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

Page 480 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Vattila

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
Sort by date Sort by votes
E

Exist50

Platinum Member
Aug 18, 2016
2,445
3,043
136
moinmoin said:
I'm not sure what's ironic about that. Is it also ironic that with ADL E cores can match P cores per thread with SMT enabled, but not with it disabled?
Click to expand...
Yes, it is.

The bigger core, for both AMD and Intel, justifies its existence by having better single thread performance. And yet as commonly deployed, that advantage is rendered moot by the usage of SMT. I would certainly call that irony. Or rather, a reason to fundamentally reevaluate where SMT makes sense.
 
Upvote 0 Downvote
utahraptor

utahraptor

Golden Member
Apr 26, 2004
1,052
199
106
This waiting game is absolutely brutal. Waiting for Zen 4, then Raptor Lake then Raptor Lake 6 GHZ and finally Zen 4 3d cache to make a decision on what to build is agonizing. I'm 99% sure I am going with 3d, but it the reviews bomb maybe I won't. I just hope there is not an announcement about the Raptor Lake refresh prior to the Zen 4 3d release date 😂
 
  • Like
Reactions: Joe NYC
Upvote 0 Downvote
N

nicalandia

Diamond Member
Jan 10, 2019
3,330
5,281
136
Exist50 said:
And this is a good example of the current issue with SMT. If you turn it on for the throughput, your per thread performance falls drastically. But leave it off, and something like Bergamo can still deliver great throughput with respectable per-thread performance. Ironically, Bergamo with SMT disabled is almost certainly faster than Genoa with it left on, per thread.
Click to expand...
By default it will have SMT On. So 256T per CPU Or 512 Threads per 2S System
 
Upvote 0 Downvote
T

Timmah!

Golden Member
Jul 24, 2010
1,418
630
136
utahraptor said:
This waiting game is absolutely brutal. Waiting for Zen 4, then Raptor Lake then Raptor Lake 6 GHZ and finally Zen 4 3d cache to make a decision on what to build is agonizing. I'm 99% sure I am going with 3d, but it the reviews bomb maybe I won't. I just hope there is not an announcement about the Raptor Lake refresh prior to the Zen 4 3d release date 😂
Click to expand...

When are the 3D reviews supposed to happen?
 
Upvote 0 Downvote
biostud

biostud

Lifer
Feb 27, 2003
18,248
4,757
136
utahraptor said:
This waiting game is absolutely brutal. Waiting for Zen 4, then Raptor Lake then Raptor Lake 6 GHZ and finally Zen 4 3d cache to make a decision on what to build is agonizing. I'm 99% sure I am going with 3d, but it the reviews bomb maybe I won't. I just hope there is not an announcement about the Raptor Lake refresh prior to the Zen 4 3d release date 😂
Click to expand...
The good thing about choosing the AM5 platform is, that you can upgrade to zen5 in a year ;)
 
  • Like
Reactions: lightmanek, Joe NYC, Tlh97 and 2 others
Upvote 0 Downvote
moinmoin

moinmoin

Diamond Member
Jun 1, 2017
4,950
7,659
136
Exist50 said:
The bigger core, for both AMD and Intel, justifies its existence by having better single thread performance. And yet as commonly deployed, that advantage is rendered moot by the usage of SMT. I would certainly call that irony. Or rather, a reason to fundamentally reevaluate where SMT makes sense.
Click to expand...
The evaluation of using SMT should already happen by process schedulers. If better single thread performance is desired that thread should get the core alone. SMT only comes into the picture once two threads share a core.

Simplified there are essentially two reasons to put two threads on one core:
  1. All cores are already in use. The additional thread would stall otherwise, using SMT allows it to run even if for two threads this means running slower than when having a whole core all alone. This increases throughput.
  2. Energy saving. Every core powered on uses power. Putting a second thread on an already active core doesn't increase power usage as much as activating a new core for that purpose. So while this means running slower than when having a whole core all alone, the result is also higher efficiency than running two threads on two separate cores. This increases power efficiency.
In the optimal case the schedulers should ensure that using SMT offers the above advantages and avoids the disadvantage of shared cores whenever single thread performance is desired. Whether they manage that is a different story.

So going back to 128C/128T Bergamo beating 64C/128T Milan, following equation should be natural:

128c/256t > 128c/128t > 64c/128t > 64c/64t

Exist50 said:
It will support SMT, but I'd be very curious to see how it's actually deployed.
Click to expand...
In shared VM servers as used in the cloud I'd consider disabling SMT a best practice for reproducibility of performance alone, but even more so due all the recent security vulnerabilities.
 
  • Like
Reactions: Tlh97 and BorisTheBlade82
Upvote 0 Downvote
H

Hitman928

Diamond Member
Apr 15, 2012
5,262
7,890
136
Racan said:
From these results I expect that the 3D cache will help Zen 4 a lot. If only they could clock as high...

Also explains the garbage gaming performance results from sites like computerbase and pcgameshardware who use DDR5 5200 for Zen 4.
Click to expand...

The reason for the bad gaming results from those sites has been known pretty much the whole time, it's just that certain people with an agenda tried to gaslight everyone into believing otherwise.

Zen4-3D should show the largest gains on sites like those where slow memory was used though who knows how the whole 1CCD with vcache situation will work out for the 2CCD SKUs. Other sites who use faster memory should show less gains though obviously their results were closer to RPL to begin with. I'm still of the opinion that, at least at the top of the line, performance between Zen4-3D and RPL will be so close that it won't really matter. Heck, we're pretty much already there if using tuned memory for both. The difference will be that you can save a decent chunk of money on memory with Zen4-3D and it should also be significantly more efficient.
 
Last edited:
  • Like
Reactions: alexruiz, lightmanek, inf64 and 2 others
Upvote 0 Downvote
IEC

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,330
4,917
136
Timorous said:
This is what happens when you pair Zen 4 with slow ram.
Click to expand...

$570 USD current Amazon.com pricing vs $345 for 7700X (+Jedi Survivor game). +2.5% performance uplift for the 13900K spending almost double on the core components.
1676467140335.png

With such a small margin of victory for Raptor Lake, X3D doesn't need much to make it a resounding win for gaming.
 
  • Like
Reactions: igor_kavinski, lightmanek, moinmoin and 6 others
Upvote 0 Downvote
S

scineram

Senior member
Nov 1, 2020
361
283
106
Racan said:
From these results I expect that the 3D cache will help Zen 4 a lot. If only they could clock as high...

Also explains the garbage gaming performance results from sites like computerbase and pcgameshardware who use DDR5 5200 for Zen 4.
Click to expand...
AKA running in spec. At stock.

this probably shows how AMD botched the fabric clocks. At the usual 3:2 speed DDR5-5200 gives only 1733 MHz IF clocks.
 
Upvote 0 Downvote
Heartbreaker

Heartbreaker

Diamond Member
Apr 3, 2006
4,227
5,228
136
IEC said:
$570 USD current Amazon.com pricing vs $345 for 7700X (+Jedi Survivor game). +2.5% performance uplift for the 13900K spending almost double on the core components.

With such a small margin of victory for Raptor Lake, X3D doesn't need much to make it a resounding win for gaming.
Click to expand...

He makes it pretty clear, bang/buck wasn't the point of the video. He just wanted to test memory scaling.

You can get lower tier Raptor Lake parts and get essentially the same gaming performance if looking for better gaming bang/buck:
relative-performance-games-1280-720.png
 
  • Like
Reactions: Henry swagger
Upvote 0 Downvote
I

itsmydamnation

Platinum Member
Feb 6, 2011
2,769
3,144
136
  • Like
Reactions: lightmanek
Upvote 0 Downvote
E

exitorious

Junior Member
Aug 8, 2019
9
42
91
itsmydamnation said:
if you look at the HUB above , they have 7200 CL32 , run that at 6000 you should be able to run CL 28 , if your super lucky and pump some volts maybe even CL 26
Click to expand...

I tried some Teamgroup 7200 cl34 ram that I use with my 13900k at 7200 and it wasn't stable no matter what I did (even stock speed) with an Asus Strix x670E-E board. I have a MSI Carbon x670e board now I haven't installed yet, so maybe I'll give it another try with that this weekend.
 
  • Like
Reactions: Tlh97, lightmanek, scineram and 2 others
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom