Announcement

Collapse
No announcement yet.

Intel's Open-Source Compute Stack Continues Work Towards Multi-GPU Support

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

  • Intel's Open-Source Compute Stack Continues Work Towards Multi-GPU Support

    Phoronix: Intel's Open-Source Compute Stack Continues Work Towards Multi-GPU Support

    While Intel's large open-source Linux graphics driver team has been pushing a lot of code over the past number of months for bringing up their DG1 graphics and other current/forthcoming discrete graphics offerings, one area that is still in its infancy is around the multi-device handling. At least from the compute side, there is some recent progress being made for multi-device support...

    http://www.phoronix.com/scan.php?pag...ti-Device-Work

  • #2
    Typo:

    Originally posted by phoronix View Post
    With this week's Intel Graphics Compilter update
    Were you thinking of filters?

    Comment


    • #3
      The only way this will matter is if those 2+ Gpus are identified as one for state trackers. Like DG1-D with 2x shaders and 2x shared memory. They should avoid extended Gpu to Gpu memory sharing via PCIe tho. Just letting Vulkan to share work via Slices is not the optimal solution and will not help if you want to laser solder Gpu Dies.

      Comment


      • #4
        Originally posted by artivision View Post
        The only way this will matter is if those 2+ Gpus are identified as one for state trackers. Like DG1-D with 2x shaders and 2x shared memory. They should avoid extended Gpu to Gpu memory sharing via PCIe tho. Just letting Vulkan to share work via Slices is not the optimal solution and will not help if you want to laser solder Gpu Dies.
        As mentioned in the article, That's for computing not for gaming.
        Thare are many computing systems with nodes in form of 1CPU+4(Nvidia)GPUs.

        Comment

        Working...
        X