progresssive and poor DR performance.

[super gnu]

my two rendernodes are not performing well with progressive.. cpu usage is all over the place.. often dropping to 20-30%..


ive tried upping the ray bundle size, first to 256, then 512, now finally 1024. to be honest it seems worse now than before.

ive got 10gb networking, so that shouldnt be a bottleneck.

[vlado]

The other thing to try is higher min shading rate - 8 or 16, for example. It seems to work well for the scene that you sent me. Not as well as the bucket renderer, but there isn't much to do about that.

Best regards,
Vlado

[super gnu]

ok a ray bundle size of 10000 seems to help.. slaves are pegged at 100% now, but unless its my imagination, main workstation is using more ram. and seems very unresponsive.

edit - its just reached the end of a pass, and now the main workstation has dropped to 6% cpu usage... no.. make that 3%

i guess its preparing huge bundles of work to send to slaves.

[super gnu]

just saw your reply Vlado.. ive ended up using a min shading rate of 32.. seemed to give me a better balance of aa to shading. i could of course turn it higher.. or lower.. is there a sweet spot wrt dr or just the higher the better.

[vlado]

is there a sweet spot wrt dr or just the higher the better.

It depends on how heavy the scene is generally. There is always some drop between the progressive passes as the main machine needs to update the image and recalculate a new AA mask.

Best regards,
Vlado