# of samples

[ClassicD]

Hi,
Does the above make any significat difference in the results, eg

GPU1=score500 #ofsample=1000
GPU2=score600 #ofsample=600

Would the more samples indicate better performance?
Thanks

[Richard (PassMark)]

Nearly all data in the charts comes from user submissions from our PerformanceTest software. The results are an average of all submissions. As such the more samples we get the more accurate the results become, whereas videocards with only a few samples have less accurate results.

In your example, since both GPU's have high number of samples, you can conclude that GPU2 should out perform GPU1.

[ClassicD]

Great thanks

[dune.kb]

sorry, new here - should be moved perhaps to another topics. Anyway, is it true/possible, that extremely low results can impact average CPU scores with low number of submissions?

For example, Intel Core i3-8145U 's average CPU Mark seems should be something like 6.000, but now it is just 4851 (https://www.cpubenchmark.net/cpu_loo...5U+%40+2.10GHz)
or AMD Ryzen 5 PRO 2400G (https://www.cpubenchmark.net/cpu.php...en+5+PRO+2400G) should score 10.000+, instead of current 9892.

Do you monitor/filter/exclude these abnormal (extremely low and multiple) samples either automatically or manually?

[David (PassMark)]

Yes, low (and high) scores can change the average. But once the number of sample is significant, one or two extreme results aren't significant.

But there are several filters. Some automatic and some manual. For example, overclocking is excluded automatically in nearly all cases from the main charts. Spammers are picked up in a semi-automated fashion.

I manually checked the i3-8145U results. There were a few duplicate low results. Now deleted. But I think it is very unlikely that the average will every get to 6,000. Proper average would seem to be around 5,400 for the range of machines that use that CPU.

10.000+, instead of current 9892

There was 1 low result (now removed). But it only moved the average by around 0.5%. So result for the Ryzen 5 PRO 2400G looks to be OK (now 9970). Are you really worried about that 0.3% variation?

[dune.kb]

thanks for the quick answer and good explanation. No further worries - these were just examples (I was able to get quickly) - glad we make it clear...

[bluepraha]

does anyone know how to show #of samples in 3 processors in practice see sshot? It's just a single core performance for coreldraw that doesn't know multithread.
Any advise for that?
thx Peter

[Richard (PassMark)]

does anyone know how to show #of samples in 3 processors in practice see sshot? It's just a single core performance for coreldraw that doesn't know multithread.
Any advise for that?

Not exactly sure what is the question...

The number of samples refer to the number of submissions from users with the particular CPU. The Single Thread Rating is the average thread rating of the submitted samples for the CPU running the single thread test. If you would like to see individual results for a CPU, you can use the Manage Baseline feature within PerformanceTest.

[bluepraha]

Not exactly sure what is the question...

The number of samples refer to the number of submissions from users with the particular CPU. The Single Thread Rating is the average thread rating of the submitted samples for the CPU running the single thread test. If you would like to see individual results for a CPU, you can use the Manage Baseline feature within PerformanceTest.

I see... I thought it was related to some test parameter, not to the feedback received from users
do you mean these 3 processors overclock? To "turbo speed"?
thx Richard
Peter

[David (PassMark)]

There is no information about overclocking in the table above.
"Turbo" speeds are safe clock speeds defined by the CPU vendor. Not overclocking.