Jump to content

Welcome to INSYDIUM ONLINE

The place to talk about X-Particles and Cycles 4D

Sign in to follow this  
VVG

CPU+GPU render (hybrid) tile size

Recommended Posts

HI All, explain to me the formula (rule) for choosing the size of a tile (bucket size) in a hybrid render
CUDA (CPU + GPU) for maximum performance.

i have CPU AMD 1950x 16T/32C + GTX 1080Ti+GTX 1070

Edited by VVG

Share this post


Link to post
Share on other sites
3 hours ago, Mohamed Sakr said:

for RT it doesn't matter, for PV I would prefer 128x128 or 64x64, you need to test which one will give you higher performance.

You'd be surprised, but in most tests 16x16 or 32x32 is the best option. (without denoiser)

Share this post


Link to post
Share on other sites

In the new version of Blender Cycles they made it so the tiles size really doesn't slow down the GPU anymore if the tile size is not big enough.  It seems the new Cycles 4D also has this.  This means if you have a really fast GPU in comparison to your CPU you can set a smaller tile size and not have the frame of the render waiting to have the CPU finish the tiles it's working on near the end of the render.  In the past it was faster to have the GPU working only because you would have to set a bigger tiles size to not get a slow down and if you had slower CPU working with that the GPU would quickly finish all the tiles, but the ones the CPU was working on.  This would delay the render finishing because it's waiting for the CPU.

Long story short smaller tile sizes will probably be better.  I use 32x32 as default now.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×