

[Markfw]

The article is about gaming.
The person you replied to was about gaming.
You talked about gaming perf when you said "Considering how the 7950x absolutely kills the 5950x, and double the cache from the 5800x3d, I would believe anything from this processor", as you said as a reply to the person talking about gaming.

The 10% figure is from the 3DCenter meta review data.

want to link that article ?

 

[moinmoin]

I don't think the extra V-cache on the second chiplet is going to be well used if at all by the other chiplet.

As the L3$ is a victim cache for the given CCX it obviously won't be used by other CCXs. Worst case the content of both CCXs' L3$es eventually end up being duplicates of each other, though I don't expect that to be likely to happen.

 

[Geddagod]

Don't bother with that type of posters, The 7600X is already beating the 13900K in gaming with good RAM Kits. The 7800X3D will just treat the 13900K like a redheaded stepchild.

Doesn't the 13900k have a better mem-controller than the 7600x? Maybe a tuned 7600x can beat a stock 13900k, but does, on average, a tuned 13900k lose to a tuned 7600x? I HIGHLY doubt it.

 

[Geddagod]

want to link that article ?

Launch-Analyse Intel Raptor Lake (Seite 3) | 3DCenter.org

Mittwoch, 26. Oktober 2022 / von Leonidas Die Anzahl der wertbaren Spiele-Benchmarks hat auch mit diesem Launch (positiverweise) zugenommen. Inzwischen ist es bei den meisten Hardware-Testern Standard, Spiele unter tatsächlich

www.3dcenter.org

 

[Carfax83]

Don't bother with that type of posters, The 7600X is already beating the 13900K in gaming with good RAM Kits. The 7800X3D will just treat the 13900K like a redheaded stepchild.

Care to back up this comment?

 

[Markfw]

Launch-Analyse Intel Raptor Lake (Seite 3) | 3DCenter.org

Mittwoch, 26. Oktober 2022 / von Leonidas Die Anzahl der wertbaren Spiele-Benchmarks hat auch mit diesem Launch (positiverweise) zugenommen. Inzwischen ist es bei den meisten Hardware-Testern Standard, Spiele unter tatsächlich

www.3dcenter.org

Again with the gaming crap. You said its faster, not in what and by who. And that's one website. Anadtech found 1.5% different in gaMING.

You are such a troll.

 

[BorisTheBlade82]

Again with the gaming crap. You said its faster, not in what and by who. And that's one website. Anadtech found 1.5% different in gaMING.

You are such a troll.

No, he is not. You are, as mostly, pushing things too far.

 

[nicalandia]

No, he is not. You are, as mostly, pushing things too far.

Please don't side with known unbiased posters.

 

[Geddagod]

Again with the gaming crap. You said its faster, not in what and by who. And that's one website. Anadtech found 1.5% different in gaMING.

You are such a troll.

How am I a troll? You replied to a person posting an article about GAMING, while just generalizing performance. How is it a stretch to assume you were also talking about gaming?
That is actually NOT one website. A meta review, be definition, is a combination of multiple reviewers.
Anandtech is JUST one reviewer. This meta review combines 20? I think reviewers.

 

[Carfax83]

Again with the gaming crap. You said its faster, not in what and by who. And that's one website. Anadtech found 1.5% different in gaMING.

You are such a troll.

The 3D center analysis is a meta-analysis of multiple reviews, including Anandtech. Anandtech also used a 6900XT during their gaming portion which bottlenecks these ultra fast CPUs, so it's near useless for gauging gaming performance for both Zen 4 and Raptor Lake.

Also, having a contrary opinion isn't trolling. He backed up his comment and just because you don't like his opinion he's a troll? That's ridiculous, and just goes to show how Anandtech forums has become an AMD echo chamber over the years since AMD took back market share with Zen 2 and Zen 3.

Thirdly, the V-cache CPUs are primarily marketed towards improving gaming performance in the consumer CPUs, so I don't know why you're so surprised he's talking about gaming.

 

[Geddagod]

Care to back up this comment?

I mean Intel's mem controller on the 13900k supports higher speeds than the one on the 7600x. Which imo is pretty impressive considering how Intel also has raptor lake supporting DDR4 as well, but that's not important.
Edit: 5600 MHz on 13900k vs 5200 Mhz on 7600x

 

[Markfw]

The 3D center analysis is a meta-analysis of multiple reviews, including Anandtech. Anandtech also used a 6900XT during their gaming portion which bottlenecks these ultra fast CPUs, so it's near useless for gauging gaming performance for both Zen 4 and Raptor Lake.

Also, having a contrary opinion isn't trolling. He backed up his comment and just because you don't like his opinion he's a troll? That's ridiculous, and just goes to show how Anandtech forums has become an AMD echo chamber over the years since AMD took back market share with Zen 2 and Zen 3.

Thirdly, the V-cache CPUs are primarily marketed towards improving gaming performance in the consumer CPUs, so I don't know why you're so surprised he's talking about gaming.

Talking about Raptor lake in a Zen 4 thread is not trolling ? And vcache primarily gaming ? Its started in Milan-x, and THEN went to Zen 3. There are a lot more places that cache helps other than gaming, you guys act the world revolves around gaming.

Edit: and the 5800x3d is known as the best gaming chip out there.

 

[Carfax83]

I mean Intel's mem controller on the 13900k supports higher speeds than the one on the 7600x. Which imo is pretty impressive considering how Intel also has raptor lake supporting DDR4 as well, but that's not important.
Edit: 5600 MHz on 13900k vs 5200 Mhz on 7600x

His comment was totally incorrect and that's why he hasn't backed up his statement with benchmarks. There maybe an outlier perhaps where the 7600x beats the 13900K, but overall, the 7600x is not in the same league as the 13900K.....and that's for both gaming and productivity.

 

[Geddagod]

Talking about Raptor lake in a Zen 4 thread is not trolling ? And vcache primarily gaming ? Its started in Milan-x, and THEN went to Zen 3. There are a lot more places that cache helps other than gaming, you guys act the world revolves around gaming.

I talked about Raptor Lake because that's what the article that you were replying too compared the 7950x3D to raptor lake. The tittle is literarily "7950x3D is 30% faster than the 13900k", to Exist 50 said I doubt it will be that high, to which YOU replied about the 7950x3D " I would believe anything from this processor. "
Also you talked about the 7950x3d FIRST and THEN Milan-X. Milan-X was at the end of your comment.

 

[Markfw]

How am I a troll? You replied to a person posting an article about GAMING, while just generalizing performance. How is it a stretch to assume you were also talking about gaming?
That is actually NOT one website. A meta review, be definition, is a combination of multiple reviewers.
Anandtech is JUST one reviewer. This meta review combines 20? I think reviewers.

So if you take 30 crap reviews and 2 good ones. then the average is not worth reading. Its does not matter how many, its how many of the GOOD review sites say what. And for the most part they all agree that Zen4 and Raptor lake trade blows, but Raptor lake does it using WAY more power.

 

[Geddagod]

So if you take 30 crap reviews and 2 good ones. then the average is not worth reading. Its does not matter how many, its how many of the GOOD review sites say what. And for the most part they all agree that Zen4 and Raptor lake trade blows, but Raptor lake does it using WAY more power.

What makes those other reviewers crap?
Btw, the metareviews from 3Dcenter.org are usually supported by large-data set benchmarks from other popular (maybe you consider them "good") benchmarkers such as hardware unboxed.
For example the 5800x3d is ~15% faster than the 5800x in both hardware unboxed 41 game benchmark, as WELL as the meta review from 3Dcenter.