incredibly slow Cuda renders

[Bertjenkins]

Hi

I am trying to start using the GPU production renderer. I have a gtx 980 on OSX yosemite 10.10.5, maya 2016. Vray 3.30.01
(I'm using the geforce gtx 980 with cuda.)

I tried rendering a scene with quite heavy geometry. In viewport 2.0 the scene shows up super fast and looks pretty good. The hardware 2.0 renders are pretty good too- obviously the depth mapped sadows are limited- but it's fast! As soon as i switch to vray IPR or vray cuda render it takes absolutely aaaaaaaaages to build the BVH tree. Most of the time it takes so long I just end up giving up. Is there a faster way to do this? Should I be using vray proxies? At the moment viewport 2.0 is the better option for me, for this particular project.

Any help on speed optimization would be much appreciated.

[t.petrov]

We've fixed a bug related to slow building of BVH in the night builds.
Would it be possible to send us the scene, so we can test if we've fixed the problem or it is still present?
If you can share the scene please write to support@chaosgroup.com .

[Bertjenkins]

are large scenes ok? its 100mb

[vlado]

Our support guys can get you FTP info to upload it. Thanks!

Best regards,
Vlado

[Bertjenkins]

Just an update... the scene had maya materials... i ran the vraytuner material swapper on them (to convert to vray mtls) and it drastically improved things. so far seems to be running nice and fast, even with mostly maya lights. I wish i didn't have to swap all the materials over (as i can get some really fast passes with viewport 2.0) but i guess i can just set this up with render layers. And sofar the vray materials are running OK in viewport 2.0. anyway i will carry on testing and see if the bvh tree thing comes back. In the meant time the only slightly annoying thing is that GI is always on by default, even when you turn it off?