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

Page 459 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,227
1,660
136
I'd say serious professionals don't browse when their machine is doing critical work. Also, I would like to know how much of performance difference these cases you listed will have, or if it would be noticeable in the practical use.
As a work from home professional, I have an assigned workstation I have to use for everything (slack, video meetings, browsing docs/wiki/etc., compiling, other extremely computational heavy work). I don't get the luxury of "not browsing when my machine is doing critical work". I instead get to deal with it being largely unresponsive and everyone else gets to watch and hear my video stutter.
 

leoneazzurro

Golden Member
Jul 26, 2016
1,004
1,595
136
That's called lack of budget. In my case, I'd think twice before making one of my coworkers risk to lose a long simulation or making stuttering calls because of stupidly forcing someone to use the same machine for doing everything.
 

eek2121

Diamond Member
Aug 2, 2005
3,043
4,263
136
Zen 5 CPU-Z single core benchmark is supposedly 910. Will share source soon.


EDIT: Sorry,, I put the twitter link below, had to get to a desktop. I wanted to link the original baidu post, but have had issues.

That is for the engineering sample of Zen 5. For reference I get around 740-750 on my stock 7950X, so about 21% higher. Not bad.

Oh and if this was already posted, apologies. Sometimes this thread moves quite quickly and it can be hard to keep up.
 
Last edited:

StefanR5R

Elite Member
Dec 10, 2016
5,681
8,240
136
There are benchmarks where the 7950X beats the 7950X3D. That should NOT be happening.
Is there a remote possibility that these benchmarks included workloads which are power-limited?

I have an assigned workstation I have to use for everything (slack, video meetings, browsing docs/wiki/etc., compiling, other extremely computational heavy work). I don't get the luxury of "not browsing when my machine is doing critical work". I instead get to deal with it being largely unresponsive and everyone else gets to watch and hear my video stutter.
Too little RAM? Bad mass storage performance? Bad scheduling of GPGPU workloads? Wrong scheduling prioritization of interactive processes vs. batch processes? [Edit: Malware scanner needlessly intercepting I/O of the computing application?]

These things do not happen because of a dual-CCX CPU with different sizes of L3$ on the CCXs.
 
Last edited:

Abwx

Lifer
Apr 2, 2011
11,163
3,859
136
That's called moving the goalposts. I work with a lot of companies that have full time remote employees and no one gets more than one assigned workstation.

That s not the point, no workstation doing critical work should be used to say browse the net, dunno wich are those firms and employees that do not have the means to buy a standard laptop or SFF PCs for marginal tasks.
 

LightningZ71

Golden Member
Mar 10, 2017
1,658
1,940
136
AMD needs to have a shared L4 V-cache for both CCDs.
While it looks good on paper, there are several notable issues with it in practice.

The stacked VCache on the CCD has considerable bandwidth available to transfer data to the CCX. It is also so close that the latency hit is only a few clock cycles. A VCache on the IOD would have substantially less bandwidth between it and the CCX, resulting in a much reduced benefit. Also, one of the things that makes caches valuable is their low latency. Crossing the IF link to the IOD, having the memory controller process the request to the VCache would introduce many more clock cycles of latency, killing a portion of it's best feature.

There's also a power hit. If they really wanted to support VCache on the IOD, they would have to substantially increase the bandwidth on the IF link, likely via widening it and running it faster. Also, if it was really that useful, it would see a much higher duty cycle. All of that results in CONSIDERABLY more energy usage for that data transfer, eating into the package power budget and producing extra waste heat.

Finally, adjusting the memory controller on the IOD to accommodate the VCache would make it more complicated, introduce additional latency for EVERY memory access request from a CCX, and result in more heat production from the IDO, which is already on a trailing node for power and thermal purposes while simultaneously making it more difficult for that IOD to dissipate that waste heat.

