by t_3 » Mon Dec 26, 2011 1:26 pm
t_3
Mon Dec 26, 2011 1:26 pm
imo there is no workaround for such scenarios. if you need a gpu accelarated viewport and a cuda accelerated renderer, they need to share a non-expandable set of resources (vram). afaik there is no ability to "swap" out vram to system ram to extend a given limit (also not with cuda 4.1). thus the best way for heavy projects is to have at least 2 gpus; then use one for the host and the other for the renderer while working interactively, and do the final render "offline" e.g. in octane standalone - what should be fine for stills but still a bit problematic for animations.
regarding arion - there is a very illustrative sentence in the newsletter:
"And if you cannot make your scene fit in your GPU memory, you can still render it on the CPU at a very nice speed"
they seem somehow managed to get rid of the texture count limits (at least they say so), but that's already all about; "unlimited materials and objects" doesn't sound to be that big compared to octane, because i haven't seen such limits in octane (yet). and last but not least their max plugin isn't "live" yet (just an exporter imo), because they just announced a xsi-plugin as their first fully integrated plug...
„The obvious is that which is never seen until someone expresses it simply ‟
1x i7 2600K @5.0 (Asrock Z77), 16GB, 2x Asus GTX Titan 6GB @1200/3100/6200
2x i7 2600K @4.5 (P8Z68 -V P), 12GB, 1x EVGA GTX 580 3GB @0900/2200/4400