DAV1D vs. LIBGAV1 Performance - Benchmarking Google's New AV1 Video Decoder

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • phoronix
    Administrator
    • Jan 2007
    • 67317

    DAV1D vs. LIBGAV1 Performance - Benchmarking Google's New AV1 Video Decoder

    Phoronix: DAV1D vs. LIBGAV1 Performance - Benchmarking Google's New AV1 Video Decoder

    With the surprise code drop of Google developing a new open-source AV1 video decoder as "libgav1", I set out this Saturday to run benchmarks on various systems for seeing how the performance is looking for this CPU-based decoder in relation to the more well known DAV1D decoder...

    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
  • tildearrow
    Senior Member
    • Nov 2016
    • 7097

    #2
    They really should have just used dav1d instead of reinventing the wheel....

    Comment

    • sturmen
      Junior Member
      • Jun 2017
      • 19

      #3
      Originally posted by tildearrow View Post
      They really should have just used dav1d instead of reinventing the wheel....
      I, for one, welcome increased "competition" in the space since I think different projects using different approaches can challenge as well as teach. I wonder what inspired Google to invest engineers' time into this. At first I thought it may be licensing, but dav1d is BSD licensed (not copyleft, business-friendly). Maybe Google just has engineers who think they can do better than dav1d and what we've seen so far all foundational.

      Comment

      • andyprough
        Senior Member
        • Feb 2012
        • 2451

        #4
        Wow, those numbers are pathetic. Sounds like another Google project doomed for the dustbin of history.

        By the way, the picture of your basement is insane, Michael. Time for you to go into competition with AWS I'm thinking.

        Comment

        • microcode
          Senior Member
          • Mar 2013
          • 2364

          #5
          I wonder what advantage they imagine a future version of gav1 having over dav1d. it'll be interesting to see if they come out with a writeup of some sort about why they think it's worth it, probably that would be after they've done the obvious optimizations. I must admit I don't see why they've chosen to limit themselves to profile 0 and 1, maybe this is meant as a basis for a hybrid (custom hardware co-designed with the software) decoder?
          Last edited by microcode; 05 October 2019, 04:46 PM.

          Comment

          • jpegxguy
            Junior Member
            • Dec 2018
            • 11

            #6
            Originally posted by sturmen View Post

            I, for one, welcome increased "competition" in the space since I think different projects using different approaches can challenge as well as teach. I wonder what inspired Google to invest engineers' time into this. At first I thought it may be licensing, but dav1d is BSD licensed (not copyleft, business-friendly). Maybe Google just has engineers who think they can do better than dav1d and what we've seen so far all foundational.
            Competition works for a product made by a company, or for pricing, but for open source? They should've contributed into dav1d instead. See ffmpeg vs libav. What's the point?

            Comment

            • geearf
              Senior Member
              • Dec 2011
              • 2150

              #7
              dav1d has started quite earlier as far as I understand it, I totally expected it to be faster for now, what matters is how fast they'll both be when they'll be considered stable...
              The graphs are weird though, why one graph per codec instead of a combined graph? That would be easier to read.

              Comment

              • sturmen
                Junior Member
                • Jun 2017
                • 19

                #8
                Originally posted by jpegxguy View Post

                Competition works for a product made by a company, or for pricing, but for open source? They should've contributed into dav1d instead. See ffmpeg vs libav. What's the point?
                Especially with regards to AV1: the same philosophy that begot SVT-AV1/rav1e/libaom. A single piece of software shouldn't be everything to everyone. People should be free to pursue different approaches ("Write it in Rust!" "Assembly for every supported platform!" "Portable C only!" "Do it JavaScript!") without being told how to to spend their time, and this exploration lets different projects learn from each other: they can adopt what works & ignore what flounders.

                If everyone had to contribute to the currently most successful project, we would have lost out on innumerable FOSS projects started by people who just wanted to try something different.

                The 19417bbfe819a6d39e02ac809524ca0725143072 commit message specifically mentions Android as an area of focus for libgav1; it remains to be seen what it brings to the table that dav1d doesn't. Time will tell!

                Comment

                • r1348
                  Senior Member
                  • Jul 2007
                  • 636

                  #9
                  Originally posted by andyprough View Post
                  By the way, the picture of your basement is insane, Michael. Time for you to go into competition with AWS I'm thinking.
                  I know it was hyperbole, but as someone working in AWS data centers, trust me: not even close.

                  Comment

                  • Royi
                    Senior Member
                    • Feb 2018
                    • 174

                    #10
                    Without comparing the output quality the FPS measure is meaningless.
                    Are they comparable? One could balance more quality for less speed which great.

                    Please also compare some output frames.

                    Comment

                    Working...
                    X