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

[DisEnchantment]

Speculate at will

 

[Nothingness]

side note: Postgres added AVX-512 support to some areas and managed more than a 6x performance increase in some cases. SQL databases can benefit significantly by utilizing AVX-512.

Side note: the speedup is for CRC32

Do you have anything else showing the benefit of AVX-512 on Postgres?

 

[Josh128]

Guys, have you seen this? This is some next level smoke and mirrors from AMD regarding Ryzen 9000 gaming testing, lol.

AMD promises Windows 11 patch boosting Zen5, Zen 4, and Zen 3 gaming performance - VideoCardz.com

AMD clarifies discrepancies in gaming performance between official data and reviews The company says that Windows patch will improve gaming performance, explains why ‘admin’ account was used for internal testing. AMD published a blog post discussing the disparity between the official gaming...

videocardz.com

 

[DisEnchantment]

Guys, have you seen this? This is some next level smoke and mirrors from AMD regarding Ryzen 9000 gaming testing, lol.

Fine wine aging so soon?

Another round of Testing marathon from tech tubers, content for weeks. Subscribe to patreon.

 

[jdubs03]

It still isn’t gonna really change the delta that much from Zen 4 to Zen 5. Everything gets a boost.

(Though AMD does say there is a 5–8% bump over the 7000 series, but count me skeptical)

 

[Hitman928]

Fine wine aging so soon?

Another round of Testing marathon from tech tubers, content for weeks. Subscribe to patreon.

Small bump in gaming performance over Zen 4 from new Windows build if true.

Their gaming numbers compared to RPL only look good because they used the baseline profile for the 14900k/14700k, which is not recommended by Intel.

 

[CouncilorIrissa]

Wendell's take.

Some of the performance swings on Linux compared to Windows are just wild lol

 

[Abwx]

Small bump in gaming performance over Zen 4 from new Windows build if true.

Their gaming numbers compared to RPL only look good because they used the baseline profile for the 14900k/14700k, which is not recommended by Intel.

Baseline profile doesnt change the gaming perfs, there s enough TDP left since games do not exhaust the 253W limit by far, this happen only in MT apps.

 

[Hitman928]

Baseline profile doesnt change the gaming perfs, there s enough TDP left since games do not exhaust the 253W limit by far, this happen only in MT apps.

That's just simply not true. Maybe some games don't see a difference, but in general, there is a noticeable decrease in performance.



*I edited the HWUB slide because this was done before Intel clarified exactly what the baseline profile was supposed to be and so only the Gigabyte board here is actually the baseline profile.

 

[Abwx]

That's just simply not true. Maybe some games don't see a difference, but in general, there is a noticeable decrease in performance.

View attachment 105936
View attachment 105937
*I edited the HWUB slide because this was done before Intel clarified exactly what the baseline profile was supposed to be and so only the Gigabyte board here is actually the baseline profile.

Even if there s a difference that s still Intel s real perfs, i mean previously it was an exageratly ocked CPU whose perfs werent sustainable on the long run withhout degradations, so AMD is not at fault here, that s rather Intel that was setting their CPU outside of a safe operating area.

 

[Hitman928]

Even if there s a difference that s still Intel s real perfs, i mean previously it was an exageratly ocked CPU whose perfs werent sustainable on the long run, so AMD is not at fault here, that s rather Intel that was setting their CPU outside of a safe operating area.

For the CPUs in question, Intel Baseline profile is not what Intel recommends using, they are supposed to use Intel Performance profile. Intel CPUs being unstable is a separate issue.

 

[jdubs03]

HWUB going in on AMD:

https://twitter.com/x/status/1826405081119687149

https://twitter.com/x/status/1826402112726859849

 

[Abwx]

For the CPUs in question, Intel Baseline profile is not what Intel recommends using, they are supposed to use Intel Performance profile. Intel CPUs being unstable is a separate issue.

They are not very clear on the subject, and for the reasons we know.
Igor s Lab stated on may that baseline profile is Intel s default setting, and he provided Intel recomendation on the matter, in some way they want their cake while still eating it.

Intel® requests system and motherboard manufacturers to provide end users with a default BIOS profile that matches Intel®
recommended settings.
– Suggested profile name “Intel Default Settings”.
– Intel requests customers to implement the “Intel Default Settings” profile as the BIOS default profile by May 31, 2024.

Intel plays with the name and the data: The “Intel Baseline Profile” becomes “Intel Default Settings” | igor´sLAB

The problem with the Intel desktop processors of the 13th and 14th generation, i.e. Raptor Lake-S and Raptor Lake-S-Refresh, is not getting any smaller and the bad news does not stop. At this point…

www.igorslab.de

 

[Hitman928]

They are not very clear on the subject, and for the reasons we know.
Igor s Lab stated on may that baseline profile is Intel s default setting, and he provided Intel recomendation on the matter, in some way they want their cake while still eating it.

Intel plays with the name and the data: The “Intel Baseline Profile” becomes “Intel Default Settings” | igor´sLAB

The problem with the Intel desktop processors of the 13th and 14th generation, i.e. Raptor Lake-S and Raptor Lake-S-Refresh, is not getting any smaller and the bad news does not stop. At this point…

www.igorslab.de

Default settings does not equal baseline profile. You should read the whole release and post in the Intel thread if you’re still not clear about it.

 

[positivedoppler]

For the CPUs in question, Intel Baseline profile is not what Intel recommends using, they are supposed to use Intel Performance profile. Intel CPUs being unstable is a separate issue.

which one is the fastest cpu will not slow cook to death profile

 

[Markfw]

which one is the fastest cpu will not slow cook to death profile

As far as I am concerned the 12900k is the fastest Intel that will not die quickly, and the 9950x is the king in everything by far, except possibly gaming, and there its close until the 9800x3d comes out. The Intel supporters won't give up until they all cook their CPUs, but I have quit listening, its their loss.

And the avx-512 solution of the 9950x is just phenomenal, and quite a few apps do support it.

 

[Abwx]

Default settings does not equal baseline profile. You should read the whole release and post in the Intel thread if you’re still not clear about it.

IIRC AMD used the default settings in their comparisons, so they just did what was requested by Intel from customers, truth is that Intel ask the reviewers to cheat in its favour by not enabling the very settings they are requesting users to implement.

Here we have a situation that is as if AMD would recommend to the reviewers to use PBO when benching the 9950X, but still asking them to display the standard 200W PPT in their charts for the chip power comsumption for those tests.

 

[Hitman928]

IIRC AMD used the default settings in their comparisons, so they just did what was requested by Intel from customers, truth is that Intel ask the reviewers to cheat in its favour by not enabling the very settings they are requesting users to implement.

Here we have a situation that is as if AMD would recommend to the reviewers to use PBO when benching the 9950X, but still asking them to display the standard 200W PPT in their charts for the chip power comsumption for those tests.

Again, this is not accurate. You should take this to the Intel thread if you want to understand Intel’s new recommended settings. The takeaway, though, is that AMD did not use the recommended settings for Intel in their comparison.

 

[Hitman928]

which one is the fastest cpu will not slow cook to death profile

TBD with the new microcode.

 

[Abwx]

Again, this is not accurate. You should take this to the Intel thread if you want to understand Intel’s new recommended settings. The takeaway, though, is that AMD did not use the recommended settings for Intel in their comparison.

To the contrary, it belong to this thread since we re talking of AMD s Zen 5 benching methodology, and FTR their own default setting is PPT = 200W and of couse no PBO enabled.

 

[Hitman928]

To the contrary, it belong to this thread since we re talking of AMD s Zen 5 benching methodology, and FTR their own default setting is PPT = 200W and of couse no PBO enabled.

I’m not talking about AMD’s settings, I’m saying this isn’t the thread to analyze Intel’s default settings. That’s already been done in the other thread and if someone isn’t clear on it, the Intel thread is the place to discuss it.

 

[Abwx]

I’m not talking about AMD’s settings, I’m saying this isn’t the thread to analyze Intel’s default settings. That’s already been done in the other thread and if someone isn’t clear on it, the Intel thread is the place to discuss it.

The other thread is not fit to dicuss about Zen 5 perfs in respect of Intel, moreover of AMD s metrics regarding Intel s settings, the point is that AMD comparisons make sense in repect of the competitor s requested settings, that s the debate that is highlighted in my posts, i wont bother about Intel s variable geometry recomendations to reviewers, that s why i posted AMD s methodology slide for the purpose

 

[RnR_au]

Wendell's take.

Some of the performance swings on Linux compared to Windows are just wild lol

The money shot I guess;

 

[jdubs03]

The money shot I guess;

View attachment 105945

Would be nice to see the same testing done with the 7700X. Can’t really see the full picture of fps increases without a 1 for 1 comparison.

 

[Thunder 57]

The money shot I guess;

View attachment 105945

Maybe there are tangible improvments to be had in the future.

 

[Hitman928]

The other thread is not fit to dicuss about Zen 5 perfs in respect of Intel, moreover of AMD s metrics regarding Intel s settings, the point is that AMD comparisons make sense in repect of the competitor s requested settings, that s the debate that is highlighted in my posts, i wont bother about Intel s variable geometry recomendations to reviewers, that s why i posted AMD s methodology slide for the purpose

That’s fine, it just means AMD is purposefully handicapping Intel in their comparison.