iMac Pro Different benchmarks of iMac Pro 10xCore?

[lorddevil]

When comparing for example Geekbench numbers - I can see that they vary on fresh machines between 4.8k and 5.3k - which I really wonder why this is happening.

On a fresh machine without anything running, I am getting around 5000-5100 constantly. My room temperature is 22°C. So does this solely depend on the temperature around the imac - which causes those differences?

Also from all the benchmark on the net and from my own - the processor never goes over 4.2Ghz - but they advertised 4.5 Ghz Turbo Boost. I tried many different single workloads - but 4.2 is the most you can see (Intel Power Gadget - 1ms refresh resolution).

The earlier benchmarks on the inet with the 5.3k score - maybe their single core was reaching the 4.5Ghz and thus having a higher score and they changed the firmware afterwards? Or it was just related to the older version of Geekbench...

Just wondering where this discrepancy is coming from and why we can't go over 4.2Ghz...

Any ideas?

 

[nick82]

Log to a file in Power Gadget and you'll see it register operations at 4.5Ghz.
I've never seen the graph show any higher than 4.2.

As for Geekbench, I got 5380/35796 on a run with nothing else running (idle at <1% CPU usage).
From what I see, Geekbench won't get anywhere near stressing the machine on a power/heat basis.
I'm using the trial of Geekbench, not sure if that matters.

At idle, my machine is landing ~38 degrees I wouldn't expect a room at 22 would cause a problem.

Nick

 

[jerwin]

From what I see, Geekbench won't get anywhere near stressing the machine on a power/heat basis.
I'm using the trial of Geekbench, not sure if that matters.

At idle, my machine is landing ~38 degrees I wouldn't expect a room at 22 would cause a problem.

Last I checked, Prime95 was the best option for stressing a CPU.

 

[Trebuin]

Thanks for the information on IPG capping out at 4.2. I cover a little bit about the CPU there. The main difference in these benchmarks are where the CPU temp starts at. Also, the RAM has an impact. Geekbench will not use 100% of the CPU and only takes a small snapshot so you'd need a much more processor intensive test to see how these really differ.

 

[lorddevil]

I just tried prime95 too on 1 processor - its the same it never goes to 4.5Ghz - seems like its capping at 4.2/4.3 checking the log files of the intel gadget.

 

[Fernand]

PLEASE, folks, the gold standard for comparing is still the free Geekbench 3 in free 32 bit mode. Everybody can run it without having to buy anything. And if you look at Everymac.com you can compare any mac back to 1,1 cMP days even. It doesn't give any less accurate measurement that any other benchmark, you can measure your Mars flight in centimeters or inches, but don't mix them. Geekbench 4/64 bit gives radically higher single core numbers relative to GB3/32.

Y'all lucky owners of the latest gear, please, please, get the free GB3/32 and post those numbers identified as such, so we can compare. This is a separate issue from detailed turbo options.

If you want to also post BG4/64 numbers, and different stress tests, that's great, but please identify them as such.