Announcement

Collapse
No announcement yet.

Intel SVT-AV1 0.6 Released With AV1 Decoding, SIMD Optimizations

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

  • #11
    Originally posted by tildearrow View Post

    I have. However, reports say SVT-AV1 produces poor quality results...
    Relative to speed, I think. The last I saw, I think it was supposed to have quality in between x264 and x265, which isn't exactly bad, per se, it's just really slow in comparison to using those codecs if you're going to have that quality anyway.

    Comment


    • #12
      Originally posted by tildearrow View Post

      I have. However, reports say SVT-AV1 produces poor quality results...
      I don't understand such claims. An encoder has to function according to the specification. The quality should depend on the encoding settings only.

      Comment


      • #13
        Originally posted by log0 View Post

        I don't understand such claims. An encoder has to function according to the specification. The quality should depend on the encoding settings only.
        That's exactly the point, the settings are tuned low in order to speed things up. Lots of things the spec defines as optional are turned off. Etc.

        Comment


        • #14
          Originally posted by smitty3268 View Post

          That's exactly the point, the settings are tuned low in order to speed things up. Lots of things the spec defines as optional are turned off. Etc.
          Well, then the complaint should be that the default settings produce poor results, not the encoder.

          Comment


          • #15
            Originally posted by log0 View Post
            Well, then the complaint should be that the default settings produce poor results, not the encoder.
            The encoder is what's producing the poor visual quality results in comparison with other AV1 encoders. No reason to try and make it more complicated than that.

            Comment


            • #16
              Originally posted by log0 View Post
              > poor quality results

              I don't understand such claims. An encoder has to function according to the specification. The quality should depend on the encoding settings only.
              What you are describing is a decoder. The encoder is a decision maker! Just like two people speaking the same language will use different words to tell the same story, so do encoders. An encoder has to solve millions of hard problems per second in trying to be as concise as possible, and has the full freedom to do absolutely whatever it wants as long as a hypothetical compliant decoder can decode it.

              For a format as complex as AV1, trying to brute-force the problem will give you those reference encoder run times, thousands of times slower than real-time, due to the combinatorial explosion. Remark that for a decoder, the features add up to make it slow, but the combinatorial explosion is uniquely an encoder problem.

              In the other end of the spectrum, it is very easy to make a fast encoder by simply doing bad decisions (i.e. dropping features). This is what Rav1e started with. But to be both fast and have decent quality, you have to have good heuristics for everything. This is what separates a good encoder from a bad one.
              Last edited by andreano; 07-05-2019, 01:35 PM.

              Comment

              Working...
              X