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!

 

[Thunder 57]

I'm curious, where is Henry Swagger? he is inactive for 1 month. I'm still waiting to hear his optimistic comments about Intel revenues, Zen 4 performance etc.

Funny, yes. But perhaps more concerning is that you are keeping tabs on him.

 

[Vope45]

Did the bridie say anything about the V-Cache launch?

Announced the same day as raptor lake

 

[Zucker2k]

IPC chart for 8c/16t@4GHz, how high those architectures clock is irrelevant there.
Also, clocks obviously don't scale linearly in games.

Their comparison are vs SKUs, not architectures. 5800X3D (+15%) was showcased vs 5800X, 7600X (+5%) vs 12900K (DDR5-6000CL30 on both) and 7950X (+15%) vs 5950X. It would be a lie to say the 2-5% 5950X has over 5800X are statistically irrelevant. It's pretty obvious they want to imply Zen4 SKUs edge out 5800X3D.

I'm curious to see how all these pan out with 5800x3d vs 7950x v 13900k in a couple of months. Let's see who's the gaming King.

 

[FangBLade]

Funny, yes. But perhaps more concerning is that you are keeping tabs on him.

He is old friend, i know him very well from other forums, he is one of the biggest (word that isn't allowed here), or can i say supporter of Intel?

 

[Thunder 57]

You're right. .What interests me is that amd chose to include more games this in the geo mean ipc calculation. We'll see if the actual ipc uplift is 13% in a wide range of apps and games, which I hope it should be.

Agreed. I'd like to see h264 and h265 encoding numbers.

 

[Abwx]

Actually, I would like to see sustained runs of ANY benchmark that is stressful, for 10 minutes or more. I want to see sustained performance and efficiency. This is just a short run.

They have published datas on efficency at different power levels, obviously they chosed to use a higher power than was intended at first, if you look for much more efficient computing you should set the CPU at lower powers :

AMD Ryzen 7000: Vier neue CPUs bieten +13 % IPC und bis zu 5,7 GHz Takt

AMD hat tief gestapelt und zur Vorstellung überrascht: Die neuen Ryzen 7000 sind nicht nur schneller, sondern zum Teil auch günstiger.

www.computerbase.de

 

[Kaluan]

You're right. .What interests me is that amd chose to include more games this in the geo mean ipc calculation. We'll see if the actual ipc uplift is 13% in a wide range of apps and games, which I hope it should be.

Zen3 app IPC over Zen2 is ~12% (while general IPC is 19%), Zen4 over Zen3 does indeed seem like ~9% (13% general). So it's just more of the same IMHO. Unless people want to also call Zen3 IPC slides "foul play"/misleading, I see no issue with the Zen4 ones.

PS: Always worth noting that AMD does IPC 8c/16c@x clocks, not ST IPC like tech reviewers typically do. TBH I prefer AMD's way.

 

[Thunder 57]

I'm curious to see how all these pan out with 5800x3d vs 7950x v 13900k in a couple of months. Let's see who's the gaming King.

I'm going with the 7900X. More than enough cores and more cache per core than the 7950X. 13900k might put up a good fight too. But after riding that AMD presentation high I'm guessing 7900X.

 

[inf64]

Zen3 app IPC over Zen2 is ~12% (while general IPC is 19%), Zen4 over Zen3 does indeed seem like ~9% (13% general). So it's just more of the same IMHO. Unless people want to also call Zen3 IPC slides "foul play"/misleading, I see no issue with the Zen4 ones.

PS: Always worth noting that AMD does IPC 8c/16c@x clocks, not ST IPC like tech reviewers typically do. TBH I prefer AMD's way.

Ian verified the ~19% uplift in SPEC when he reviewed Ryzen 5000, so that checks out. I expect the same to be true for Zen 4.

