Paul (pavel.delete@this.noa-labs.com) on November 25, 2020 2:08 am wrote:
> Per Hesselgren (perhesselgren.delete@this.yahoo.se) on November 24, 2020 9:04 am wrote:
> > anon5 (anon.delete@this.anon.com) on November 20, 2020 10:51 am wrote:
> > > https://www.phoronix.com/scan.php?page=article&item=apple-mac-m1
> > >
> > > A lot rosetta2 bench
> >
> > Interesting but why no decompressing test?
> > If you compress anything you also (sooner or later) will decompress it.....
>
> Seeing going that far even under emulator on compression tests, the first thing coming to mind
> is them having really impressive cache management performance. It is certainly not due to cache
> sizes only, as it soundly beats X86 CPUs with huge, and or even bigger caches as well.

This is more or less coherent of what people were telling about their A SoCs as well. Not only caches are huge, but they are getting the most of bang for the buck out of them.

If Apple would've removed all the cruft from the die like sensor IP, DSPs, neural thing, GPU, and etc, they would've made M1 a brilliant server CPU, with a leading performance/size ratio.