[rob rhodes]

To be fair I dont think NL posted its limits for maxwell when V1 was released in 32bit, it was limited to 2GB or whatever windows limited it to and people still used it on dual core systems for production. Arion by all accounts has a much smaller footprint so if you can get a 4GB card its not just like 4GB of standard RAM in Fry or maxwell, also no multilight feature in arion to save RAM but with quick feedback i think the idea is you adjust all your lights in realtime to start with rather than throw a load of lights in and adjust after. You can still adjust the iso though. I too was/am still a bit sceptical about how it would work on my current GPU but I dont think its RCs fault, I just think its just a bit ahead of the hardware manufacturers like NL was with MW, but as soon as the 4-6GB cards start to become more wide stream I think RC will have the experience and base platform to keep ahead of the competition when it comes to unbiased rendering. Biased rendering though will blisteringly quick when Vray RT GPU is released think ill wait for that!

[Micha]

...Biased rendering though will blisteringly quick when Vray RT GPU is released think ill wait for that!

I'm curious for biased GPU rendering too, since I think, clever shortcuts in the calculation are not a problem and help during tight deadlines. But what can be expected by Vray GPU? I use light cache and irradiance map on a small render farm of five computers (Vray for Rhino). Interiors in high res are done within 10 .. 60 min. What kind of GI calculations are supported by Vray RT? If Vray GPU will stick at PPT (light cache), than I'm not sure it will be so fast to get a scene noisefree. Will be IrrMap available for GPU rendering? Or is the speed gain of the GPU so high, that the IrrMap isn't needed?

[Jan]

It's PPT, but I think that it's very fast. You can see the past year presentation at siggraph:

http://www.youtube.com/watch?v=Soyr1sQmro4

Regards