GPU Next 1.1 - Num samples per thread reduced

[Morne]

Hi guys

I'm trying out an old CPU project on the new Next 1.1 GPU. Everything default, except using bucket mode. I get a "warning: Num samples per thread reduced to 113951, rendering might be slower"

What's that suppose to mean and why am I getting it?

[savage309]

This is log messsage for us (devs), don’t worry about it too much.

Best,
Blago.

[ThomasHorsley]

Hi Morne,

I just got the same message when testing out (2x) 2080Ti s with a 980Ti during rendering in bucket mode.
When using only the two 2080 Ti, they use about 6.5GB each during rendering (according to GPUz) - in this case they render happily using about 97% GPU load. No errors / warnings.

However I noticed when trying to render the same scene using all 3 cards, the scene does loads onto the 980Ti, using only about 5.8GB, and the 2080Ti also reduce the amount of memory they use.
Textures set to 'full size' - not 'MipMapped'

Most important point, is that the render took longer when using all 3 cards ! - the usage of the 2080 Ti went down, hovering around 85% which is probably why it took longer to render.
And of course, the last buckets were allocated to the slower 980Ti - so the 2080Ti's sat idle.

Shame that Vray doesn't re-allocate the last buckets to the unused GPUs - ie) 'notice/check' that there are idle GPUs, then re-allocate the last buckets, or better still, subdivide them half/quarter like in CPU mode and then re-allocate.