1920 pixels width
100 samples
GTX Titan: 53 sec
GTX 560 : 1min 57sec
GTX Titan + GTX 560: 56sec
Using two GPUs is counterproductive! That’s not good. It’s needed to find a way to better use all the computing resources.
If I could make ANY blender wish come true, it would be a more responsive viewport, but I’ve seen enough people bring it up with the developers that I imagine it’s already a pretty high priority.
EDIT: Just to be clear, I’m satisfied with the cycles viewport performance. With the GPU, it is even faster than wireframe for me. All the viewport options other than cycles are what gets slow for me when I have large scenes.
yes, it is good. But with interior scenes, where there is lot of indirect light, if you want to be productive you want that the GPUs you bought are used at their best, becouse they are needed to get a fast preview during lighting phase or shader set-up.
Just wondering if there is a way to specify (user preferences) which card to use for 3d viewport cycles render (presumably a single card from what I glean from this thread) AND specify which cards to use for F12 cycles render (i.e. a combination of cards for maximum performance during test/final render). I don’t want to have to manually change user preferences each time I render.
This would be great, and while we’re at it, let the slower cards render smaller blocks. It’s faster to use all three of my cards at once, but a five minute render with 30 blocks inevitably finishes 29 blocks and then the final block takes another 30 seconds on the slowest card all by itself.