Hi,
After building the GPU code I thought it may be interesting to compare CPU and GPU performance. I'm running on Ubuntu with a Ryzen 5600X (6 core) and a nvidia GTX 1070Ti (~2400 core). For a 20000 iter, 150 walker, 15 param code:
CPU: est 56hrs, 10.7s/it
GPU: est 50hrs, 9.05s/it
It's nice to see that the time it took to build the gpu code is reclaimed on my first run. I'm curious if upgrading to the latest cards e.g. RTX 3080 with ~8000 cores would there be a 3x speed increase? Also would changing the GPU on a desktop require a whole new build of galario?
Hi,
After building the GPU code I thought it may be interesting to compare CPU and GPU performance. I'm running on Ubuntu with a Ryzen 5600X (6 core) and a nvidia GTX 1070Ti (~2400 core). For a 20000 iter, 150 walker, 15 param code:
CPU: est 56hrs, 10.7s/itGPU: est 50hrs, 9.05s/itIt's nice to see that the time it took to build the gpu code is reclaimed on my first run. I'm curious if upgrading to the latest cards e.g. RTX 3080 with ~8000 cores would there be a 3x speed increase? Also would changing the GPU on a desktop require a whole new build of galario?