AMD Zen 3 Ryzen Deep Dive Review: 5950X, 5900X, 5800X and 5600X Tested

www.anandtech.com

"IPC wise, looking at a histogram of all SPEC workloads, we’re seeing a median of 18.86%, which is very near AMD’s proclaimed 19% figure, and an average of 21.38% - although if we discount libquantum that average does go down to 19.12%. AMD’s marketing numbers are thus pretty much validated as they’ve exactly hit their proclaimed figure with the new Zen3 microarchitecture. "



There are some obvious outliers, just like we see in Zen 4 slide (although AMD used various different benchmarks instead of SPEC shown above).

 

[dnavas]

Agreed. I'd like to see h264 and h265 encoding numbers.

If there's one area that Intel might win, it's QuickSync.
The announced pricing is quite reasonable, seems to me. Yes, low end is a little high, but if you're sensitive to pricing, you won't be doing an AM5/DDR5 build quite yet. The 7950X is reasonable, count me surprised.

 

[Vope45]

Ian verified the ~19% uplift in SPEC when he reviewed Ryzen 5000, so that checks out. I expect the same to be true for Zen 4.

AMD Zen 3 Ryzen Deep Dive Review: 5950X, 5900X, 5800X and 5600X Tested

www.anandtech.com

"IPC wise, looking at a histogram of all SPEC workloads, we’re seeing a median of 18.86%, which is very near AMD’s proclaimed 19% figure, and an average of 21.38% - although if we discount libquantum that average does go down to 19.12%. AMD’s marketing numbers are thus pretty much validated as they’ve exactly hit their proclaimed figure with the new Zen3 microarchitecture. "

View attachment 66783

There are some obvious outliers, just like we see in Zen 4 slide (although AMD used various different benchmarks instead of SPEC shown above).

I reckon amd actually calculate the ipc uplift from spec, then build up the testing suite accordingly.

 

[Thunder 57]

If there's one area that Intel might win, it's QuickSync.
The announced pricing is quite reasonable, seems to me. Yes, low end is a little high, but if you're sensitive to pricing, you won't be doing an AM5/DDR5 build quite yet. The 7950X is reasonable, count me surprised.

Quicksync and GPU encode is fine and good (enough) but nothing beats CPU.

As for the prices, remember back when Zen 3 launched they had 7nm for CPU's, GPU's and consoles. There wasn't enough to go around. Without consoles gobbling up 5nm, AMD may have more supply and was generous enough to keep prices low on the high end. I'd still like to see a $200-249 chip in there but they will make their money on early adopters first as always.

 

[desrever]

I'm going with the 7900X. More than enough cores and more cache per core than the 7950X. 13900k might put up a good fight too. But after riding that AMD presentation high I'm guessing 7900X.

The 7950X is cheaper per core than every other 7000 CPU.

 

[inf64]

I reckon amd actually calculate the ipc uplift from spec, then build up the testing suite accordingly.

They use SPEC regularly, I'm surprised they haven't listed it in the endnotes this time. Once we get a more detailed look into the Zen 4 core (a la deep dive Ian did in the past), we'll see how the IPC breakdown goes.

 

[Hans Gruber]

We live in weird times. Even if Raptor Lake runs the entire table with performance victories. Once the 3D v-cache chips are out. Intel will get smoked in gaming performance by Zen 4. The 5800x3D was gimped by voltage limitations because the onboard cache could not handle above 1.2v. That problem is gone in Zen 4. So the 7700X3D chip will have stock 5.4ghz boost speeds with 3d cache vs. 4.5ghz against the 5800X3D chip. That's not counting the IPC gains and the DDR5 memory advantage.

I was just picking the middle CPU (7700X). From what I understand 3D v-cache will be available in all the Zen 4 CPU's.

 

[CakeMonster]

I'm going with the 7900X. More than enough cores and more cache per core than the 7950X. 13900k might put up a good fight too. But after riding that AMD presentation high I'm guessing 7900X.

