output_cpu = torch.nn.AvgPool2d(3, stride=2)(input_tensor) # Performing the average pooling on GPU output_gpu = torch.nn.AvgPool2d(3, stride=2)(input_tensor.cuda()) output_cpu, output_gpu
Im only using a 220mm rad as well that is eating the gpu heat as well so thermal throttling may be an issue. But like I said I didn't see single core boost up and thats a low temp test with 1 core. I think I also said I had the x63 kraken,? Maybe not I dont remember but...
For the "no performance impact", i'll upload a small YT video later. My FPS does tank by 20-25 as soon as benchmark is enabled and jump back up by 20-25 as soon as I disable the benchmark. Hardware polling rate set to 2000ms Fan control polling rate to 5000ms Temp hysteresis ...