Announcement

Collapse
No announcement yet.

Combined Quadro K4200 & Geforce Titan X experiences

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Hi, it seems I have overlooked something myself as well, Vray will not let me set the K4000 as a openCL device (it worked as one before installing the Titan X). So this means the (non) usage of the quadro K4000 is not distributed rendering but driver related in this case. I'm using the 353.62 drivers. Could you tell me where this log is usually saved when Vray is running in server mode?

    Also, back to my original concern (I don't mind my quadro K4000 not being used in a workstation), are you sure Vray RT server will use all GPU's in a renderslave mode? How did you confirm this? And finally, if you are doing it, does your cpu have an onboard gpu? (Setting the onboard GPU as primary adapter would be the only workaround for the issue I see on other sources.)

    Click image for larger version

Name:	opencl devices 2.png
Views:	2
Size:	61.4 KB
ID:	858156
    Last edited by auke; 24-08-2015, 07:58 AM. Reason: Added image with the failing of activating the K4000

    Comment


    • #17
      This tool only sets one environment variable. If it doesn't succeed, it means that the account with which the tool has been started does not have enough rights to change environment variables. The only interaction the tools has with the driver is when querying for devices, and if it shows them both everything is okay.

      Try to start ocldeviceselect with "Run as administrator" or with different account. Sometimes doing that via remote desktop is not working, I think (you can try VNC or attaching monitor just once to set up which GPUs to be used using ocldeviceselect).

      If that does not help, you can try editing the env var manually (instead of using ocldeviceselect) - it is should be something like VRAY_OPENCL_PLATFORMS_x64 with value "{NVIDIA CUDA,GPU,1,1};" (without the quotes).

      Remember to restart all processes related to V-Ray after updating the env var successfully. If you are not sure which processes to restart (for example, V-Ray spawner, etc), it will be best to restart the whole machine, just in case.

      Once the env var has the correct value (meaning, both of the GPUs in ocldeviceselect are "ticked" when it is being initially started) V-Ray RT GPU should be able to use all the GPUs that the machine has. You don't need onboard GPU on the CPU for that.

      For nVidia GPUs you should use CUDA, since it will be much faster.
      Last edited by savage309; 24-08-2015, 08:17 AM.
      V-Ray fan.
      Looking busy around GPUs ...
      RTX ON

      Comment


      • #18
        Thanks! Might try it the harder way next. Also just tried the remote rendering again to another system, which has only 1 GPU and no onboard graphics and it worked fine as you told, so no more worries on that subject! I'll let you know if I run into other issues.

        Comment

        Working...
        X