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: https://forums.anandtech.com/threads/rapydmark-cpu-benchmark.2599240/post-41425508

The current champ!

 

[clemsyn]

On high settings 🙂 CPU undervolted and default settings. Memory increased to 5000 from 4800 and no Cathedral 3d

 

[Det0x]

EDIT: putting results of 8C vs 8C side by side i don't even know what to think about this benchmark and whatever virtues it has:

Maybe its better to use my 8 core numbers from the post right above yours as i saved both upper and lower numbers

*edit*
Nevermind, i will do the comparison myself 🙂

Alder lake @ 5 ghz (?) vs Zen3 @ 4.85 ghz

 

[Makaveli]

View attachment 53557

The screenshot shows you are still on Medium for the Stress level you need to change that to High

 

[JoeRambo]

Maybe its better to use my 8 core numbers from the post right above yours as i saved both upper and lower numbers 🙂

16C vs 8C is another variable. 8C vs 8C is good to compare "peak IPC". But hard to compare results when some lines just don't make sense:

For example
Matrix Multiplication is 2.4s vs 9.4s
Mem cache performance 28.2s vs 11.6s

I find it hard to believe ADL is 4x worse at matrix multiplication, that usually is about memory subsystem and then proceeds to beat 2.5x is said mem/cache performance.

We might need 8C without HT perf point, to see what really is going on.

 

[Ed1]

Duh, ok here high mode (12600k XMP1 3600@c16)

 

[Makaveli]

Based on the two 12600k results I see one with DDR4 3600 vs DDR5 5000

This benchmark does benefit from the extra bandwidth.

 

[clemsyn]

Based on the two 12600k results I see one with DDR4 3600 vs DDR5 5000

This benchmark does benefit from the extra bandwidth.

Interesting, and its over 34 seconds.

 

[Makaveli]

@igor_kavinski

i'm creating an Excel speadsheet with a graph to track all of these, which I wouldn't mind sending to you so you can add it the first post. The only thing is I would like people to most their memory speeds since it seems to affect this benchmark. For now just like DDR4 3200 don't need CL numbers.

 

[JoeRambo]

Alder lake @ 5 ghz (?) vs Zen3 @ 4.85 ghz

Even more ridiculous Matrix Multiplication difference 🙂

 

[Ed1]

Maybe we can compare specific memory tests. I pasted those below with memory in the test name.
Also, remember I am still using fresh Win10 but looking at core utilization all looks good with this test.

Test Type: Fill Memory 256 MB
Iterations: 60K | Performance: 1,552 it/s (16th) | Time: 38.636 s

-------------------------------------
Test Type: Copy Memory 256 MB
Iterations: 60K | Performance: 2,683 it/s (16th) | Time: 22.358 s

-------------------------------------
Test Type: Memory cache performance (max 4 threads)
Iterations: 40 | Performance: 11 it/s (4th) | Time: 13.936 s
Last result: SizeMem= 8 kB Iterations= 65536 Time= 2 ms
SizeMem= 16 kB Iterations= 32768 Time= 1 ms
SizeMem= 32 kB Iterations= 16384 Time= 1 ms
SizeMem= 64 kB Iterations= 8192 Time= 7 ms
SizeMem= 128 kB Iterations= 4096 Time= 7 ms
SizeMem= 256 kB Iterations= 2048 Time= 7 ms
SizeMem= 512 kB Iterations= 1024 Time= 7 ms
SizeMem= 1 MB Iterations= 512 Time= 8 ms
SizeMem= 2 MB Iterations= 256 Time= 14 ms
SizeMem= 4 MB Iterations= 128 Time= 15 ms
SizeMem= 8 MB Iterations= 64 Time= 18 ms
SizeMem= 16 MB Iterations= 32 Time= 47 ms
SizeMem= 32 MB Iterations= 16 Time= 59 ms
SizeMem= 64 MB Iterations= 8 Time= 64 ms
SizeMem= 128 MB Iterations= 4 Time= 78 ms

Test Type: Memory Encrypt/Decrypt AES256 (1 MB)
Iterations: 8K | Performance: 870 it/s (16th) | Time: 9.188 s

-------------------------------------
Test Type: Memory Encrypt/Decrypt Blowfish448 (1 MB)
Iterations: 8K | Performance: 1,290 it/s (16th) | Time: 6.200 s

-------------------------------------
Test Type: Memory Compress/Decompress 1 MB (Gzip)
Iterations: 8K | Performance: 1,044 it/s (16th) | Time: 7.656 s

 

[Markfw]

Even more ridiculous Matrix Multiplication difference 🙂

By ridiculous, do you mean you don't think its right ? I do know that CPU is really fast in primegrid (calculating primes)

 

[igor_kavinski]

@igor_kavinski

i'm creating an Excel speadsheet with a graph to track all of these, which I wouldn't mind sending to you so you can add it the first post. The only thing is I would like people to most their memory speeds since it seems to affect this benchmark. For now just like DDR4 3200 don't need CL numbers.

That's a great idea! Thanks 🙂

 

[nicalandia]

View attachment 53557

Can you get it on HIGH?

 

[Det0x]

16C vs 8C is another variable. 8C vs 8C is good to compare "peak IPC". But hard to compare results when some lines just don't make sense:

For example
Matrix Multiplication is 2.4s vs 9.4s
Mem cache performance 28.2s vs 11.6s

I find it hard to believe ADL is 4x worse at matrix multiplication, that usually is about memory subsystem and then proceeds to beat 2.5x is said mem/cache performance.

We might need 8C without HT perf point, to see what really is going on.

Here you go, i'm down to 8 out of 32 threads now 😛




Same mem settings as before, 1900:3800 CL13 tuned
8 threads 4975mhz

 

[Ed1]

Can you get it on HIGH?

I did, it's 8 posts up from yours 🙂

Ed

 

[igor_kavinski]

Mem cache performance 28.2s vs 11.6s

I find it hard to believe ADL is 4x worse at matrix multiplication, that usually is about memory subsystem and then proceeds to beat 2.5x is said mem/cache performance.

That mem/cache performance may explain why you find the ADL snappier than your 5950X.

 

[clemsyn]

The one to the left is memory speed at default (4800) but with aggressive timings (36-36-36-72) while the one on the right is overclocked to 5400 default timings.

 

[Ed1]

I ran the test another time, this time with Asus default "optimized" limits setting before I had it set to Intel limits 150w. even though it never passed 125w (well below the Intel 150w) the score increased (clocks all same as before)

 

[Makaveli]

Sample so far.

 

[igor_kavinski]

Sample so far.

View attachment 53581

That i3 shouldn't be there. It was done on Medium. Or you could add High or Medium in parentheses.

 

[Makaveli]

That i3 shouldn't be there. It was done on Medium. Or you could add High or Medium in parentheses.

You are right I forgot to remove it, I've updated the chart. I'd rather just keep all the high results and not medium.

Also what is the best way to deal with the excel spreadsheet?

i've seen people have them editable online and accessible to everyone but i've never done that any suggestions?

 

[JoeRambo]

Here you go, i'm down to 8 out of 32 threads now 😛
Same mem settings as before, 1900:3800 CL13 tuned
8 threads 4975mhz

I also did 8 no HT test, even went to CL13, so we can do as Apples to Apples now, so pretty much 5Ghz vs 4.975Ghz for AMD





So 329s vs 368s with no HT.

Also the mysteries with Matrix Multiplication and Memory Cache performance remain, should not be THAT different.

 

[igor_kavinski]

Also what is the best way to deal with the excel spreadsheet?

i've seen people have them editable online and accessible to everyone but i've never done that any suggestions?

That's all well and good until people editing stuff left and right and it goes out of control 😀

Maybe one chart with the best score of each CPU+Mem type.

Another chart for all other scores done with different memory speeds/timings etc.

 

[Det0x]

Sample so far.

View attachment 53582

Nicely done 🙂
My best 5950x time with all threads enabled are 135.978 seconds

1 CCD -> 8 cores / 16 threads = 230.797 seconds


1 CCD -> 8 cores / 8 threads = 368.309 seconds

 

[Makaveli]

That's all well and good until people editing stuff left and right and it goes out of control 😀

Maybe one chart with the best score of each CPU+Mem type.

Another chart for all other scores done with different memory speeds/timings etc.

I was thinking about that think I can host it on google and only have modify access given to you and myself.

Those other suggestions are good aswell but I won't have time to dig into multiple charts as i'm at work right now 🙂 that is something however you can run with.