RapydMark CPU benchmark

[igor_kavinski]

RapydMark for Windows (frikiscape.com)

Synthetic highly parallel benchmark.




My company's dual socket Ivy Bridge-E beat Comet Lake in total time, only due to more cores. Should be fun to see how 5950X and i9-12900K fare in this benchmark. My guess is that 5950X will win.

Update: 5950X did win! But i9-12900K put up a good fight despite fewer cores.

Performance comparison chart so far thanks to Makaveli:



Public view-only link for the sheet: RapydMark64 - Google Drive

CCD comparison by Det0x:

1 CCD -> 8 cores / 8 threads = 368.309 seconds
1 CCD -> 8 cores / 16 threads = 230.797 seconds
2 CCD's -> 16 cores / 32 threads = 135.978 seconds

Interesting HT/no-HT comparison:

User​	CPU​	Cores​	Threads​	Time​
JoeRambo​	i9-12900K​	8​	16​	212.656​
JoeRambo​	5800X​	8​	16​	263.424​
Det0x​	5950X​	8​	16​	230.797​
Det0x​	5950X​	8​	8​	368.309​
JoeRambo​	i9-12900K​	8​	8​	328.88​

 

[igor_kavinski]

That Xeon result on Win Server 2019 got me very suspicious (never seen the Hash bench run so fast before) so I installed that OS on my 245KF and it does 184 seconds with stock clocks! Changed the power profile to high performance and completion time lowered to 170 seconds!

Thing of particular note is, there's no extra drivers installed. No internet connection and Arrow Lake is flaming fast out of the box. Going to install .Net 4.8 runtime on it now to try and see how fast I can get it to run with an actual overclock.

 

[igor_kavinski]

Well, dotnet 4.8 is a no go on server OS. May work with internet but not in the mood to contaminate my fresh installation.

Went with OC through the BIOS. Managed to get it down to 163 seconds. Not bad for a simple OS change improving the best score from 183 to 163 seconds.

@Det0x I think Win Server 2019 is calling you

P.S. install it on a separate partition and it will dual boot. Then just benchmark without installing any drivers or connecting to internet. Wanna hear how your experience goes.

 

[Micrornd]

Not bad for a simple OS change improving the best score from 183 to 163 seconds.

Have you tried Windows Pro for Workstations, I think you'll see some improvement there as the CPU scheduler is tweaked/optimized for a workstation (single to quad CPU) environment.

 

[igor_kavinski]

Have you tried Windows Pro for Workstations

I have it installed on my Zen 2 Epyc and that CPU seems to struggle hard with the Hash test. Gonna put Server 2019 on that when I get the time to see if things improve for it.

 

[igor_kavinski]

That's the best I can get out of the 245KF without trying to kill it.

 

[samuelmar]

Something VERY interesting:

View attachment 115843

Ice Lake Xeon having trouble with the Hash test too.

BUT

Crazy max performance unlocked with 7-Max !!!

View attachment 115845



Roughly 17% improvement in calculation time.

@Micrornd , please try 7-Max too.

Amazing and awesome to see the 7-Max boost in action.

 

[igor_kavinski]

View attachment 116158

Ma bootiful Arro Lack outdone by my old 12700K running in Windows Server 2025!





So annoying!

 

[igor_kavinski]

Your 64 core does not beat that ??

Example showing how Zen 2 server performance is nerfed in the latest Windows Server 2025 (based on the Win11 kernel):






37% increase just from using an OS kernel that properly supports Zen 2!

 

[ArrogantHair]

285K P56E50. This is with my P Cores kinda tuned.

 

[ArrogantHair]

I also just upgrade my 5800x server to a 7950x - also upgraded to Windows Server 2025 from 2019. I should have run before.
Had to get the 7950x as a bundle to get my 5090. I am so poor now. lol.

 

[igor_kavinski]

View attachment 121007

You are really close!

Hash test score is about right. Maybe run from a RAMdrive and you could get down to the annoyingly elusive 99 seconds.

 

[ArrogantHair]

Error on the File Encrypt. Not sure why. Really hates the hash.

 

[igor_kavinski]

Error on the File Encrypt. Not sure why. Really hates the hash.

The file encrypt test fails sometimes. That's normal. I hate the Hash test too

 

[ArrogantHair]

The file encrypt test fails sometimes. That's normal. I hate the Hash test too

What CPU is your AMD Engineering sample?

 

[Det0x]

could you record Rapydmark running on your 9950X3D using the Win11 Snipping tool or any other screen recorder? I just like seeing Rapydmark progress bars whizz by quickly and I don't have any CPU that can do anything close to 100 secs or less. Hope you can post such a video in the Rapydmark thread or give me a cloud link and I will upload to my Youtube channel and share the link in the Rapydmark thread.

Low effort PBO run, straight from boot into running Rapydmark (no trickery with 7-max/ram drives etc)

• 9950X3D @ custom watercooling
• Asus X870E APEX
• G.SKILL 6000MT/S CL26 @ 8700MT/s CL32 -> 1.74VDD / 1.55 VDDQ @ Air cooling + custom heatsinks
• 2200mhz FCLK

Screenshot


But i dont understand why your so fixated on this benchmark.. 😅

In my opinion its not a good comparison/benchmark, as you have large run to run variation on the subtests which will influence the end score/time greatly, together with "outdated" instructions sets (?)
Dont think X3D sees any benefit in this benchmark either, and i dont think there are many that use it in the first place

 

[Det0x]

What CPU is your AMD Engineering sample?

If you talk about this:

Then its my 9950X ES
*edit*
Cache for the lose right ?

 

[ArrogantHair]

Low effort PBO run, straight from boot into running Rapydmark (no trickery with 7-max/ram drives etc)

• 9950X3D @ custom watercooling
• Asus X870E APEX
• G.SKILL 6000MT/S CL26 @ 8700MT/s CL32 -> 1.74VDD / 1.55 VDDQ @ Air cooling + custom heatsinks
• 2200mhz FCLK

Screenshot


But i dont understand why your so fixated on this benchmark.. 😅

In my opinion its not a good comparison/benchmark, as you have large run to run variation on the subtests which will influence the end score/time greatly, together with "outdated" instructions sets (?)
Dont think X3D sees any benefit in this benchmark either, and i dont think there are many that use it in the first place

Cache for the win. Nice!

 

[ArrogantHair]

Its my 9950X ES

I meant the 64 core one that Igor shared

 

[igor_kavinski]

I meant the 64 core one that Igor shared

It's got a really long ES name. Checking in CPU-Z, it shows it's a Threadripper which is obviously wrong since it's not a Threadripper mobo or even socket. SP3 so it's an Epyc mobo. Supposed to be ES version of this: https://www.amd.com/en/support/down...sors/epyc/epyc-7002-series/amd-epyc-7662.html

 

[igor_kavinski]

But i dont understand why your so fixated on this benchmark.. 😅

The super annoying kid inside me loves it! He's the one who makes me do and love totally irrational stuff that an adult would never approve of

I think you could improve the score because the Fill Memory test is running a bit slow at over 7 seconds. If it completes in 4 seconds or less, you can easily beat the 100 seconds barrier no problem.

Thanks for the video! It's especially nice to watch something that can't be observed under normal circumstances. Hopefully Zen 6 and Nova Lake will make this benchmark totally useless and it will complete so fast that I will lose interest in it

 

[ArrogantHair]

It's got a really long ES name. Checking in CPU-Z, it shows it's a Threadripper which is obviously wrong since it's not a Threadripper mobo or even socket. SP3 so it's an Epyc mobo. Supposed to be ES version of this: https://www.amd.com/en/support/down...sors/epyc/epyc-7002-series/amd-epyc-7662.html

If you talk about this:
View attachment 121028
Then its my 9950X ES
*edit*
Cache for the lose right ?

I missed that there were two in the thread. I guess it is more of the encrypt/decrypt and other tests that favour the architecture. Under 100 is a Port Royal Hall of Fame score.