I just sent the scene to support's email.
Announcement
Collapse
No announcement yet.
Slow Lens Effects updates with high-res images
Collapse
X
-
Aleksandar Mitov
www.renarvisuals.com
office@renarvisuals.com
3ds Max 2023.2.2 + Vray 7
AMD Ryzen 9 9950X 16-core
64GB DDR5
GeForce RTX 3090 24GB + GPU Driver 565.90
-
I just finished a Team Viewer session with the support (thanks so much for the help!). The problem was the small amount of memory of my GPU. Turns out 12 megapixel images require at least 3 GB of GPU RAM for hardware accelerated lens effects and my GPU has only 4 GB. When it already has about 1 GB filled before doing lens effects, it's almost certain that it will fall back to CPU calculation because of insufficient RAM. And it frequently does.
So conclusion - I would need to upgrade my GPU if I want to work more smoothly with 12 MP images and higher.
By the way, does lens effects benefit from more than 1 GPU as RT does? If I put 2 GPUs will lens effects use both? Same question goes for Denoiser. I have access to another GTX 980 but it's inside another PC here and I don't want to go through the hassle of disassembling them just to test it out.Aleksandar Mitov
www.renarvisuals.com
office@renarvisuals.com
3ds Max 2023.2.2 + Vray 7
AMD Ryzen 9 9950X 16-core
64GB DDR5
GeForce RTX 3090 24GB + GPU Driver 565.90
Comment
-
The denoiser can utilize multiple GPUs only when used through the standalone vdenoise.exe (-useGpu=2) application (or through the Lele UI script).
By default the denoiser in Max uses only the most appropriate GPU - it checks which one are most likely to perform the best and usually stays away from GPUs that have attached display on it.
Lens effects are not utilizing multiple GPU devices at the moment.
I have added a note both for that and to check if there is a way to make the GPU lens effect use less memory.
Btw, both the denoiser and the lens effects (when they fit on the GPU and dont fallback to CPU) are usually much faster than CPU, that it may not make much sense to even do multi-GPU calculations (going from 20 seconds on CPU to 2 seconds on 1 GPU is 18 seconds less; going from 2 seconds on GPU to 1 second on 2 GPUs is just 1 second less). The denoiser algorithm however turned out more suitable for GPU calculations than the lens effects and it has much bigger gain from GPU calculations (to the point that I don't see the CPU-based denoising very feasible at all).
Best,
Blago.Last edited by savage309; 15-02-2017, 12:07 PM.V-Ray fan.
Looking busy around GPUs ...
RTX ON
Comment
Comment