Announcement

Collapse
No announcement yet.

Radeon HD 3870 X2 Open-Source Support

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

  • Radeon HD 3870 X2 Open-Source Support

    Phoronix: Radeon HD 3870 X2 Open-Source Support

    In a git commit this afternoon by Joachim Deguara, an AMD Linux software engineer, support for the new Radeon HD 3870 X2 has been added. As we shared yesterday, the Radeon HD 3870 X2 is the new high-end ATI graphics card but is composed of two Radeon HD 3870 GPU cores...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    slightly off-topic but has someone thought of / tested / tried / figured out if it works yet,
    to put an X2 card in each PCIe socket that supports PCIe 16x ?
    Imagine a mobo with 4 PCIe 16x slots and 4 of these puppies in it, e.g. 8 GPU cores !!!
    (Driver nightmare ought to be negligible beyond support for 3 GPUs, e.g. one, two, three,many ..... or am I overlooking something?)

    Comment


    • #3
      They only have 1 crssfire connector so you can only connect up to 2 cards afaik

      Comment


      • #4
        Originally posted by Swoopy View Post
        slightly off-topic but has someone thought of / tested / tried / figured out if it works yet,
        to put an X2 card in each PCIe socket that supports PCIe 16x ?
        Imagine a mobo with 4 PCIe 16x slots and 4 of these puppies in it, e.g. 8 GPU cores !!!
        (Driver nightmare ought to be negligible beyond support for 3 GPUs, e.g. one, two, three,many ..... or am I overlooking something?)
        I think you're overlooking the fact that this arrangement isn't simply a linear growth in the number of GPUs; the interconnection between them also becomes more complex when more are added. With hardware Crossfire bridging, there are at least two layers of interconnect and device abstraction: mainboard-to-card and GPU-to-GPU. Various limitations could exist in each of these layers in the driver architecture and/or in the hardware itself.

        Comment


        • #5
          Stuffing a lot of GPUs into a single box seems pretty workable for Stream Computing applications, but getting good performance scaling on graphics gets harder as the number of GPUs goes up.
          Last edited by bridgman; 23 February 2008, 11:12 PM.
          Test signature

          Comment

          Working...
          X