Announcement

Collapse
No announcement yet.

Nouveau Can Beat NVIDIA With Cairo In Select Cases

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

  • Nouveau Can Beat NVIDIA With Cairo In Select Cases

    Phoronix: Nouveau Can Beat NVIDIA With Cairo In Select Cases

    Chris Wilson has shared his testing experience of Cairo with NVIDIA ION hardware on the open-source Nouveau driver and the closed-source NVIDIA blob. In certain situations, the Cairo performance does better with Nouveau than the official NVIDIA Linux driver...

    http://www.phoronix.com/vr.php?view=MTI4MTA

  • #2
    How does one interpret those bars some of which grow downwards some upwards?

    Comment


    • #3
      Originally posted by mark45 View Post
      How does one interpret those bars some of which grow downwards some upwards?
      I'm pretty sure that the top is 11 x Faster while the bottom is 11 x Slower (or in the red). Weird graph.

      Comment


      • #4
        Originally posted by mark45 View Post
        How does one interpret those bars some of which grow downwards some upwards?
        Relative to not GPU accelerated path I think.

        Comment


        • #5
          Oh, I noticed at the bottom right "relative to image" whatever that means. And yes, this approach sucks balls, just put another bar "Image".

          Comment


          • #6
            Unsurprising. Nouveau has always been a few orders of magniture faster than the nvidia driver on 2D acceleration. The nvidia driver has only ever had an advantage in 3D acceleration.

            Comment


            • #7
              Originally posted by carewolf View Post
              Unsurprising. Nouveau has always been a few orders of magniture faster than the nvidia driver on 2D acceleration. The nvidia driver has only ever had an advantage in 3D acceleration.
              "A few orders of magnitude" means "at least a hundred times", do you really mean that?

              Comment


              • #8
                Originally posted by mark45 View Post
                "A few orders of magnitude" means "at least a hundred times", do you really mean that?
                Yes. I remember trying the XRender mode of kwin on the proprietary driver. It was slower than the generic SVGA driver. Which means it was slower than not accelerated at all. Incredible. Of course with the nouveau driver it was the other way around, with OpenGL mode being too slow.

                Comment


                • #9
                  they know
                  https://devtalk.nvidia.com/default/t...e-regression-/

                  Comment


                  • #10
                    Originally posted by kokoko3k View Post
                    I love that one of the solutions was to, basically, make cairo less powerful (specifically, not allow it draw all of the gradient types that it currently supports).

                    Comment


                    • #11
                      Originally posted by mark45 View Post
                      How does one interpret those bars some of which grow downwards some upwards?
                      The bar represents the cairo-image backend, and being above the bar represents a speedup. As I understood ickle's explanation to me the main differences between the image backend and the xlib backend is that the image backend can't be threaded and doesn't perform the tesselation step (this being a big computational win).

                      Comment


                      • #12
                        Originally posted by mark45 View Post
                        "A few orders of magnitude" means "at least a hundred times", do you really mean that?
                        Actually, common usage of 'order of magnitude' outside of astronomy generally means doubled or halved, so "a few orders of magnitude" would imply at least eight times (doubled three times). In astronomy, where the term originated, an order of magnitude is technically 2.512 (or its inverse, about .398) times so "a few orders of magnitude" would mean at least around 16 times (about 15.851). Sometimes people use 'order of magnitude' to mean 'ten times,' but it's neither the most common nor the most correct meaning.

                        Comment


                        • #13
                          Originally posted by CFWhitman View Post
                          Actually, common usage of 'order of magnitude' outside of astronomy generally means doubled or halved, so "a few orders of magnitude" would imply at least eight times (doubled three times). In astronomy, where the term originated, an order of magnitude is technically 2.512 (or its inverse, about .398) times so "a few orders of magnitude" would mean at least around 16 times (about 15.851). Sometimes people use 'order of magnitude' to mean 'ten times,' but it's neither the most common nor the most correct meaning.
                          Que? The 2.512 value comes from Apparant Magnitudes and not from Orders of Magnitude. An order of magnitude is base 10 which means that two orders of magnitude is 100x and three is 1000x.

                          https://en.wikipedia.org/wiki/Order_of_magnitude

                          Comment


                          • #14
                            Originally posted by F.Ultra View Post
                            Que? The 2.512 value comes from Apparant Magnitudes and not from Orders of Magnitude. An order of magnitude is base 10 which means that two orders of magnitude is 100x and three is 1000x.

                            https://en.wikipedia.org/wiki/Order_of_magnitude
                            I'm not as familiar with the mathematical application as the astronomical one. So you are correct about the common mathematical application. However, it is still not the only meaning. An order of magnitude in astronomy is still 2.512 times. Also, in general, any progression that is based on multiplication rather than addition can be referred to as orders of magnitude (sometimes this is referred to as geometric progression). It's not by any means always ten times (even the article you cited makes this clear).

                            Comment


                            • #15
                              Originally posted by CFWhitman View Post
                              Also, in general, any progression that is based on multiplication rather than addition can be referred to as orders of magnitude (sometimes this is referred to as geometric progression). It's not by any means always ten times (even the article you cited makes this clear).
                              In which case the onus falls squarely on the writer to define the term "order of magnitude". In my experience, the term means "power of 10". If the term "order of magnitude" is sufficiently ambiguous that a large percentage of users could genuinely misunderstand it, then it must be defined. If not, then we reduce the problem to what the 'average person' would interpret "order of magnitude" to mean.

                              For the record, I've never heard of subsequent terms in a geometric progression being referred to as orders of magnitude (for context, I have a Masters degree in Mathematics). It is entirely possible that my university simply didn't use this notation/description, but to suggest that it is standard is, I think, going too far. Descriptive terms such as these were inconsistent even within my own department, so what must they be like throughout the entire world?

                              As a final plea: when discussing matters of this nature, PLEASE don't cite Wikipedia. No academic would base conclusions on "the journal in which anybody can arbitrarily alter the contents", just as no reasonable doctor would base their medical advice on "the journal of what my friends have seen". Wikipedia is a good resource and what it says should not be dismissed out of hand, but it could almost be the reason for the mantra "trust, but verify.

                              Comment

                              Working...
                              X