Same demo Performance (FPS) on different Hardware?

[cerbert]

Good Day to all.

Being a total newbie I've started by tackling the demos. The first thing I did was to tamper with BasicDemo in order to change the shape, number of objects for cube and number of cubes on the scene simply by giving the executable a ".ini" file to read.

It worked pretty well and I played a bit around, trying to "benchmark" the performance of a dinamic world given the number and shape of colliding rigid bodies and the toggling of textures and (obviously) shadow.

I was a little surprised when I moved the tampered demo from this hardware
- Centrino Duo / 4GB Ram / nVidia Geforce 8600M@512MB Ram
to this hardware
- Intel I5 / 6GB Ram / nVidia Geforce GTS 360M@1GB Ram

because the FPS was exactly the same. I expected that the performance would increase on the second and most powerful machine.

So I did also tamper with the MultiThreadedDemo, but I did notice that increasing the maxNumOutstandingTasks did not effect the overall performance neither on the same machine nor comparing the two machines.

There is some "cap" in the demos? Or I need to modify part of the initialization process to have "scalable" demos?

Whoops... I'd almost forgot: I'm building with Visual Studio 2008 on a Windows 7 x64 System with ATI OpenCL SDK installed.
If anyone so desires, I'd gladly share my "hacks" of the demos (not that much of a hacking, anyway).

Thanks for your effort.

[cerbert]

Uhm... my question needs to be clarified... or the only answer is "RTFM"?

I solved the "multithread" part (I was just setting the wrong number of thread... so performance simply was not affected), but the fact that the demos run at the SAME FPS rate on two different machines remains.

The only difference is (obviously) that where the older machine needs to boost the CPU usage over 80%, the powerful one only needs to use a mere 20%...