by KeeWe » Mon Jun 24, 2019 9:00 am
KeeWe
Mon Jun 24, 2019 9:00 am
I tested a scene from Thanu Lee and got the following results:
Setup:
Win 10 Pro
C4D R19.068
Octane
2018.1.3
Nvidia Studio 430.86
1080, 1080 ti, 2080 ti:
F 195, 512 Samples, Renderregion active:
2080 ti stops @480 Samples, 1080ti stops at 500, 1080 finishes.
rendertime: 2:06
------------------------
Master + 1070:
F 195, 512 Samples, Renderregion active:
estimated time: 1:27
1080 ti and 2080 stop at 340 samples
1080 stops at 420 samples
time estimated stops at 1:10/1:10
Rendertime: 2:37 (2:43 @ first test)
Conclusion: it seems Octane splits the image in one bucket per GPU. Each GPU renders their bucket but doesn't jump to another bucket when it finishes. It isn't even a network related problem, it happens local as well. It's just not very obvious and doesn't cost much time.
I assume it would be more obvious when using my 1070 in my master, together with the 2080 ti... but I've no time to test this scnerario.
I'm sure it worked in early v4 since I used network rendering as soon as you guys introduced the all access thingy and I did not noticing any speed drops with thep project I rendered in feb/march.
I really hope you guys can work this out, because right now my slave is useless and I've to wait to buy another GPU till I can be sure it decreases instead of increases my rednertimes.
Last edited by
KeeWe on Mon Jun 24, 2019 9:26 am, edited 1 time in total.
6850k // 32 GB // 1080, 1080 Ti, 2080 Ti // Win 10 // C4D 19.068