Looking into creating a Logik Benchmark for flame.
lets use this Thread to talk about how / what /
My rough Idea…
-
Have multiple people contribute real world batch setups without media (replaced with flame generated sources). If we can get a good spread of like 10 setups, we should cover many real world scenarios.
-
add specific Benchmark batches, one that focuses on GPU, one that focuses on CPU, so again just to get a spread
-
Trigger all the batches in a project using python, which also records the time it took to render all batches, ideally with splits.
-
Create a website under Logik where people can enter benchmark results that will show a ranking e.t.c , maybe it can be auto-submitted with basic system info right from python (might be a legal issue with data collection…)
I think there should be a working group of like 3-4 People that split the tasks between them, so anyone wants to join getting stuff done! But lets first discuss if this is a way to go and if we even need this, I think we do.