I always wondered about that, as I have the 5900X and figured that might help. I've not been able to find any good analysis, everytime I've asked around people have dismissed it, so I'm leaning toward it not doing much difference, and if it did it helps less than having 8 threads instead of 6 threads fit inside one CCX.

Let me know if you know of any good analysis and/or testing on it.

 

[Saylick]

So apparently the mop cache is more than 50% larger...

https://twitter.com/x/status/1564413952108335105

 

[FangBLade]

We live in weird times. Even if Raptor Lake runs the entire table with performance victories. Once the 3D v-cache chips are out. Intel will get smoked in gaming performance by Zen 4. The 5800x3D was gimped by voltage limitations because the onboard cache could not handle above 1.2v. That problem is gone in Zen 4. So the 7700X3D chip will have stock 5.4ghz boost speeds with 3d cache vs. 4.5ghz against the 5800X3D chip. That's not counting the IPC gains and the DDR5 memory advantage.

I was just picking the middle CPU (7700X). From what I understand 3D v-cache will be available in all the Zen 4 CPU's.

Don't forget that zen 4 as architecture will be more optimized for 3dcache, double l2cache and other things will greatly benefits from additional l3 so performance increase should be even more.

 

[Vope45]

We live in weird times. Even if Raptor Lake runs the entire table with performance victories. Once the 3D v-cache chips are out. Intel will get smoked in gaming performance by Zen 4. The 5800x3D was gimped by voltage limitations because the onboard cache could not handle above 1.2v. That problem is gone in Zen 4. So the 7700X3D chip will have stock 5.4ghz boost speeds with 3d cache vs. 4.5ghz against the 5800X3D chip. That's not counting the IPC gains and the DDR5 memory advantage.

I was just picking the middle CPU (7700X). From what I understand 3D v-cache will be available in all the Zen 4 CPU's.

V cache version will only be available for the 7800x3d. I don't know if they will release 7950x3d due to mostly heat. Heat will be a big problem going forward. I'll actually wait until zen 5 unless 3d vcache part will blow raptor lake successor out of the water.

 

[inf64]

Found this on Gordon's stream( at 40:19 mark)

7950X ran the V-ray benchmark at 5.1GHz all-core turbo . There you go, we have an all core turbo of 5.1GHz in compute heavy benchmarks, so 5.1/3.9Ghz x 1.13 = 1.47 or 47% better in compute heavy stuff versus 5950X.

 

[Kaluan]

Did anyone check the Ryzen 7000 SKUs listing on AMD's website BTW? Why do they say "requires discrete GPU", I thought they come with a small RDNA2 iGPU?

 

[nicalandia]

I'm curious to see how all these pan out with 5800x3d vs 7950x v 13900k in a couple of months. Let's see who's the gaming King.

AMD Just Showed a Game where is 25%+ performance better than 5800X3D...

 

[Thunder 57]

Did anyone check the Ryzen 7000 SKUs listing on AMD's website BTW? Why do they say "requires discrete GPU", I thought they come with a small RDNA2 iGPU?

They do.

https://www.amd.com/en/processors/ryzen

Go all the way to the bottom, specifications. 2 CU's.

 

[HurleyBird]

Did anyone check the Ryzen 7000 SKUs listing on AMD's website BTW? Why do they say "requires discrete GPU", I thought they come with a small RDNA2 iGPU?

It's a typo. Even though it wasn't mentioned in the livestream, an AMD rep talked about the iGPU in the PCWorld stream.

 

[Exist50]

Found this on Gordon's stream( at 40:19 mark)

7950X ran the V-ray benchmark at 5.1GHz all-core turbo . There you go, we have an all core turbo of 5.1GHz in compute heavy benchmarks, so 5.1/3.9Ghz x 1.13 = 1.47 or 47% better in compute heavy stuff versus 5950X.

IPC and frequency aren't quite 1:1 with performance (or each other) across clock speeds because of the effectively slower memory. IIRC, it's often closer to 0.8, but that will be a great test for someone to try on release. Also some interesting cache harmonics at play.