12-Way Radeon Gallium3D GPU Comparison

[curaga]

IMO, it's better to not do it at all, unless it can be done right.. And to do it right, you need all 3. Only with all 3 can you really draw conclusions about whether a card is video memory bottlenecked by memory capacity or memory performance or not. Not sure if phoronix wants to get into all that video memory stuff, but if they do, I hope they do it right, that's all.

I agree the memory should be reported.

But having to guess with the stats is pointless. If AMD had opened the performance counters, we could have a label next to each benchmark result saying "this benchmark was mostly memory-bound" or similar. Now the best that could be done is to determine whether the bench was CPU or GPU bound, nothing more.

[curaga]

(except for the Radeon driver since it only exposes the clock information over debugfs instead of sysfs, unless I run PTS as root or remember to chmod the debug directory before testing, it won't show anything)

Eh, I tried to change that months back, but got a response that Jerome (IIRC it was him) wants to create a new way to report information.


@Kivada

GPU-Z needs updates for every new GPU, does it not? This really points to it having some built-in database of them.

[agd5f]

GPU-Z needs updates for every new GPU, does it not? This really points to it having some built-in database of them.

IIRC, GPU-Z uses vendor specific interfaces to get that information.

[curaga]

Do those interfaces change each generation? That could explain the per-GPU new versions too.