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

Page 786 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

jdubs03

Golden Member
Oct 1, 2013
1,282
902
136
Maybe there are tangible improvments to be had in the future.
The above image obscures the performance delta between the 9700X and 7700X. We get a limited picture because it pretty much focuses only on the 9700X, there’s just the two token submissions for the 7700X at the bottom.

It does look like there will be tangible improvements across-the-board (including Intel); not just for the 9000 series.

Omitting like for like comparisons just obfuscates the true impact of the targeted changes. It’s quite an unsatisfactory test.

When considering a like for like comparison, the delta is likely going to be similar.

I edited the image a bit to show the different comparisons so it’s easier to see the difference between FPS values:
1724306388271.jpeg
If we’re just considering VBS being on (because you know security), then the performance increase from 7700X to the 9700X is 10.2% for the 23H2 build, and 13.61% for the 24H2 pre-release build (but this excludes a result on the newer build for the 7700X).

Beyond that though it does look like with VBS off, the 9000 series looks like it does see a bump.
 
Last edited:

Det0x

Golden Member
Sep 11, 2014
1,465
4,999
136
Guys, have you seen this? This is some next level smoke and mirrors from AMD regarding Ryzen 9000 gaming testing, lol.

Pretty funny, I have already been running Win11 24H2 for the last two months --> all my ES benches have been on this windows version
Maybe thats why i seem to have higher scores than many other Zen5 users 🧐
 

del42sa

Member
May 28, 2013
179
311
136
Pretty funny, I have already been running Win11 24H2 for the last two months --> all my ES benches have been on this windows version
Maybe thats why i seem to have higher scores than many other Zen5 users 🧐
no one will give you as much as we can promise you :cool:

AMD states that they have seen some issues with the Branch Predictor not working as intended within Windows 11, but company remain committed to it´s claim about ZEN5 performance improvement
so their official test numbers came from what version of WIN11 or was it just made up ? :rolleyes:
 
Last edited:
Jul 27, 2020
28,039
19,144
146
Fine wine aging so soon?
But branch prediction optimizations that will also affect Zen 3 and 4???

Why are those two getting these optimizations so late then?

What could be the nature of these optimizations? Is the Windows scheduler going to give some kind of "hints" to the CPU to enable it to predict better?

An optimization targeting the decode clusters I would've understood but this just sounds ridiculous for AMD to be so late in getting them into Windows.
 
  • Like
Reactions: Josh128

yuri69

Senior member
Jul 16, 2013
677
1,215
136
Guys, have you seen this? This is some next level smoke and mirrors from AMD regarding Ryzen 9000 gaming testing, lol.

We 'member:
 

MS_AT

Senior member
Jul 15, 2024
869
1,765
96
But branch prediction optimizations that will also affect Zen 3 and 4???

Why are those two getting these optimizations so late then?

What could be the nature of these optimizations? Is the Windows scheduler going to give some kind of "hints" to the CPU to enable it to predict better?
That's actually the interesting part of this whole ordeal and not if CS will suddenly do 500FPS from 450 FPS ;) But we will get another batch of youtube videos that will more or less confirm, depending on game selection, that Zen5 is a failure in gaming because it doesn't beat x3d and costs more. Probably good for content creators but puzzling form AMD marketing point of view as it will only reinforce the public opinion of Zen5 being a failure...
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
Lol :
When taking the geometric mean of 59 benchmarks run across all of the Linux distributions on this AMD Ryzen 9 9950X system, Intel's Clear Linux easily took the crown.
Tapping Intel's Clear Linux netted a 16% improvement on top of the performance offered by Ubuntu 24.04 LTS
And re Lol :
Intel's software engineers advancing open-source (x86_64) performance continue doing a splendid job showing what's possible.


 

jdubs03

Golden Member
Oct 1, 2013
1,282
902
136
In case anyone is interested:

That should lessen the disappointment from a delay in release date for the 9000X3D series.
 

tsamolotoff

Senior member
May 19, 2019
256
510
136
Thank you, Intel! :D
It is not really 'free' performance - Clear Linux uses some, lets say, not universally beneficial compilation flags that some of the HPC software can't run with, i.e. -ffast-math and some other optimizations because of the FPE and/or inconsistent results
 
Last edited:
  • Like
Reactions: lightmanek
Jul 27, 2020
28,039
19,144
146
It is not really 'free' performance - Clear Linux uses some, lets say, not universally beneficial compilation flags that some of the HPC software can't run with, i.e. -ffast-math and some other optimizations because of the FPE and/or inconsistent results
Most of us here are interested in the ClearLinux Desktop so HPC software not running is moot for us.
 

MS_AT

Senior member
Jul 15, 2024
869
1,765
96
It is not really 'free' performance - Clear Linux uses some, lets say, not universally beneficial compilation flags that some of the HPC software can't run with, i.e. -ffast-math and some other optimizations because of the FPE and/or inconsistent results
I am surprised they don't compile their own stuff to order, linking everything statically, then there is really nothing from system that should be able to affect them I can think of. Unless some packages are compiled from source ad hoc using system defaults, but well those can be changed;)
 

MS_AT

Senior member
Jul 15, 2024
869
1,765
96