Same PC, different Geekbench scores between MacOS and Windows

[stefancosd]

I found something weird imo.
When I run Geekbench on my AMD Ryzen 9 5950X machine, I get different results between MacOS Sonoma and Windows 10.
The difference is more then I would expect for the Single core performance.

I am running OpenCore 0.9.5 with the latest AMD CPU patches and changed Hexadecimal to 10 because of 16 core CPU (https://github.com/AMD-OSX/AMD_Vanilla)

MacOS Sonoma
Single-Core Score: 1748
Multi-Core Score: 13111

Windows:
Single-Core Score: 2247
Multi-Core Score: 13772


Is this difference normal? If not, how to fix/improve it?

 

[Edhawk]

Depends, what was running in the background when you undertook each GB test?
How many tests did you run in each OS?
Are these an average of the scores from say 3 or 5 tests in each OS?
Or just a single test from each OS?
If just a single test in each OS, then difference like this can happen.

Assumption you are using exactly the same version of GB in both OS's.

 

[stefancosd]

Depends, what was running in the background when you undertook each GB test?
How many tests did you run in each OS?
Are these an average of the scores from say 3 or 5 tests in each OS?
Or just a single test from each OS?
If just a single test in each OS, then difference like this can happen.

Assumption you are using exactly the same version of GB in both OS's.

Nothing special running in the background and these are average scores after 10 benchmark runs.
Yes it is the same system. Dual boot using OpenCore 0.9.5

 

[Edhawk]

The difference may be accounted for by how the two OS's activate and run your AMD CPU. Windows is configured to work fully with your AMD CPU, whereas macOS isn't.

You are using a number of OC patches and fixes to get your AMD CPU working in macOS. It wouldn't take a lot for your CPU to be slightly limited in macOS and for this single-core discrepancy to show.

Your GB Single-Core score compares well with other Ryzen 9 5950X CPU's running macOS.

Search for Ryzen 9 5950X scores for macOS systems.

Yes some of the Windows and Linux scores are higher, but they are designed to work with this CPU. MacOS isn't.

Some of the Windows scores are absolutely rubbish!

Not sure how you would go about fixing any limitation. Not sure anyone else has noticed this limitation, so it may be something the AMD Kernel Patch developers can look at fixing?

@CaseySJ were you aware of this Geekbench single-core limitation/reduction with AMD CPU's? Are we asking too much to expect the GB scores to match those in Windows/Linux?

 

[CaseySJ]

...

@CaseySJ were you aware of this Geekbench single-core limitation/reduction with AMD CPU's? Are we asking too much to expect the GB scores to match those in Windows/Linux?

I think it would help to look at the GeekBench breakdown, which shows the results from each of the individual tests that constitute the overall score. If certain individual tests are systematically faster on Windows, that could offer some insight.

On the other hand we could make an argument for comparing GeekBench scores only across the same operating system -- macOS to macOS -- instead of apples-to-oranges. This would make sense if GeekBench is written in such a way that specific operating system APIs are benchmarked rather than the hardware directly (or in a consistent manner with the same API). But I doubt that GeekBench is written in an OS-independent manner because it would be difficult to isolate the operating system. And if the operating system was truly isolated, we wouldn't be seeing any difference in scores between Mac and Windows.