The end result would be a MORE expensive product that was LESS performant in it's "best case scenario" target, and wouldn't improve much for any other targets. While it may allow better thermals for the CCDs, which might give you 400-500 more boost Mhz on a single core, what you are trying to do by moving the VCache to the IOD is make ALL CORE performance improve (after all, there's no really reason to do so otherwise). Unfortunately, all core performance is typically more limited by package power and total thermal load when it comes to throughput. The added IF link power load and thermal load would serve to limit the package's ability to boost all cores enough to erase any differences you might get. To cover for main memory latency for large data sets, the VCache would have to be massive, on the order of 128-256Mb. That's also going to consume a lot of power and thermal headroom.

It just doesn't add up.
 

inf64

Diamond Member
Mar 11, 2011
3,759
4,213
136
Zen 5 CPU-Z single core benchmark is supposedly 910. Will share source soon.


EDIT: Sorry,, I put the twitter link below, had to get to a desktop. I wanted to link the original baidu post, but have had issues.

That is for the engineering sample of Zen 5. For reference I get around 740-750 on my stock 7950X, so about 21% higher. Not bad.

Oh and if this was already posted, apologies. Sometimes this thread moves quite quickly and it can be hard to keep up.
For comparison , Zen 4 achieved 1% vs Zen 3 and Zen 3 got 12% over Zen 2 ( IPC) in CPUz 1T.
Zen 5 getting ~20% in this archaic benchmark bodes well for the average IPC uplift number. Zen 4 was cited to be between 11 and 13% IPC uplift (average) while Zen 3 average was ~19%.

Edit:
 

Saylick

Diamond Member
Sep 10, 2012
3,379
7,127
136
For comparison , Zen 4 achieved 1% vs Zen 3 and Zen 3 got 12% over Zen 2 ( IPC) in CPUz 1T.
Zen 5 getting ~20% in this archaic benchmark bodes well for the average IPC uplift number. Zen 4 was cited to be between 11 and 13% IPC uplift (average) while Zen 3 average was ~19%.
CPU-Z isn’t the best benchmark to begin with, but for what it’s worth…
 

Kepler_L2

Senior member
Sep 6, 2020
459
1,891
106
For comparison , Zen 4 achieved 1% vs Zen 3 and Zen 3 got 12% over Zen 2 ( IPC).
Zen 5 getting ~20% in this archaic benchmark bodes well for the average IPC uplift number. Zen 4 was cited to be between 11 and 13% IPC uplift (average) while Zen 3 average was ~19%.
It's a terrible benchmark but AFAIK CPU-Z should be a much bigger uplift on Zen5. Also there are no IF/IMC changes so I'm skeptical of this post.
 

Abwx

Lifer
Apr 2, 2011
11,163
3,859
136
AFAIK CPU-Z should be a much bigger uplift on Zen5.

No, because it was modded in 2017 to greatly favour Intel after Zen 1 yielded much better scores than SKL, that s the most possible biaised bench, they alleged that it was due to a bug but that was plain lies, that was just $$ at play.

Edit : Zen 4 gained something like 1-2% in CPU Z comparatively to Zen 3,
that s just impossible since it gained 12% in 7 Zip and as much
in FP tests, so that s just a prove that this bench is not a real one,
just a marketing tool on Intel s behalf.

 
Last edited:

Kepler_L2

Senior member
Sep 6, 2020
459
1,891
106
No, because it was modded in 2017 to greatly favour Intel after Zen 1 yielded much better scores than SKL, that s the most possible biaised bench, they alleged that it was due to a bug but that was plain lies, that was just $$ at play.

Edit : Zen 4 gained something like 1-2% in CPU Z comparatively to Zen 3,
that s just impossible since it gained 12% in 7 Zip and as much
in FP tests, so that s just a prove that this bench is not a real one,
just a marketing tool on Intel s behalf.

CPU-Z benefits from FP latency/throughput and big schedulers/reorder buffers, which are greatly improved with Zen5. Therefore I would expect CPU-Z IPC increase to be quite large.
 

