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!

 

[TESKATLIPOKA]

Without rebooting or anything:
View attachment 66831

Thank you very much.
So It scores 16,210 points and frequency is 2377-2402MHz.
Then, 7950x at 65W TDP should score 28,205 points(+74%).
IPC is 9% higher in CB R23 based on AMD's presentation, so 174/109 = 1.6 or 60% higher frequency. That would mean 3803-3843 MHz.

P.S. Can you also test 170W TDP(230W PPT)?

 

[CakeMonster]

Then I wonder why TDP was increased to 170W, given that it doesn't scale much beyond 105 W according to AMD. I guess the power draw per core was targeted to increase per core clock speeds.

My guess is that they had pretty good intel (ahem) about the 13900K and wanted to make sure they didn't end up a few % below it (see the unconfirmed GB scores that were extremely close).

 

[Timmah!]

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.

Do you have some kind of inside info to claim this or its just an assumption/guess on your part?
It would be nice to know for sure, so i wont wait several months for 7950x3D and then they wont even release it.

 

[Abwx]

Huh, so according to AMD's numbers the 7950X score @ 65 W TDP is greater than the 5950X score @ 105 W TDP. IIRC at default settings the 5950X scores ~26,000 pts.

Yes, because at those low power, and hence frequencies, N5P should be around the 0.5 x te power at isoperf and same uarch, add the IPC and the fact that the 7950X IOD is likely to use 5-7W at this frequency while the one in the 5950X is at 15-18W.

 

[Abwx]

dp

 

[Zucker2k]

The

True and at ISO speed/Core and ISO Memory(DDR4 vs DDR4) Zen3 was a match in gaming to Alder Lake(12400K vs 5600X) due to better memory subsystem even if they were behind on IPC performance.

So my point has been, Look AMD has reach speed and Memory parity with Intel while their memory subsystem has been enhanced, this will translate to better gaming no doubt about it.

5600x enjoyed superior clocks, everywhere, over the 12400, not to mention ~40% bigger l3 cache.

 

[CakeMonster]

Do you have some kind of inside info to claim this or its just an assumption/guess on your part?
It would be nice to know for sure, so i wont wait several months for 7950x3D and then they wont even release it.

I'm 100% in the same situation.

That user does not provide more info backing it up except being confident, so its not worth anything even if they could theoretically be Lisa herself. Even semi famous 'leakers' have been caught BS'ing.

 

[tamz_msc]

Thank you very much.
So It scores 16,210 points and frequency is 2377-2402MHz.
Then, 7950x at 65W TDP should score 28,205 points(+74%).
IPC is 9% higher in CB R23 based on AMD's presentation, so 174/109 = 1.6 or 60% higher frequency. That would mean 3803-3843 MHz.

P.S. Can you also test 170W TDP(230W PPT)?

I know that with PBO the 5950X scores around 28,000-29,000 points.

 

[Det0x]

Thank you very much.
So It scores 16,210 points and frequency is 2377-2402MHz.
Then, 7950x at 65W TDP should score 28,205 points(+74%).
IPC is 9% higher in CB R23 based on AMD's presentation, so 174/109 = 1.6 or 60% higher frequency. That would mean 3803-3843 MHz.

P.S. Can you also test 170W TDP(230W PPT)?

Still with 21 hours update without closing anything.

PBO CO disabled (at 0) and PPT limited to 230w with stock auto boosting:

 

[shady28]

How come people talk about game scaling as if the cpu alone is responsible for generating the fps? I believe the newer generation of gpus will reveal the real difference between a chip like the 5600x and 12900k.

GPUs really caught up with CPUs in the past 3 generations. Should be obvious, GPU performance goes up 20-30% every 2 years while CPUs were getting +5% per year since like, Sandy bridge until Zen 3.

We already see a 20%+ delta in FPS between Zen 2 and Alder Lake with a 3080.

So yeah, when a 4070 comes out and performs like a 3090 Ti, the difference in CPUs will become very very clear for gamers.

This generation of GPUs may well make everything before Alder Lake obsolete for high end gamers, including Zen 3. There's no point in buying a 4080 if you are going to leave 20% of your FPS on the table due to your CPU.

 

[Abwx]

Still with 21 hours update without closing anything.

PBO CO disabled (at 0) and PPT limited to 230w with stock auto boosting:
View attachment 66834

At 65W, and moreover at 88W PPT, it should score quite higher than 16k.

 

[TESKATLIPOKA]

Still with 21 hours update without closing anything.

