FROM GEEKBENCH OVER CHEATING ANDROID

Somewhere over the past year or so the score for vpr on ARM64 has about doubled from what you would expect. Just for interest you can see the Apple A9 result here:

Look at the 175.vpr results on this page: These sorts of wild swings are not limited to GeekBench. > My days of not taking Geekbench seriously are certainly coming to a middle. > am starting to think that the same applies to quite a few 'macro' benchmarks. > I was always of opinion that closed source microbenchmarks are useless. > to measure how responsive those operations would be during a normal web browsing session. > several manipulations, including node insertion and deletion, and selective node class decoration, > workload is the ‘Processor Benchmarks’ page from. > and performs manipulations of that DOM.

> This workload constructs a HTML5 Document Object Model (DOM) based on the specification supplied at > Primatelabs whitepaper does not help at all here:

FROM GEEKBENCH OVER CHEATING ANDROID

> It certainly does not look like the same or even similar code is running on Android and on other platforms. > Here are examples on the same (Braswell) CPU: > Further comparisons seems to show, that it's not x86 vs ARM thing, but Windows/Linux vs Android. > U-series i3 CPUs, which belong to completely different league in all other tests. More so, ARM Cortex-A72 scores in this subtest are about equal to Kaby Lake > While comparing scores of various Atoms vs various ARMs I paid attention that ARM always wins massively Nobody in particular on Aug5:09 am wrote:

From geekbench over cheating