osiosiosi wrote:Hi
,
since i am planing to upgrade my system there are a few questions i am asking myself. Maybe somebody has helpfull info.
1. Any idea when those new RTX cards will be supported by Octane?
2. When support happens, will this be only the case for the 4.0 version? I am sasking this since many plugings are still at 3.8 and it seems 4.0 will take some time to be done (still no idea when this is sheduled?)
3. Will the NVLINK memory pooling feature be avaiable?
Nice if somebody could help with some info
Thx
1. We do have to recompile Octane for CUDA 10 for it to work on Turing, but unlike previous releases, NVIDIA has given us the CUDA 10 SDK and we have a V4 build on Turing already in house.
2. We should at minimum get V4 and possibly V3 updated in September before the cards launch with support. We are planning for V4 stable in September, assuming that RC3 fixes all known issues reported in RC1 and RC2 and no new issues are reported within a week or two following it's release.
3. NVLink on Turing GeForce RTX cards has not been tested by us yet, and we don't have all the details on how it works differently or is limited relative to Quadro/Tesla NV Link. We want to get our hands on the cards before we release this feature publicly, as we expect many Octane users will be looking to use it on 2080 Ti and 2080.