PBO CO disabled (at 0) and PPT limited to 230w with stock auto boosting:
View attachment 66834

Great work, thanks.
5950x 170W -> 29,293 points at 4379MHz.
7950x should score 39546 points.
Frequency should be ~24% higher(135/109=24) -> ~5430 MHz!

 

[Det0x]

At 65W, and moreover at 88W PPT, it should score quite higher than 16k.

PBO is stupid when you set the limit below stock values.. (soc with those settings using 21 19w alone at those runs above dont help either)

This is what i can do manually:

16 Zen3 cores @ 49 watt = 20441 points in Cinebench r23 (~lowest powerlevel i could run with my current setup)
3100mhz / 2950mhz @ 0.7 vcore under load -> ~2.125watt per core under load
20441points/49watt = 417 points per watt

16 Zen3 cores @ 64 watt = 23701 points in Cinebench r23
3575mhz / 3425mhz @ 0.781 vcore under load -> ~3.063watt per core under load
23701points/64watt = 370 points per watt

16 Zen3 cores @ 88 watt = 27037 points in Cinebench r23
4075mhz / 3925mhz @ 0.887 vcore under load -> ~4.5625watt per core under load
27037points/88watt = 307 points per watt

 

[HurleyBird]

The MT score for AES is indeed weird for all Ryzen 7000 parts, I wonder what is happening there. It's basically lowering the total MT score(avg). It could be memory related( BW limited?) but who knows at this point. We'll have to wait for final reviews and chips in the hands of users to figure it out.

Could be something they hope to iron out with the launch BIOS.

 

[nicalandia]

The 5600x enjoyed superior clocks, everywhere, over the 12400, not to mention ~40% bigger l3 cache.

12400F vs 5600X is the closes we will get to a Zen3 vs Golden Cove at ISO Core, ISO Speed and Memory.

Just take the numbers at(Power Limit remove) for the extra speed.

Intel Core i5-12400F Review - The AMD Challenger

The Intel Core i5-12400F comes at an extremely attractive price point, yet offers performance comparable to AMD's Ryzen 5 5600X. While Intel introduced a Hybrid core design with Alder Lake, the 12400F is a P-core only design, which helps avoid potential compatibility issues with E-cores.

www.techpowerup.com

 

[Rigg]

I'm sorry but there seem to be an awful lot of typos in the specs they listed. I haven't even looked at but two of them, one says they used a 12900K with AMD EXPO timings (Zen 4 specific) and now this one says they used DDR4-3600C30 on the 5900X.


View attachment 66814

I don't get the focus on the EXPO timings. This is nothing but XMP with AMD's branding. Were AMD CPU's previously handicapped when using (intel specific) XMP timings? I'm failing to understand how testing both CPU's on the exact same 6000 Mt/s CL32 DDR5 kit is some kind of grand conspiracy to misrepresent 12900k performance.

C'mon man! If you only get 10% more 1080p gaming performance (using a 3090ti) from 100% more memory speed (3200-6400), how much difference is a 7% jump (6000-6400) going to make? Show me the benchmarks. The copium is strong with you.

 

[Abwx]

PBO is stupid when you set the limit below stock values.. (soc with those settings using 21w alone dont help either)

This is what i can do manually:

16 Zen3 cores @ 49 watt = 20441 points in Cinebench r23 (~lowest powerlevel i could run with my current setup)
3100mhz / 2950mhz @ 0.7 vcore under load -> ~2.125watt per core under load
20441points/49watt = 417 points per watt

16 Zen3 cores @ 64 watt = 23701 points in Cinebench r23
3575mhz / 3425mhz @ 0.781 vcore under load -> ~3.063watt per core under load
23701points/64watt = 370 points per watt

16 Zen3 cores @ 88 watt = 27037 points in Cinebench r23
4075mhz / 3925mhz @ 0.887 vcore under load -> ~4.5625watt per core under load
27037points/88watt = 307 points per watt
View attachment 66837

I thought the uncore was at 18W while you have 21W, not a big difference but still, this is an advantage of Zen 4 with a much lower power for the rest of the chip.

That s quite precious info, thanks for the tests, a lot can be extracted from those numbers.

 

[Zucker2k]

I don't get the focus on the EXPO timings. This is nothing but XMP with AMD's branding. Were AMD CPU's previously handicapped when using (intel specific) XMP timings? I'm failing to understand how testing both CPU's on the exact same 6000 Mt/s CL32 DDR5 kit is some kind of grand conspiracy to misrepresent 12900k performance.

