TCC mode performance

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi
I have Quadro P4000. when launching the same workload WDDM mode shows better performance than TCC
I checked the profiler and it seems that driver latency is greatly reduced in TCC mode however kernel execution times are longer in TCC mode for the same kernels.

I checked SM and memory clocks and the seem to be at max in both modes.
What could be the reason for the slowdown when using TCC mode.

Thanks.

2

Hi Liran,
Could you file a separate bug from “https://developer.nvidia.com/->my account->my bug” for tracking the issue easily and updating you bug status quickly in the future.
Could you also provide us a self-contained reproducer and the steps you run into issue?
You can send attachment to CUDAIssues@nvidia.com for file exchange.
Thanks for your cooperation.