Post by Corruptinator in FlowBlur-App 1.2 comments

Viewing post in FlowBlur-App 1.2 comments

I have a RTX 2080 SUPER and for some reason I get this error when attempting to motion blur a 60 FPS footage:

CUDA out of memory. Tried to allocate 3.91 GiB (GPU 0; 8.00 GiB total capacity; 4.13 GiB already allocated; 791.63 MiB free; 5.20 GiB reserved in total by PyTorch)

Then the software just crashes.

Deleted post3 years ago

Deleted 2 years ago

a-n-d-r-e-w3 years ago(+1)

I'm getting the same type of error with a RTX 2070 Super.

Calvare303 days ago

It seems like you don't have enough memory/ram.

LeeSlacks3 years ago

yeah same here - on i dunno 16xx gtx ? i forget what it is :D the point is .. it's broken with the same 3.91GiB number.

Katzenwerfer3 years ago

That just mean you don't have enough VRAM to process it

Usually only affected by resolution, and with the current version the only workaround would be to downscale it

Corruptinator3 years ago

I'm wondering if there should be an option to have any high-scale videos split into separate pieces so then after rendering a frame it stitches back together to compensate for lost VRAM?

Katzenwerfer3 years ago

I would guess it is totally possible. I will ask the developer about it

Aikidou3 years ago (1 edit)

definitely, dainapp is developed by grisk too and i'm pretty certain it uses tiles for the "out of vram" option. cupscale and flowframes do too

itch.io

Viewing post in FlowBlur-App 1.2 comments