C'mon man! If you only get 10% more 1080p gaming performance (using a 3090ti) from 100% more memory speed (3200-6400), how much difference is a 7% jump (6000-6400) going to make? Show me the benchmarks. The copium is strong with you.

There are AMD and Intel specific RAM settings (timings) that can lead to significant performance boosts.

 

[TESKATLIPOKA]

PBO is stupid when you set the limit below stock values.. (soc with those settings using 21w alone dont help either)

This is what i can do manually:

16 Zen3 cores @ 49 watt = 20441 points in Cinebench r23 (~lowest powerlevel i could run with my current setup)
3100mhz / 2950mhz @ 0.7 vcore under load -> ~2.125watt per core under load
20441points/49watt = 417 points per watt

16 Zen3 cores @ 64 watt = 23701 points in Cinebench r23
3575mhz / 3425mhz @ 0.781 vcore under load -> ~3.063watt per core under load
23701points/64watt = 370 points per watt

16 Zen3 cores @ 88 watt = 27037 points in Cinebench r23
4075mhz / 3925mhz @ 0.887 vcore under load -> ~4.5625watt per core under load
27037points/88watt = 307 points per watt
View attachment 66837

This is why I love this forum. Seriously great work.
Those efficiency gains are brutal, Zen4 no longer looks that impressive to me.

 

[inf64]

I thought the uncore was at 18W while you have 21W, not a big difference but still, this is an advantage of Zen 4 with a much lower power for the rest of the chip.

That s quite precious info, thanks for the tests, a lot can be extracted from those numbers.

The huge perf/watt advantage that Zen 4 has over Zen 3 is going to make Dragon Range and Phoenix unbeatable, even versus Meteor Lake. AMD will have 16 big cores with highest IPC overall that will clock like mad at 45W TDP bracket, before Meteor Lake launches. Intel will need basically 16 MTL cores, even if they squeeze some IPC out of the new core while maintaining clocks - almost an impossible task.

 

[FangBLade]

Shady, reb0rn and similar Intel army, you should celebrate instead being triggered, AMD made huge jump again, and that means Intel will need to follow it, i can safely say we won't watch skylake era and 1% perf jumps anymore, core count will increase, perf jump will be huge, they are pushing each other like nothing before. AMD is stable company now, they are increasing R&D budget every year and that's good thing.

 

[Thunder 57]

This is why I love this forum. Seriously great work.
Those efficiency gains are brutal, Zen4 no longer looks that impressive to me.

That goes with pretty much any uarch. If Intel didn't push the E cores so high they would be much more efficient. A 65w CPU will pretty much always be more efficient than a 95/105W or 170W, etc.

 

[DrMrLordX]

So as it stands the best Intel is going to bring till 2024 is the 13900KS right? So 7950X3D will reign supreme till late 2024.

We don't know what the 13900k will be like or if Intel will bother with a 13900KS. But from the looks of things Raptor Lake may not meaningfully exceed Raphael as it is. It's a good thing that Raphael-X is coming anyway. AMD may not really need it.

Cheese had access to Raptor Lake and found no IPC increases in their own workloads https://chipsandcheese.com/2022/08/23/a-preview-of-raptor-lakes-improved-l2-caches/

That is a bit worrying. Especially since that may be the last meaningful desktop update from Intel until 2024.

I don't get the focus on the EXPO timings. This is nothing but XMP with AMD's branding. Were AMD CPU's previously handicapped when using (intel specific) XMP timings?

Basically, yes. Intel's memory controllers do not like the same timings as AMD's and vice versa. XMP 2.0 can do some pretty horrible things to timings and "hidden" voltages on AM4 systems. It's about time that AMD did something to attempt to address the problem.

 

[TESKATLIPOKA]

That goes with pretty much any uarch. If Intel didn't push the E cores so high they would be much more efficient. A 65w CPU will pretty much always be more efficient than a 95/105W or 170W, etc.

True, but look at what frequency his system actually runs when he sets It manually.
At 88W, he has comparable or a bit higher frequency than stock 5950x at 142W.

 

[FangBLade]

Some people still look at 13% IPC and says it is small, they forgot that Zen 4 in the end brings biggest performance jump, yes they choose clock path instead of IPC, but still performance jump is biggest in the Zen era, and power consumption will be much better than Intel so it is a clear win, can't wait to see 3d version, they have even better power efficiency, and with second generation 3dcache + better original cache hierarchy that will better cooperate with 3dcache + ddr5 + higher clocks = we should see significant perf jump in games, RPL won't catch up it even with best ddr5 and KS model.