Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 709 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

reaperrr3

Junior Member
May 31, 2024
14
37
51
While Zen5 is certainly a far cry from what some of the earlier hype suggested, I honestly think some posts here (and some of the somewhat clickbaity YT reviews) are more negative than is justified.

- the Computerbase review clearly shows IPC is a double-digit increase over Zen4, including in games. It even blows Raptor IPC out of the water by quite a bit, despite only half the L2 and 67% the L3 per core.
Clockspeed (especially all-core) is all it's lacking, which is partially the result of the low default TDP, partially due to the backport from N3E to N4P.
- 9700X's MT problem is clearly the low TDP/PPT, it's running up to 500 MHz (!) below a 7700 and up to 900 (!) MHz below a 7700X with all cores under full MT load. I don't get why AMD didn't just go with at least 95W TDP/125W PPT for this one, would've still been cooler and more efficient than a 7700X while beating it much more clearly in MT.
- despite only a slightly better process and nearly 30% bigger cores, it doesn't consume much more power per clock than Zen4 and beats it in Perf/W.

The SKUs are somewhat poorly chosen/configured (and too expensive for the performance) and from what I read on the computerbase review, the AGESA/memory support situation was pretty much a nightmare until a couple of days ago, sure.
But that's all "only" poor SKU and launch management by AMD (which is disappointing in its own way, of course).
Zen5 as such isn't a dud as some people are claiming, and certainly no Bulldozer.
 

Ghostsonplanets

Senior member
Mar 1, 2024
691
1,113
96
per TechPowerUp, 11% uplift 9700X (with unlimited PBO) over 7700X at RPCS3 PS3 emulation, AVX 512 enabled.... what is going on !?


View attachment 104647
AVX-512 is only a small part of RPCS3. A big uplift on it won't translate to the same uplift for the PS3 emulator
No way, no way!
AMD can't "survive" selling this for another 2 years! Not for "average consumers".
Zen 6 needs to come in 12 months at most!
? AMD is fine. Core will do well on servers, it will be competitive on mobile and they'll have a cost advantage there. And it's a much more efficient core for Office/General DT tasks. The only public that will be disappointed will be the gaming one.

But that doesn't mean they'll need to rush Z6. If anything, they need to make sure Z6 improves radically over Z5.
 
  • Like
Reactions: misuspita

MS_AT

Senior member
Jul 15, 2024
246
569
96
This y-cruncher result from AT makes me giggle. [For me AVX512 is the only reason to buy the chip,
From AT's review - there are some workloads (e.g. AVX512) where the new architecture truly shines...
View attachment 104653
and this even does not use Zen5 optimized Y-Cruncher;)

BTW nice in-depth analysis from y-cruncher creator but probably not useful to most people;) http://www.numberworld.org/blogs/2024_8_7_zen5_avx512_teardown/ it explains why RPC3 did not see a boost.
 

CouncilorIrissa

Senior member
Jul 28, 2023
542
2,124
96
This y-cruncher result from AT makes me giggle. [For me AVX512 is the only reason to buy the chip,

and this even does not use Zen5 optimized Y-Cruncher;)

BTW nice in-depth analysis from y-cruncher creator but probably not useful to most people;) http://www.numberworld.org/blogs/2024_8_7_zen5_avx512_teardown/ it explains why RPC3 did not see a boost.
LOL:
The 40% IPC improvement in SpecInt (an early leak) is consistent with my tests showing 30-35% improvement in raw scalar integer that isn't memory-bound.
Could you post this in the architecture thread? This is very interesting.
 

CakeMonster

Golden Member
Nov 22, 2012
1,502
659
136
- the Computerbase review clearly shows IPC is a double-digit increase over Zen4, including in games. It even blows Raptor IPC out of the water by quite a bit, despite only half the L2 and 67% the L3 per core.
Clockspeed (especially all-core) is all it's lacking, which is partially the result of the low default TDP, partially due to the backport from N3E to N4P.
Agreed, this started yesterday with the clowns who tweeted about 'failure' comparing it to the previous generation with X3D, which is misleading on so many levels. Just because X3D cache is an impressive addition for some usages, doesn't mean the base architecture is a failure. It just means the extra cache has a huge impact in those benchmarks. The right comparison would be to the vanilla models, and from what I see so far there may be some valid criticism to some multithread benchmarks over Zen4 vanilla (prob power limited), but overall it seems like a good iterative step.

On a different topic, I'd have absolutely loved to see this architecture on N3*. But it seems both NV and AMD cheaped out, and are saving N3* for the generations launching in 2026.
 

SolidQ

Senior member
Jul 13, 2023
540
632
96
seems 9800X3D gonna have crazy efficiency
8d80857ba07b118befbe2c8c4ebcf311.png
 

tsamolotoff

Member
May 19, 2019
184
319
136
For gamers, there is the X3D variant to wait for.
I concur, if I bought a new CPU solely for gaming, I would wait for x3d anyway even if the prophesized 40% (int) ipc was really there. Extra cache is basically heaploads of (almost) free performance, no sense in rushing and buying something that can be 50% slower in games where CPU performance really matters (extraction shooters, MMOs, various competitive games that x3d especially excels at)
 
Last edited:
  • Like
Reactions: Racan

branch_suggestion

Senior member
Aug 4, 2023
392
875
96
BTW nice in-depth analysis from y-cruncher creator but probably not useful to most people;) http://www.numberworld.org/blogs/2024_8_7_zen5_avx512_teardown/ it explains why RPC3 did not see a boost.
That chart is a real eye opener, the core is future looking and as more programs are built with AVX-512 in mind, along with anything heavy on scalar INT, the thing will look better over time.
Goes to show how prevalent SSE and early AVX implementations are, especially on Windows.
 

Rheingold

Member
Aug 17, 2022
55
151
76
Regarding PBO, wider cores, and how much performance is left on the table for the 9700X:

1723039875003.png

This shows the wattage provided for the cores when using 88W or 142W PPT. The limit for the cores is always lower than the overall limit. The 7700X (blue) isn't limited during full load when using 142 watts, but the Zen 5 cores in the 9700X (black) is still being capped at that limit. This shows the increased power requirements of the wider cores, and raises the question why AMD chose to lower the TDP classes for this of all generations. It's an ecological but odd choice. This also means that there is more performance to be gained from PBO than before.
 

CakeMonster

Golden Member
Nov 22, 2012
1,502
659
136
Not releasing the reviews of the whole lineup on the same day smells like a PR disaster for AMD, the higher models should be less power limited. You really don't want those (even if few) benchmarks that show a tiny regression to get much media attention.
 
  • Like
Reactions: tsamolotoff

MarkPost

Senior member
Mar 1, 2017
322
616
136
Non-X3D Ryzen is garbage in gaming and AMD makes it seem better than it is, no surprise there. Not like Zen5 would change anything with its reused uncore and mediocre core. Though it will make vanilla Granite Ridge kind of DOA in DIY desktops.

Yeah, that’s an idiotic blunder from AMD. A whole lot of people look at the charts and see garbage performance improvements in MT loads without caring about the fact that the CPUs are running into power limits, which paints this already disappointing launch in an even worse light.
well, brain is there for some reason, to use it.
 
Last edited: