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​

 

[.vodka]

Highly tuned 3900X (PBO Fmax enhancer) + memory (4x16GB DDR4-3800 16-19-19-38-57 1T+GDM, as tight secondaries and tertiaries as Rev. E will do)

Zen3 is quite the improvement here.

 

[Det0x]

Quick look at the results, seems like there is a large difference in "fill memory" numbers.. Is this down to 16GB vs 32GB ?

View attachment 61611 vs View attachment 61612

Memory amount dont seem to affect numbers.. tested with maxmem:


@JoeRambo
MLC for you

 

[PJVol]

Can that tool disable AMD's hardware prefetchers too?

Not sure were you interested in how L2 hardware prefetcher affects the results, but anyway, here is the answer:
no, your benchmark score is not affected. Only AIDA latency is vulnerable.
(Should have noted, I can disable and enable it from windows)
For @JoeRambo - added Sandra Cache&Memory Latency test.

 

[Makaveli]

Quick look at the results, seems like there is a large difference in "fill memory" numbers.. Is this down to 16GB vs 32GB ?

View attachment 61611 vs View attachment 61612

Interesting find also wonder if its related to the difference in memory write speed.

All standard single CCD Zen 3 chips are reading 32 bytes and writing 16 bytes on each clock cycle.

Dual CCD Zen 3 can do writes 32 bytes

Don't believe the 5700G has the same limitation.

 

[PJVol]

Don't believe the 5700G has the same limitation.

It hasn't, cause it's the limitation of a single IFOP CCM->IOMS link in Zen2/3 desktop parts.

 

[YBS1]

Anyone have any idea why the Hash SHA1/256/512 MD4/5 CRC32 RIPEMD160 test can vary so wildly between systems that are otherwise roughly comparable? For instance .vodka's 3900X took 51.458 to complete that test, my 3900X completes it in about 40 seconds flat. Even more extreme is Det0x's 5950X......About 12 seconds flat, meanwhile Markfw's 5950X takes ~60 seconds, I think there was also a Ryzen 7 5800H that did it in about ~17 seconds. If you look through the results submitted so far you can see a number of systems blaze through that one test and all the other results are way behind them. There must be a BIOS or windows setting that is having a tremendous effect on this.

 

[Makaveli]

Anyone have any idea why the Hash SHA1/256/512 MD4/5 CRC32 RIPEMD160 test can vary so wildly between systems that are otherwise roughly comparable? For instance .vodka's 3900X took 51.458 to complete that test, my 3900X completes it in about 40 seconds flat. Even more extreme is Det0x's 5950X......About 12 seconds flat, meanwhile Markfw's 5950X takes ~60 seconds, I think there was also a Ryzen 7 5800H that did it in about ~17 seconds. If you look through the results submitted so far you can see a number of systems blaze through that one test and all the other results are way behind them. There must be a BIOS or windows setting that is having a tremendous effect on this.

I think its related to how the systems are tuned.

This is my score

 

[.vodka]

Anyone have any idea why the Hash SHA1/256/512 MD4/5 CRC32 RIPEMD160 test can vary so wildly between systems that are otherwise roughly comparable?

Now that you mentioned it, the results in my other post were done with sz's Ryzen balanced v4 power plan. Zen2 is very sensitive to power plans, that one nets snappy performance with lower temperatures on day to day usage. Some workloads do suffer a performance hit.

This is how it behaves with AMD's Ryzen Balanced plan.

That's more like it. (for reference, here's a quick CPU-z benchmark with this CPU as it is right now on the same power plan)

I don't know what could contribute to Zen3 results variance here, as it's not power plan sensitive (at least not as Zen2 is)

 

[Markfw]

If you notice the stress level also affects it. I did medium and it was 3.54 seconds.

 

[scannall]

Ok, Ryzen 6900HX in a laptop.

 

[igor_kavinski]

Ok, Ryzen 6900HX in a laptop.

View attachment 67688

Wow. That's 5600X level MT performance in a laptop!

I'm assuming this was done with Balanced profile. Can you do High performance (or better, create new profile and choose Ultimate performance) and re-run to see how much the score improves?

 

[scannall]

Wow. That's 5600X level MT performance in a laptop!

I'm assuming this was done with Balanced profile. Can you do High performance (or better, create new profile and choose Ultimate performance) and re-run to see how much the score improves?

Sure, here ya go.

 

[igor_kavinski]

Sure, here ya go.

View attachment 67698View attachment 67698

High performance really lets it loose!



Score of 277 seconds puts it closer to the 5800X!

 

[scannall]

High performance really lets it loose!

View attachment 67700

Score of 277 seconds puts it closer to the 5800X!

Not bad for a 45 watt CPU with laptop cooling eh?

 

[thigobr]

My results running on a Zenbook S 13, 6800U, 16GB LPDDR5 (Performance mode sustained 18W according to HWinfo) : 355.008s

 

[igor_kavinski]

Something changed in CPU-Z? Why is it showing CL19 for DDR5? Maybe an older version?

 

[igor_kavinski]

My results running on a Zenbook S 13, 6800U, 16GB LPDDR5 (Performance mode sustained 18W according to HWinfo) : 355.008s

Great score. Beats the 5800H due to DDR5. 11% improvement. Poor Zen 3 has been bandwidth starved since its launch.

 

[thigobr]

Something changed in CPU-Z? Why is it showing CL19 for DDR5? Maybe an older version?

It's really weird but HWiNFO also reads the same... It's probably because this machine is using LPDDR5 6400 instead of regular DDR5

 

[igor_kavinski]

Maybe we need to double those figures so it's running at 38-30-34-68? If correct, CL is just OK but the sub timings are great.

 

[Det0x]

Don't seem like there is so much improvement for me in this benchmark..
PBO CO auto clocking 7950x

Could optimize further but other testing to do

 

[igor_kavinski]

Don't seem like there is so much improvement for me in this benchmark..
PBO CO auto clocking 7950x
View attachment 68686
Could optimize further but other testing to do

Thanks! I was actually gonna ask you in this thread for your score

17% improvement is in line with the expected IPC improvement over Zen 3.

What do the clocks and temps look like in this benchmark for the various cores?

 

[eek2121]

7950X -- Stress Level High, 240mm AIO, 64GB DDR5 6000 CL32 - 171.285S for all tests. (Version 1.2a)

EDIT: Excluding the graphics test that would be 82.547 seconds.

 

[igor_kavinski]

7950X -- Stress Level High, 240mm AIO, 64GB DDR5 6000 CL32 - 171.285S for all tests. (Version 1.2a)

Page 4 - RapydMark CPU benchmark

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

24% improvement over Markfw's 5950X. Good.

Looks like it's not easy to touch Det0x's time

 

[Markfw]

My first run with the new 6000 cl30 memory was disappointing at 138.559, so not going to bother to post. Not sure whats up.

 

[igor_kavinski]

My first run with the new 6000 cl30 memory was disappointing at 138.559, so not going to bother to post. Not sure whats up.

That's still a great score compared to your 5950X.