0
$\begingroup$

I recently got my hands on a laptop with a GPU, and was playing around with it when I noticed there's a "GPU Subdivision" option in the preferences.

I'm assuming that enabling this offloads the work of generating subdivision to the GPU instead of relying on the CPU for that task. I'm also assuming that the GPU would be faster at this job than the CPU. However I've not had much luck putting that to the test. I tried applying high levels of subdivision to a high-poly mesh (a torus with lots of polys), but attempting to get the subsurf level up to the point where it would be noticeable that subsurf generation was slowing things down ended up pushing the machine to the point where Blender crashed, regardless of whether GPU or CPU was selected.

In the interests of determining whether it would be better to use GPU for subsurface generation or the CPU, is there a simple way of benchmarking the relative performance of either? As noted above, my attempt to stress test this with a high poly + high level subsurface crashed Blender in both GPU and CPU mode.

Improve this question
$\endgroup$
5
  • 3
    $\begingroup$ Hello, GordonM. I think a general Blender question about bench-marking GPU subdiv. is fair, here .. asking about the M3 in particular is verging on off-topic.. we're not equipped to answer questions on particular hardware components, so please make that 'incidental'. But I'm sure you're aware that there may be other bottlenecks in your system; memory, in particular? (We're sliding out of Blender SE's territory, here, hence the downvote, I suspect). If it can still be relevant to your needs, I would try to revise your question to one that can be answered here. $\endgroup$
    – Robin Betts ♦
    Commented Dec 13, 2023 at 5:36
  • $\begingroup$ Apart from that being off-topic as it is more a kind of hardware question, I don't even know what benefit you would have from benchmarks. When I'm not about to buy hardware but already have it, I make my own tests what works best or fastest - a benchmark that tells me my fast performance is actually supposed to be the slower performance is of no use for me. Also render times in tests are irrelevant for me when I can measure my own. $\endgroup$
    – Gordon Brinkmann
    Commented Dec 13, 2023 at 10:18
  • $\begingroup$ @RobinBetts I've updated the question to remove mention of specific CPU/GPU models. $\endgroup$
    – GordonM
    Commented Dec 14, 2023 at 22:20
  • $\begingroup$ @GordonBrinkmann I'm asking in order to determine if enabling GPU subsurf is the best setting, or whether CPU subsurf is the way to go. I'm not interested in absolute performance figures, just the performance of GPU mode relative to CPU mode. $\endgroup$
    – GordonM
    Commented Dec 14, 2023 at 22:21
  • $\begingroup$ @GordonM I understand, but as I said: just test the difference yourself when you already have the hardware. If your hardware makes no noticeable difference in scenes it can handle, why care about it? And if in one scenario GPU or CPU makes no difference but in another it does, I'll go with the one that performs better in the second and use it as well in the first where it doesn't matter, so I don't need to switch settings. And a general statement "GPU is better" or "CPU is better" cannot be made without the specific hardware. GPU is perhaps supposedly better, but not necessarily always. $\endgroup$
    – Gordon Brinkmann
    Commented Dec 14, 2023 at 22:35

0

Reset to default

You must log in to answer this question.

Browse other questions tagged .