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​

Update (17 APRIL 2025): Det0x's video of his 9950X3D benchmark run: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/rapydmark-cpu-benchmark.2599240/post-41425508

The current champ!

 

[igor_kavinski]

Really good score!

Makes me wish you would try it on Win Server 2025 with a ramdisk to see if you can get under 100 seconds

 

[igor_kavinski]

I have 56,56,58,58,56,56,57,57 as my daily, but the 2 cores at 58 crash the Memory Encrypt/Decrypt AES256 test. No issues with Karhu or OCCT. Not sure where the instability is. Will test.

I had that problem with my 9950X3D when I had curve offset set at -37. Kept increasing until I got to -29 and then that test passed. I think maybe the 58 capable cores need a slight voltage boost to pass this test.

 

[ArrogantHair]

Hmm. Will try that - currently I have the 58 @ 1.425v but offset -0.030v or -0.045v depending on bench. The 1.425v is from Skatterbencher, but may be too high/low depending with offset. The default auto voltage was 1.398v.

 

[ArrogantHair]

Makes me wish you would try it on Win Server 2025 with a ramdisk to see if you can get under 100 seconds

If I can find time lol.

 

[igor_kavinski]

Hmm. Will try that - currently I have the 58 @ 1.425v but offset -0.030v or -0.045v depending on bench.

Probably the negative offset is the one that test hates.

 

[igor_kavinski]

If I can find time lol.

Let me try to motivate you a bit by saying that if you've been exposed to Win10/11 too much and never tried a server OS before, the first thing you will notice is that it "feels" VERY responsive. Moving the mouse around, clicking the Start menu and launching programs, it feels like some burden's been lifted from your PC and everything is working effortlessly. I prefer Win 2019 to Win 2025 but sadly, Ryzen seems to like Win 2025 better.

 

[igor_kavinski]

RapydMark Benchmark Competition

1) Download here: RapydMark for Windows (frikiscape.com) 2) Switch stress level to HIGH 3) Select and run all benchmark tests 4) Listed by TOTAL TIME is how we will score the results 4) Post your results in a format like this:

www.overclock.net

Custom cooled delidded 285K nabbed the top spot.

I dunno. Seems like outside of LN2, this score will be pretty hard to beat.

 

[Det0x]

No LN2 needed

btw X3D compared to regular Zen5 sees no benefit, so seeming this benchmark does not care about memory bandwidth/latency too much


Still on custom watercooling with a mora 👍

 

[igor_kavinski]

And you know what? You still have room to improve because I can sometimes get less than 12 seconds on my 9950X3D in the Hash test on Win Server 2025!

But probably some other test will regress so possibly you may not see a net improvement.

EDIT: It's funny that few months ago, going below 100 seconds seemed pretty daunting and now within a few days apart, the goal has been shifted to breaking the 90 seconds barrier!

Like, I don't know but did Det0x chant "YOU CAN DO IT, BABY!" over and over during this run???

 

[igor_kavinski]

btw X3D compared to regular Zen5 sees no benefit, so seeming this benchmark does not care about memory bandwidth/latency too much

It does seem to care a lot about the MCLK/UCLK ratio though. I noticed that high MCLK and half UCLK was better than MCLK=UCLK for this benchmark.

If AMD is able to sustain a UCLK of 5000 for DDR5-10000 with Zen 6, that will really make this benchmark fly!

But a more realistic UCLK 4000 with DDR5-8000 would still be a great improvement.