Announcement

Collapse
No announcement yet.

The Results Of Optimizing Radeon's VRAM Behavior

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

  • The Results Of Optimizing Radeon's VRAM Behavior

    Phoronix: The Results Of Optimizing Radeon's VRAM Behavior

    At the beginning of the year Lauri Kasanen had issued a call for help via Phoronix for collecting Radeon video memory statistics as he was looking to optimize Radeon's vRAM handling for his university thesis. The code and thesis are now available...

    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
    The kernel is fully backwards compatible with old mesa, so you should be able to compare just by changing mesa and the vram limit. I should note that there's a big ioq3 regression currently in mesa git[3], so if your comparison mesa is too far back, it could seem like it was caused by my work, when it's in reality in master too
    If you're not sticking with the same version and only changing the VRAM pressure patches, you're testing it wrong.

    Comment


    • #3
      Great Job.

      PS At what stage of education are You? Still some left or going to work (hint: AMD )?

      Comment


      • #4
        Just read the whitepaper, VERY nice job =).

        Glad to hear that your min-max allocator was merged into the kernel, and I hope that your work on the AI will lead to optimizations in the driver.
        All opinions are my own not those of my employer if you know who they are.

        Comment


        • #5
          under video RAM pressure he's scored a 1~2% frame-rate improvement
          Well, that's a bit dissappointing, considering the original post on phoronix said
          For r600g, much of the remaining gap to fglrx can be explained by less than optimal VRAM use.
          But improving the worst cases is of course a result that's good anyway.

          Comment


          • #6
            Originally posted by gigaplex View Post
            If you're not sticking with the same version and only changing the VRAM pressure patches, you're testing it wrong.
            I wanted to make sure there wasn't a comparison to 10.1, for example.

            Originally posted by przemoli
            PS At what stage of education are You? Still some left or going to work (hint: AMD )?
            Once this work is credited, nothing more, I'll have my degree then. Now I'm taking a vacation.

            Originally posted by ChrisXY
            Well, that's a bit dissappointing, considering the original post on phoronix said
            ...
            But improving the worst cases is of course a result that's good anyway.
            It's not that hard to have some VRAM pressure. If you have a compositor and a web browser open in the background, they could easily be taking 25% of your VRAM. Fragmentation is also an issue for < HD7k.

            It just happens that q3-based games, designed for 64-128mb VRAM, don't really cause any pressure on current cards. If you can easily fit every texture in OA and still have plenty of room, there's not much to optimize at that level. More recent games are of course heavier, bringing those situations to current cards.

            Comment


            • #7
              Will be this feature implemented on mesa drivers?

              Comment


              • #8
                Originally posted by Azrael5 View Post
                Will be this feature implemented on mesa drivers?
                Do you mean "does it work on mesa drivers" or "will it be merged"? It's been tested on r600g, radeonsi should work. Merged, cannot say, code still needs some cleanups before even proposing.

                Comment


                • #9
                  Originally posted by curaga View Post

                  It just happens that q3-based games, designed for 64-128mb VRAM, don't really cause any pressure on current cards. If you can easily fit every texture in OA and still have plenty of room, there's not much to optimize at that level. More recent games are of course heavier, bringing those situations to current cards.
                  Was testing done with more demanding engines?

                  Comment


                  • #10
                    Originally posted by curaga View Post
                    Do you mean "does it work on mesa drivers" or "will it be merged"? It's been tested on r600g, radeonsi should work. Merged, cannot say, code still needs some cleanups before even proposing.
                    Yes I know it is a feature under development so I would like to know if this improvement will be merged in open drivers.

                    Comment

                    Working...
                    X