|
|
|
|
|
|
by v1sea
793 days ago
|
|
|
It was probably from my workloads being relatively small that I could get away with 90Hz read on the cpu side. I'll need to dig deeper into it. The metrics I was seeing were showing 200-300 microseconds of GPU time for physics calculations and within the same frame the cpu reading from that buffer. Maybe I'm wrong, need to test more. |
|
|
If the only reason you were "wrong" was because you intuitively understood that it wasn't worth a large amount of valuable human time to save a small amount of cheap machine time, you were right in the way that matters (time allocation) and should keep it up :)