Surprsingly low score for new AMD EPYC2 7502P (32 cores) sonly 1100 ST points.

[user3141]

Hi,

Reading all the good reviews of the new AMD EPYC2 processors I considered buying the 32 core variant 7502P. but when I looked up its passmark score,

https://www.cpubenchmark.net/cpu.php...+7502P&id=3538

I was very surprised to see its preformance rated at a score of only 24000 and 1100 single thread.

From all the other published benchmarks and all theoretical considerations, the performace should be more like 2000 single thread and 35000 total performance for this CPU.

It is also inconsitent by a factor of two when compared to the listing of the corresponding 64 core variant (7742)

https://www.cpubenchmark.net/cpu.php...C+7742&id=3547

which scored more than double the single thread score (2300) even at a lower base clock of 2.3 Ghz.

Does anyone have any explanation for the discrepancy ?

[Richard (PassMark)]

As noted, the number of cores isn't important for the single threaded test. The base clock speed is also not much a factor as when running on a single thread test (assuming cooling is adequate) the cores should be running at or close to max turbo speeds. Some differences that can factor between the EPYC 7502P and EPYC 7742 CPUs is cache size and turbo clock speed, and to a small extent RAM (while the CPU tests are mostly independent, some test do interact with system RAM). So with half the cache and 1.5% lower clock speed, it should perform worse, but half is probably incorrect.

We don't have either processors to test, and there is only 1 results submitted for 7502P and likewise, 2 results for the 7742. It could just be a bad result for 7502P. If you end up getting the CPU, you can help contribute to the charts by submitting your results.

[user3141]

Thanks for the answer.

From my theoretical considerations and other published benchmarks the 7502P (32-cores @ 2.5Ghz base) should deliver a total passmark score of around 35000 points.

[QUOTE=Richard (PassMark);n45479
We don't have either processors to test, and there is only 1 results submitted for 7502P and likewise, 2 results for the 7742. It could just be a bad result for 7502P. If you end up getting the CPU, you can help contribute to the charts by submitting your results.
[/QUOTE]

Actually I was just about to buy the 7502P until I saw the strange result of 24000

Now I see the number of samples is only one (I thought it was two) so just an outlier seems plausible, as you said.

Will probably buy the CPU anyway and report the results here.

Thanks & Regards,
b