Benchmark Tests?

[Muaddib]

Hi all,

Has anyone run any common benchmark tests on the Pandora yet?
When I install Linux on a new system I run this one:
bench
...followed by this one:
bandwidth
...which produces a graph.
Surely there is a good OpenGL benchmark out there... maybe this one?
OLGbench

Thanks.

 

[mali]

Those won't work, as Pandora has got an ARM SoC and uses OpenGL ES. There are nBench results floating around somewhere, though.

 

[quadomatic]

I'm not sure what a benchmark would be good for. Most of the software running on the Pandora will be optimized for the Pandora specifically, while benchmark software would not be. Not sure what the comparison would be good for, cept maybe for determining the performance impact of different OSes and Kernels running on Pandora.

 

[juniorm33]

To test the impact of under/overclocking.
Comparison to other devices.
Stress testing.
A benchmark can be optimized for Pandora.

Obviously it's all dependent on the benchmark, and actual usefulness will vary.
Also I'm not saying a bench is the only or best way to do any/all of what I listed, just saying what it could be good for.

 

[mali]

A Quake timedemo with default config is my favourite benchmark. It works on many devices.

 

[juniorm33]

Like you said there may be nBench figures about, but are they at different frequencies?
Guess you can probably work it out once you have a baseline, don't know.

And Quake timedemo is a benchmark in it's own right in my view rightly or wrongly :lol:

Edit: sounds like I'm disagreeing with you mali, I'm not

 

[mali]

Like you said there may be nBench figures about, but are they at different frequencies?

Yep, notaz(I think) did some tests at different speeds. IIRC, that discussion was about floating point performance, which isn't very good on the A8, but this can be compensated by using NEON.

 

[juniorm33]

Some nBench figures

 

[Muaddib]

Those won't work, as Pandora has got an ARM SoC and uses OpenGL ES. There are nBench results floating around somewhere, though.

Neither having an ARM nor having OpenGL would prevent either of the two tests that I listed from running, if the Pandora runs Linux.

 

[lulzfish]

It won't tell you anything important, since nobody uses benchmarking programs for daily activity like web browsing or gaming.
Benchmarks only tell you how good your system is at running benchmarks.

 

[mali]

edit: got it, nvm.

 

[Muaddib]

It won't tell you anything important, since nobody uses benchmarking programs for daily activity like web browsing or gaming.
Benchmarks only tell you how good your system is at running benchmarks.

Incorrect. An artificial benchmark like 'bandwidth 0.21' can identify the limits of hardware, and help identify flaws in that hardware that need correcting. Maybe those flaws can't be fixed short-term, but it is better to know and be honest about flaws, rather than to conceal them.