Announcement

Collapse
No announcement yet.

NVIDIA Releases Updated Vulkan Driver

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

  • #11
    Originally posted by blackout23 View Post

    10% TOTAL market share. That also includes all AMD and Intel cards. For NVIDIA internally that doesn't have to care about AMD and Intel the share is much larger. Also a GTX 590 or 580 is still faster than popular cards like the 750 Ti or their Tegra chips.
    Missing the point yet again. the 590 and 580 may be still decent today, but that's not what the bulk of the Fermi cards is made of. And adding support to the whole family just because the cards that make up maybe 5% of that family are still decent today doesn't look like a sound decision to me.
    Also, faster than Tegra? Is that really supposed to be an argument in favour of Fermi?

    Originally posted by bridgman View Post


    Please stop trolling. We are not dropping support for GCN parts. You know this.

    If you are talking ONLY about dropping Linux Catalyst's Crossfire support (that didn't really help on modern games anyways) then you could call that "dropping a feature" but it's a heck of a stretch calling that "dropping support".

    For some saying "anyone from the open source community can add Vulkan to GCN 1.0" because we won't commit to it, can be interpreted as "dropping support". </trolling>

    Comment


    • #12
      Originally posted by bug77 View Post

      Missing the point yet again. the 590 and 580 may be still decent today, but that's not what the bulk of the Fermi cards is made of. And adding support to the whole family just because the cards that make up maybe 5% of that family are still decent today doesn't look like a sound decision to me.
      Also, faster than Tegra? Is that really supposed to be an argument in favour of Fermi?
      I thought Vulkan drivers are so super tiny and fast to develop, because there is no heuristics involved anymore. Also they only figured that out after last SIGGRAPH?
      Why did they add preliminary support in the first place, because you could query device properties on Fermi with the first Vulkan driver?

      Comment


      • #13
        Originally posted by atomsymbol

        0.8^3 = 0.512 (Alt+F2 in KDE4)
        Even if it's not the case here you shouldn't trust calculating with Alt+F2 in KDE4 to much because it does errors sometimes, i found that out the hard way when things didn't add up and i didn't understand why until i did a manual calculation and saw that it did calculate thing wrong and it wasn't any advanced calculation.

        Comment


        • #14
          Originally posted by blackout23 View Post

          I thought Vulkan drivers are so super tiny and fast to develop, because there is no heuristics involved anymore. Also they only figured that out after last SIGGRAPH?
          Why did they add preliminary support in the first place, because you could query device properties on Fermi with the first Vulkan driver?
          Tiny and fast to develop compared to the OpenGL driver maybe. Software that has to work flawlessly on many types of hardware is never trivial to write.

          Comment


          • #15
            Does this new driver have support for X.org 1.18 ?

            Comment


            • #16
              Originally posted by ileonte View Post
              Does this new driver have support for X.org 1.18 ?
              No. You can "IgnoreABI" "1", but X crashes sometimes.

              Comment


              • #17
                Originally posted by atomsymbol

                0.8^3 = 0.512 (Alt+F2 in KDE4)
                Yeah, it's the same in KDE5. KCalc was the one disagreeing (it interprets successive ** differently from what I expected).

                Comment


                • #18
                  Originally posted by bridgman View Post

                  Please stop trolling. We are not dropping support for GCN parts. You know this.

                  If you are talking ONLY about dropping Linux Catalyst's Crossfire support (that didn't really help on modern games anyways) then you could call that "dropping a feature" but it's a heck of a stretch calling that "dropping support".
                  pre-GCN (r600 etc) are considered ancient right?

                  Comment


                  • #19
                    Originally posted by bug77 View Post
                    For some saying "anyone from the open source community can add Vulkan to GCN 1.0" because we won't commit to it, can be interpreted as "dropping support". </trolling>
                    Sure, but that's not what was said. Graham was talking specifically about adding SI support to amdgpu, not adding Vulkan support to SI.

                    At the time we were exploring a few different options, including extending amdgpu back to SI and extending the libdrm-amdgpu code to run over radeon IOCTLs on SI.
                    Last edited by bridgman; 24 February 2016, 12:22 PM.
                    Test signature

                    Comment


                    • #20
                      Originally posted by 89c51 View Post
                      pre-GCN (r600 etc) are considered ancient right?
                      eydee was specifically talking about GCN (at least that's how I interpreted his reference to "7990"), so not sure where you're coming from here.

                      Can you give me a bit of context ?
                      Last edited by bridgman; 24 February 2016, 12:20 PM.
                      Test signature

                      Comment

                      Working...
                      X