Tweet
Hi everybody,
it seems as if vray gpu rendering is scaling its performace non-linear depending on render time. I just want to understand if this is correct and - if so, why and how?
For example i have scene, where my workstation takes 1min 14sec per frame on an RTX 3090 while Rendernode1 takes 2min 30 with its RTX 2070 Super and Rendernode2 takes about 4min on its 970 GTX. That means I have a ratio of 1 : 2 (3090 to 2070) and 1:4 (3090 to 970). All rendered with cuda, GPU only. If I higher values for longer rendertimes the ration changes up to 1:4 (3090 to 2070) and 1:8 (3090 to 970). The time for writing output is same in all cases and nodes and takes round about 10sec. So my questions are:
Does anybody else see this effect?
Is this an issue?
I also found out, that the rendertimes are almost 70% higher, if I use CUDA with GPU and CPU (all nodes Ryzen 3900). Usually the rendertime has to decrease with more cores, even if these additional cores have to be handled too. The higher rendertime with GPU+CPU affects both cases, low settings with short render times AND higher values with longer render times.
I use 3ds max 2021 with vray 5.00.5 (due to some issues on later version i can't switch within projects), in all cases either GPU Memory nor RAM are used on its limit.
it seems as if vray gpu rendering is scaling its performace non-linear depending on render time. I just want to understand if this is correct and - if so, why and how?
For example i have scene, where my workstation takes 1min 14sec per frame on an RTX 3090 while Rendernode1 takes 2min 30 with its RTX 2070 Super and Rendernode2 takes about 4min on its 970 GTX. That means I have a ratio of 1 : 2 (3090 to 2070) and 1:4 (3090 to 970). All rendered with cuda, GPU only. If I higher values for longer rendertimes the ration changes up to 1:4 (3090 to 2070) and 1:8 (3090 to 970). The time for writing output is same in all cases and nodes and takes round about 10sec. So my questions are:
Does anybody else see this effect?
Is this an issue?
I also found out, that the rendertimes are almost 70% higher, if I use CUDA with GPU and CPU (all nodes Ryzen 3900). Usually the rendertime has to decrease with more cores, even if these additional cores have to be handled too. The higher rendertime with GPU+CPU affects both cases, low settings with short render times AND higher values with longer render times.
I use 3ds max 2021 with vray 5.00.5 (due to some issues on later version i can't switch within projects), in all cases either GPU Memory nor RAM are used on its limit.
Comment