[Bf-cycles] CUDA error on Azure NC6 VM with Tesla K80

[Thomas Dinges]

My best guess is, that the binary, which you copied, doesn't match the 
Cycles version of the standalone repository.

Make sure they match, then it should work. Do you compile Cycles 
standalone yourself? Then you should be able to compile the 
corresponding CUDA kernel alongside.


Am 11.12.2016 um 08:42 schrieb Prayaag Kasundra:
>
> We are trying to run cycles on new Azure N series (NC6 ) Virtual 
> Machines which has Nvidia Tesla K80 GPUs
>
> -I created fresh new VM with Windows Server 2016 OS and installed 
> Tesla K80 drivers from Nvidia website
>
> -I installed CUDA tookit 8.0 (Win 10 installer)
>
> - Now after running –list-devices option I saw K80 in the device list 
> along with CPU .
>
> -After that I ran a render with device=cuda and got this error 
> http://prntscr.com/dhz3g9 . So I copied kernel_sm_37.cubin file from 
> blender/scripts/addon/cycles/bin folder of latest blender build zip.
>
> - After that the above error went away but then new error appeared 
> http://prntscr.com/dhz3wd
>
> How do I fix this ? Please help.
>
> Thanks,
>
> Prayaag
>
>
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> https://lists.blender.org/mailman/listinfo/bf-cycles

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20161211/4651d55f/attachment.htm