Comparios Geekbench performance - Win7 vs. Androidx86 on x86 lenovo Netbook

ruthan

Member
Apr 3, 2014
92
3
71
Benchmarked netbook was Thinkpad e120 with Win7 home, Android (4.4.2 RC1) was downloaded from androidx86 project and installed on ext2 partion of standart WD 320GB.

Results:
https://dl.dropboxusercontent.com/u/2741044/lenovo 30436bg - android ws windows - geekbench.pdf

What is most interesend what Android is quite comparable with Windows, without any optimalisation just after recomplication for PC x86. There are even benchmarks, were is Android significantly quicker.

If you ask on android on classic x86 nebook HW experience, it is quite pleasant.
Wifi working, sound working, touchpad working, physical keyboard working and its faster than of some mobile / tablet.

GPU - i using Intel2000 HD with OpenGL ES 3 and mesa 9.2devel, most of games running, but there some which not - especialy some Unity games, but others 3D games running. Epic citadel running, 3Dmark running, Carmaggeddon running, Modern Combat 4 running.
 

code65536

Golden Member
Mar 7, 2006
1,006
0
76
What's surprising is that there are these significant differences at all.

The OS should have no bearing on the performance of a process running within it. What does the OS do that could be relevant here? Schedule threads and manage memory. The memory tests are pretty consistent, so we're down to the scheduler. I find it hard to believe that the scheduler would matter this much. Nor am I convinced that compiler optimizations for the Android OS build would matter here. We're talking about synthetic benchmarks, after all, not a test of the OS's widgetry.

I'm actually leaning more towards differences in the benchmark program. Sure, it's nominally the same benchmark program, but is the code behind each test identical across platforms? How does the compiler optimization compare between the compiler used for the Windows version (which I would assume to be MSVC, which is pretty good at optimization) and the compiler used for the Android version of the benchmark program?
 
Last edited:

code65536

Golden Member
Mar 7, 2006
1,006
0
76
what's with the android side only reading 3GB of RAM?

Oh, good catch; I hadn't noticed that. Probably because the Android side is 32-bit; Google probably roped off and reserved a chunk of address space, so it can't use the 4GB. The 32-vs-64 bit difference probably accounts for the the performance discrepancies seen.
 

BrightCandle

Diamond Member
Mar 15, 2007
4,762
0
76
3GB sounds suspiciously like a 32 bit v 64 difference, which depending on the benchmark might make quite a bit of difference. It will literally be using different code paths on Windows 7 (64 bit) and Android (32 bit). The difference we see could very well be the 64 bit instruction set and registers making a difference mostly.

There are some definite differences for some programs Windows v Linux just due to the way the schedulers go about dealing with thread communication and stalling on locks but you don't see those in general benchmarks all that often (or with much impact).

I suspect the difference is just 32 v 64.