Geekbench 5 question

[novo]

TLDR: Geekbench should run better when nothing else is running, however, when I had the CPU blasting, I saw much better results.

Okay, so I'd like some clarity regarding Geekbench 5.

I had Better Touch Tool running, and for some reason, it was requiring a lot of CPU. Looking at Intel Power Gadget I was running an average frequency of around 4 GHz, without dipping. Running Geekbench in this state would somehow give me a really high score for the OpenCL benchmark (29715).

When I turned off Better Touch Tool, the GPU stabilized and I ran the test again, resulting in a much lower benchmark of (25548)

So, what's the deal here, anyone knows?

 

[maflynn]

I would run the benchmark 10 times and then take the average. Too many variables when you do a one off.

 

[bsbeamer]

Suggest you run the benchmark, wait one minute, then run again. Wait 3-5 minutes, then run again, Wait 5-10 minutes, then run again. Then immediately run again. Repeat on cycles and you'll get more true to life performance results.

Personally REALLY do not like GB5 and wish there were better benchmarks for METAL that could be used on a variety of machines.

 

[novo]

Yeah,
I don’t really understand how other machines get in the 30-32k OpenGL region with the same machine as people who get around 24-25k Doesn’t make a whole lot of sense to me how it can be that way.

 

[leman]

The deal is that Geekbench is an unreliable benchmark and it’s use should be avoided. It’s not clear what exactly it measures and the results show ridiculously high variance.