Announcement

Collapse
No announcement yet.

Rav1e 0.2 Rust AV1 Encoder Released With 40~70% Improvements Over Previous Release

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

  • Rav1e 0.2 Rust AV1 Encoder Released With 40~70% Improvements Over Previous Release

    Phoronix: Rav1e 0.2 Rust AV1 Encoder Released With 40~70% Improvements Over Previous Release

    Compared to the inaugural Rav1e 0.1 release just over one month ago, Rav1e 0.2 was released on Wednesday with 40~70% better performance depending upon the encode settings...

    http://www.phoronix.com/scan.php?pag...e-0.2-Released

  • #2
    I wish you had inserted 0.1 benchmarks just for comparison.

    Comment


    • #3
      Originally posted by tildearrow View Post
      I wish you had inserted 0.1 benchmarks just for comparison.
      https://github.com/xiph/rav1e/issues/1856

      Here are some. The results for the OdroidN2(arm64) and Core2Quad are outdated. Phoronix test suite for Rav1e 0.2 is running at the moment on that Hardware

      Comment


      • #4
        These numbers look like they are not using multi-threading at all. No scaling with core count whatsoever is visibile. At the same time, the high clocking parts perform best.

        Comment


        • #5
          Is it faster than hand optimized assembly yet? ;-)

          Comment


          • #6
            Looking for another 70% improvement...

            Comment


            • #7
              Originally posted by GruenSein View Post
              These numbers look like they are not using multi-threading at all. No scaling with core count whatsoever is visibile. At the same time, the high clocking parts perform best.
              From my experience with Rav1e does it scale with the Tiles and looking at the rav1e profile it does use 4 tiles so it will only use 4 Cores https://github.com/phoronix-test-suite/test-profiles/commit/f38dcb8f946ca73a4ce3e57d12764e208c99ac86

              Comment


              • #8
                Originally posted by rene View Post
                Is it faster than hand optimized assembly yet? ;-)
                Yes, because you don't have to rebuild your life after the monthly security breach caused by human errors in the c/assembly.
                In Rust We Trust.

                Comment


                • #9
                  Originally posted by elatllat View Post

                  Yes, because you don't have to rebuild your life after the monthly security breach caused by human errors in the c/assembly.
                  In Rust We Trust.
                  _ lol_

                  Comment


                  • #10
                    Originally posted by GruenSein View Post
                    These numbers look like they are not using multi-threading at all. No scaling with core count whatsoever is visibile. At the same time, the high clocking parts perform best.
                    Multithreading is actually not a requirement. The most heavy usage these encoders see are from websites like Facebook, Youtube, TikTok and whatnot. Those do not need multicore support.because they're encoding so much stuff at once, even if a job only uses one CPU, they'll still saturate whatever resources they have.
                    Multithreading would only be useful for the casual home user. It will only be tackled if it's not too much of a hassle. Which, if Rust lives up to its promises, won't be a problem.

                    I'm hoping this will be added at some point, but I'm not holding my breath. At least now right now.

                    Comment

                    Working...
                    X