x86 and ARM benchmark

[ndtech]

x86 and ARM benchmark.

The Geekbench benchmark now provides us a good opportunity to compare CPU performance of different systems.

I suggest we collect Geekbench results for all available CPU types at the same clock speed.

We can use 2.4 GHz for Intel and AMD CPUs.
So we can compare all results with results of mainstream ARM cores.

If the BIOS allows you to change CPU frequency, you can make required test.
Also if the BIOS allows you to disable CPU cores and Hyper-Threading (HT), you can make additional tests that show how the performance is scaled with additional core and Hyper-Threading.

And we can collect all results in Geekbench 4 benchmark database.
CPU frequency must be fixed at 2.4 GHz.
All tests:

- 1 core HT-off
- 1 core HT
- 2 cores HT-off
- 2 cores HT
- all cores HT
- all cores HT at 3.6 GHz

Additional test at 3.6 GHz can show how the performance is scaled with frequency change. We can use any high frequency, but 3.6 GHz (2.4 GHz + 50%) is good choice.
All cores must work at fixed frequency. So you must disable Turbo Boost features in BIOS.
It's better also to switch the power plan in Windows to "High Performance".
If you make the test with 2 cores in CPU (e.g. Ryzen) with multiple clusters, use 2 cores in same cluster.


Now we can use results of ARM systems from Geekbench database for comparison.

Apple A10 @ 2.34 GHz
http://browser.primatelabs.com/v4/cpu/2079398

HiSilicon Kirin 960, 4x Cortex-A73 @ 2.36 GHz + 4x Cortex-A53 @ 1.84 GHz
http://browser.primatelabs.com/v4/cpu/2105715

 

[imported_jjj]

Those results are polluted by encryption and memory tests and the benchmark is calibrated on Intel so the totals are very misleading and one needs to look at individual tests.
Just perf without power and area is rather pointless too.

Anyway, an Intel Pentium N3700 that turbos to 2.4GHz does about 1100 as an average score in ST https://browser.primatelabs.com/v4/cpu/1338413