Abwx

Lifer
Apr 2, 2011
11,163
3,859
136
CPU-Z benefits from FP latency/throughput and big schedulers/reorder buffers, which are greatly improved with Zen5. Therefore I would expect CPU-Z IPC increase to be quite large.

There s no FP bench where RPL has better IPC than Zen 4 excepted in the ICC compiled Cinebench, so why did CPU Z IPC increase only by 1% in Zen 4 when all other FP benches were at least at 10% ?

Beside Zen 4 had even better IPC improvement in INT code than in FP, so this has nothing to do with either FP or INT.

This has likely do with some instructions not being used on AMD, same as Povray
that gimp AMD by 16-18% since it use AVX2 only with Intel CPUs.
 

Kepler_L2

Senior member
Sep 6, 2020
459
1,891
106
There s no FP bench where RPL has better IPC than Zen 4 excepted in the ICC compiled Cinebench, so why did CPU Z IPC increase only by 1% in Zen 4 ?

This has likely do with some instruction not being used on AMD, same as Povray
that gimp AMD by 16-18% since it use AVX2 only with Intel CPUs.
It uses SSE in both Intel and AMD.
 

Abwx

Lifer
Apr 2, 2011
11,163
3,859
136
It uses SSE in both Intel and AMD.

As said it was updated in 2017 after Zen 1 performed too well for their liking, and when it comes to SSE - SSE4.2 AMD implementation is very good, just look at CB R15 wich use up to SSE 4.2 insructions.

I think you missed the link i posted in my previous post :

 

Geddagod

Golden Member
Dec 28, 2021
1,202
1,166
106
There s no FP bench where RPL has better IPC than Zen 4 excepted in the ICC compiled Cinebench, so why did CPU Z IPC increase only by 1% in Zen 4 when all other FP benches were at least at 10% ?

Beside Zen 4 had even better IPC improvement in INT code than in FP, so this has nothing to do with either FP or INT.

This has likely do with some instructions not being used on AMD, same as Povray
that gimp AMD by 16-18% since it use AVX2 only with Intel CPUs.



🤡 🤡🤡
 

Abwx

Lifer
Apr 2, 2011
11,163
3,859
136
Reactions: spursindonesia

gdansk

Platinum Member
Feb 8, 2011
2,478
3,373
136
Zen 5 CPU-Z single core benchmark is supposedly 910. Will share source soon.


EDIT: Sorry,, I put the twitter link below, had to get to a desktop. I wanted to link the original baidu post, but have had issues.

That is for the engineering sample of Zen 5. For reference I get around 740-750 on my stock 7950X, so about 21% higher. Not bad.

Oh and if this was already posted, apologies. Sometimes this thread moves quite quickly and it can be hard to keep up.
I think it's real (because it aligns well with what I guessed lol)
 

Geddagod

Golden Member
Dec 28, 2021
1,202
1,166
106
There s Povray that is included in Spec_FP, and as said it doesnt use AVX2
for AMD, so much for the clowning, i guess that you didnt even notice that this
bench was in SPEC.

FTR Zen 4 perform 16-18% better in Povray once AVX 2 is enabled for this CPU.
Ah yes, that one subtest explains the 15% difference found by geekerwan lol.
I think it's real (because it aligns well with what I guessed lol)
Based
 

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,227
1,660
136
CPU-Z isn’t the best benchmark to begin with, but for what it’s worth…
View attachment 99838
Part of the reason AMD is disproportionately slower than intel in the CPU-Z ST test is that on AMD Ryzen CPU's, it forces CPU 0 always. Even if you try to change that, it forces it again.

On intel it runs ST on the fastest core automatically. So unless your fastest core on Ryzen happens to be core 0, you're getting a score that is potentially 5-10% slower than it should be.

What this means is even comparing Ryzen vs. Ryzen the ST score isn't comparable, because every chip has different fastest cores. Some samples will score higher due to fastest core being core 0 while others will score poorly.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |