Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

Page 300 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Vattila

Senior member
Oct 22, 2004
799
1,351
136
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).

Untitled2.png


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! :)
 
Last edited:
  • Like
Reactions: richardllewis_01

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,438
14,406
136
What I am looking for is 12900k vs 5950x, but Markfw only has 12700.
I am interested in CB R23 score at different number of threads, because in reviews you either have 1T or Max number of threads and nothing between.



.
The problem is that my 12700F and all but one 5950x are on linux. The one 5950x on windows is an older motherboard, so it does not work that well, but still I would have no ADL cores to test. My only testing has been with DC applications, and I disabled the E-cores since they don't work as well as AVX512 in primegrid. They are strong cores, but the power usage is much higher than Zen 3 per core.
 
  • Like
Reactions: Grazick and Drazick

uzzi38

Platinum Member
Oct 16, 2019
2,556
5,531
146
For example, stock R5 5600X hits 4.5 on all cores or 12 threads boost.


R5 7600X, hm probably minimal 5-5.1ghz all core boost no problem.
I was talking about the 7950X but sure.
 
  • Like
Reactions: Tlh97 and Kaluan

uzzi38

Platinum Member
Oct 16, 2019
2,556
5,531
146
Everyone is discussing only desktop, but In my opinion mobile is where AMD will shine.
Both with Phoenix(strong IGP) and with Raphael(16C).
Intel won't give you a CPU with more than 96EU so Phoenix will win by default.
It's likely Intel will release 8P16E mobile CPU, but power will be a problem as is with ADL and AMD will double the number of cores.
Power is an issue because you have to cool the chip. In reality things are going to be closer than you think
 

Abwx

Lifer
Apr 2, 2011
10,842
3,295
136
I was talking about the 7950X but sure.

35% uplift for a 7950X vs 5950X should be easy to achieve because of the latter s relatively low frequency in MT.

When it comes to the 7600X vs 5600X that appear as much difficult since the 5600X run at 4.5GHz all cores, this would mandate roughly 5.5GHz for a 7600X if we are to use as basis 10% better MT IPC for Zen 4.
 

Gideon

Golden Member
Nov 27, 2007
1,598
3,520
136
4.5Ghz @ 230W is something that 5950X can do today in CB / Blender, expect Zen 4 to clock higher than this given the new node and the speed oriented design.

Yes. We know the IPC (ballpark) and based on that AMD needs to hit around 5 ghz all core, just to hit the demoed 35% MT gain.

The rumored base clock is 4.7Ghz so it shouldn't be a problem (just look at the 5950X base clock)
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,326
2,806
106
Power is an issue because you have to cool the chip. In reality things are going to be closer than you think
For comparison:
8C16T AMD Ryzen 9 PRO 6950HS - 11,515 in CB R23 (40W Sustained)
6P8E20T Intel Core i9-12900HK - 16,530 in CB R23 (45W Sustained)
8P8E24T Intel Core i9-12900HX - 18,337 in CB R23 (86W Sustained)

I think 23,000 points in CB R23 should be possible within 55-65W for Zen4.
If Intel by adding 8 more E cores can increase performance by 25% to be on par with Zen4 and still stay at 86W, then good job Intel.
 

Kaluan

Senior member
Jan 4, 2022
500
1,071
96
4.5Ghz @ 230W is something that 5950X can do today in CB / Blender, expect Zen 4 to clock higher than this given the new node and the speed oriented design.
I can tell you now that 4.5GHz all core is going to be extremely easy to achieve.
Greymom55 seemed to state that 5,1GHz all-core is not a problem for Ryzen 7000. And when asked if that's for dual or single CCD SKUs he just replied "8 cores can go 5,5GHz"

Stock/PBO/manually tuned/whatever... make of that what you want, just thought it was pertinent to bring up.

PS: There's also the 5,5GHz supposedly all-core ES in light workloads (gaming) AMD demoed a few months back.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,326
2,806
106
That s not at 45W, the number is for the first run, in the power charts we can see that the laptop consume 105-110W in Cinebench with an external monitor...
You are right, the numbers for ADL are much higher than for Zen 3.
R9 PRO 6950HS uses only Ø64.7 (49.2-68) in CB R15 with external monitor.
i9-12900HK uses Ø106.1 (102-113) in CB R15 with external monitor.
i9-12900HX uses Ø177.1 (166.7-232) in CB R15 with external monitor.

edit: I checked some reviews at ultrabookreview.com and 6800U vs 6900HX vs 12900H looks like this in CB R15 at 15th run.
R9-6900HX (35W)i9-12900H (35W)R7-6800U (25W)
CB R15192516611598
 
Last edited:

moinmoin

Diamond Member
Jun 1, 2017
4,926
7,608
136
I would not agree for that math, as E core use a lot less power so in multithread work it can push the limits while the zen and P cores will be limited by the power more so
also I almost never see point in thread in task I need
This is getting off topic in a Zen 4 thread, but ADL E-cores are actually equal to less efficient than ADL P-cores at the frequency they are (ab-)used.

Chips and Cheese did some article about that in January:

I discussed that in the Intel Lakes & Rapids thread (where this discussion belongs):
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
171watts needed atm for 5950x @ 4500mhz allcoreView attachment 66287
You're idling at 30c and maxing at 60c at 4.5Ghz all-core in CB?

I'll wait for @pakotlar to provide a more realistic ballpark figure on consumption and temps. Preferably, untweaked as well.

So, from the feedbacks, it says the base is 4.7Ghz for the 7950x? That's 1Ghz higher than the base clock of 5950x at 3.7Ghz, right? On a denser node and at a higher tdp. What's it going to need to cool those at 4.7Ghz, let alone 5.0 to 5.1Ghz all-core, sustained on blender?
 

uzzi38

Platinum Member
Oct 16, 2019
2,556
5,531
146
You're idling at 30c and maxing at 60c at 4.5Ghz all-core in CB?

I'll wait for @pakotlar to provide a more realistic ballpark figure on consumption and temps. Preferably, untweaked as well.

So, from the feedbacks, it says the base is 4.7Ghz for the 7950x? That's 1Ghz higher than the base clock of 5950x at 3.7Ghz, right? On a denser node and at a higher tdp. What's it going to need to cool those at 4.7Ghz, let alone 5.0 to 5.1Ghz all-core, sustained on blender?
1661176700657.png

(I'm actually not even joking. Because [REASONS], you should expect the chip to run at TjMax under a full all-core workload even with a hecking big AIO on it.)

EDIT: For clarification, I don't know if Cinebench will be enough to do it, it's going to be this way for heavy all-core workloads
 
Last edited:
  • Wow
Reactions: Elfear

moinmoin

Diamond Member
Jun 1, 2017
4,926
7,608
136
That might be possible with an all-core OC, but PBO on my 5950X for example has it running at ~4.3GHz at 200W+, so this isn't a good indicator of power.
If 171W for 4.5Ghz all core on 5950X is possible with all core OC, but PBO needs significantly more, doesn't that just point to plenty room for optimization of the automatic behavior programmed in the firmware? Manual optimization will always be hard to beat, but the aim should be to bring automatic optimization as close as possible.
 

Hitman928

Diamond Member
Apr 15, 2012
5,159
7,595
136
So, from the feedbacks, it says the base is 4.7Ghz for the 7950x? That's 1Ghz higher than the base clock of 5950x at 3.7Ghz, right? On a denser node and at a higher tdp. What's it going to need to cool those at 4.7Ghz, let alone 5.0 to 5.1Ghz all-core, sustained on blender?

Base clock of the 5950x is 3.4 GHz. That is what it should run if you turn off boosting so that it keeps to its rated TDP (105 W). With boosting turned on, there isn't a guaranteed base clock but it seems to be around 3.8 GHz or so, depending on silicon quality, voltage delivery from the MB, etc.
 

Det0x

Golden Member
Sep 11, 2014
1,027
2,953
136
You're idling at 30c and maxing at 60c at 4.5Ghz all-core in CB?

I'll wait for @pakotlar to provide a more realistic ballpark figure on consumption and temps. Preferably, untweaked as well.
You do understand each CCD were only using ~75watt with that 170 PPT figure ? (SOC was using ~20w)

75watt over ~83.7mm2 is not too hard to cool when you are running custom watercooling.. So yeah, delta was 30 degrees between idle and load.

That might be possible with an all-core OC, but PBO on my 5950X for example has it running at ~4.3GHz at 200W+, so this isn't a good indicator of power.
With all automatic PBO tuning i need 220watt to hit same 4500mhz in Cinebench r23.
220 watt.jpg
Optimally could pretty much have ran 4650/4500mhz as first CCD0 is ~150-200mhz better than second CCD1 at same vcore..
Really hope they fix this PBO drawback with Zen4, as all Zen3 dual CCD cpus are always limited by their worse CCD1 as they can only run allcore clockspeeds and not separated "8core + 8core".
 
Last edited:

Abwx

Lifer
Apr 2, 2011
10,842
3,295
136
AMD state that in Cinebench the 7950X perform >35% better than a 5950X@142W PPT as well as being 1.25x more efficient, that means that the >35% better perf is at 154W PPT for the 7950X, it cant be more clear than this...

17-1080.8ba48c82.png



 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,326
2,806
106
AMD state that in Cinebench the 7950X perform >35% better than a 5950X@142W PPT as well as being 1.25x more efficient, that means that the >35% better perf is at 154W PPT for the 7950X, it cant be more clear than this...

17-1080.8ba48c82.png



If 5950X has 3760 Mhz in CB R23, then Zen4 with 8% better IPC in CB R23 would need 25% higher clocks and that is 4700 Mhz for 35% higher performance.
If It manages 35% higher CB R23 score at ~154W PPT, then I have to wonder at 230W PPT how much faster will It get.

Saylick: 15% higher clocks would mean 5.4GHz. Performance would be at best 135 *1.15= 155 or 55% higher than 5950X or ~39500 in CB R23. It looks a bit too good to be true.
 
Last edited:

inf64

Diamond Member
Mar 11, 2011
3,680
3,943
136
Stock motherboard settings with PBO disabled in bios
CPU used only 128 watt PPT, seems to be limited by the stock TDC and EDC limits in Cinebench r23
Max allcore effective clocks were 3760mhz

So to sum up with regards to Zen 4 :

16 x Zen 3 cores at 3.76Ghz score 25.5K points while capped at 142W PPT.

At the same 142W PPT, Zen 4 should score 25.5K x 1.25 ~=31.87K points. Normalize for MT IPC gain (~10%) : 31.87 / 1.1 ~=28.97K points => 16C Zen 4 should be running this hypothetical test at 28.97/25.5~=1.136 or 13.6% higher clock than Zen 3 part => 3.76Ghz x 1.136~= 4.27Ghz.

In theory, a Zen 4 16C @ 4.3Ghz SKU *should* score 31.87 (~32K) points in R23 while consuming the same 142W PPT. This leaves a lot of TDP headroom for all core boost if the new maximum for PPT is ~230W. I expect an all core boost to be app dependent but ~5-5.2Ghz should be doable.
 
Last edited: