Announcement

Collapse
No announcement yet.

Ubuntu 22.04 LTS Changes Default For NVIDIA Driver Back To Using X.Org Rather Than Wayland

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

  • Originally posted by skeevy420 View Post
    Basically, using gasoline and internal combustion engines is more environmental friendly than EVs and coal plants. And since it's basically established science that gasoline and internal combustion engines are bad for the environment, replacing them with EVs powered by coal that's 50-200% more dirty no matter how you fudge the numbers is clearly not the correct solution.

    Using green vehicles without creating green energy first is like putting a fresh coat of paint over black mold spores. You didn't fix anything; you just hid the problem from view.
    I believe those numbers are approximately right under certain conditions. They aren't relevant to what you'd see in a car, though, because you'll never get perfect efficiency. That doesn't even take into account transmission costs - it requires less power to transmit electricity through a wire than it does to ship gas around on a truck to petrol stations, for example.

    There are so many possible variables it's really complex to get an actual good comparison, but I believe I've seen studies that show an electric car on pure coal power is still slightly better than a gas powered car - but it's close enough that it can tip either way very easily depending on numerous local factors. But if, for example, you get 50% coal power and 50% non-coal, it's pretty easily in favor of the electric car at that point.

    You've also done things unrelated to CO2 like transfer all the smog from cars out of the city, and transformed it into pollution around a power plant, where it can hopefully be better managed and is farther away from large population centers, which is presumably good for health.

    Anyway, there are certainly numerous ecological issues you can point to with electric cars if that's what you want to do. There needs to be a big focus on battery recycling and clean mining going forward.
    Last edited by smitty3268; 25 April 2022, 09:25 PM.

    Comment


    • Originally posted by smitty3268 View Post

      I believe those numbers are approximately right under certain conditions. They aren't relevant to what you'd see in a car, though, because you'll never get perfect efficiency. That doesn't even take into account transmission costs - it requires less power to transmit electricity through a wire than it does to ship gas around on a truck to petrol stations, for example.

      There are so many possible variables it's really complex to get an actual good comparison, but I believe I've seen studies that show an electric car on pure coal power is still slightly better than a gas powered car - but it's close enough that it can tip either way very easily depending on numerous local factors. But if, for example, you get 50% coal power and 50% non-coal, it's pretty easily in favor of the electric car at that point.

      You've also done things unrelated to CO2 like transfer all the smog from cars out of the city, and transformed it into pollution around a power plant, where it can hopefully be better managed and is farther away from large population centers, which is presumably good for health.

      Anyway, there are certainly numerous ecological issues you can point to with electric cars if that's what you want to do. There needs to be a big focus on battery recycling and clean mining going forward.
      Also, times have already changed reagrding coal. For example in the US in 2021, on average coal is only 22% of "utility-scale electricity generation".
      Renewables (including hydro) are 20%, nuclear 19%, natural gas 38%.

      Comment


      • Originally posted by tildearrow View Post

        Didn't you say you wouldn't engage into any Wayland discussion anymore?

        Well then, what's this?
        I've not touched on Wayland in this topic in any shape or form. I have kept my promise. This is me arguing with people who are talking BS about completely orthogonal things.

        Comment


        • Originally posted by wertigon View Post

          The market is screaming for a working Wayland render path. Nvidia, thus far, cannot provide one. The only explanations I can find are that Nvidia are either incompetent, or are otherwise doing their best to exert willful malice. Would love to be proven wrong though.
          You have been proven wrong many times. The Linux graphics stack (i.e. mesa/gbm) are using a design that is so outdated that NVidia drivers don't even support it anymore and thus it has been posted many times.
          Last edited by mdedetrich; 26 April 2022, 06:42 PM.

          Comment


          • Originally posted by mdedetrich View Post

            You have been proven wrong many times. The Linux graphics stack (i.e. mesa/gbm) are using a design that is so outdated that NVidia drivers don't even support it anymore and thus gas been posted many times.
            Yes, it really is so technologically inferior that Mesa-based drivers on AMD Radeon cards are beating their Nvidia counterparts.

            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

            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


            If it was so technologically inferior as you claim Nvidia would be at least 10% better in each tier. Care to explain why it isn't?

            The only way to settle this properly though is to implement a mesa driver for Nvidia cards. But since we have already established that Nvidia doesn't really care about Desktop Linux, nor are unlikely to start caring anytime soon, well...

            Comment


            • Originally posted by wertigon View Post

              Yes, it really is so technologically inferior that Mesa-based drivers on AMD Radeon cards are beating their Nvidia counterparts.

              https://www.phoronix.com/scan.php?pa...rx6600xt-linux
              https://www.phoronix.com/scan.php?pa...rx6500xt-linux

              If it was so technologically inferior as you claim Nvidia would be at least 10% better in each tier. Care to explain why it isn't?

              The only way to settle this properly though is to implement a mesa driver for Nvidia cards. But since we have already established that Nvidia doesn't really care about Desktop Linux, nor are unlikely to start caring anytime soon, well...
              Rearead what everyone else has said about implicit vs explicit sync and then come back. The tl;dr since Linux at least currently only supports implicit sync for every part of the stack there is a performance penalty if drivers have to implement manual workarounds for implicit sync (which is the main reason NVidia drivers are slower but its only for Wayland, with X11 since by design NVidia has much more control they don't have the same problem). The contrary is not true, i.e. if the Linux stack fully supported explicit sync than you can implement implicit sync without any performance penalty (which is why explicit sync is far superior).

              Comment


              • Originally posted by mdedetrich View Post

                Rearead what everyone else has said about implicit vs explicit sync and then come back. The tl;dr since Linux at least currently only supports implicit sync for every part of the stack there is a performance penalty if drivers have to implement manual workarounds for implicit sync (which is the main reason NVidia drivers are slower but its only for Wayland, with X11 since by design NVidia has much more control they don't have the same problem). The contrary is not true, i.e. if the Linux stack fully supported explicit sync than you can implement implicit sync without any performance penalty (which is why explicit sync is far superior).
                Yes, I did reread. I think this is pretty much monolithic vs microkernel debate all over again. Linux is still a monolithic kernel and that proved to be the superior option ,though there are still Microkernel enthusiasts out there. The argument is pretty much philosophical tech wankery, and while interesting academically, it does nothing to solve any problems.

                You, on the other hand, have yet to supply a sufficient reason why Nvidia cannot create a mesa driver since this is obviously what they have to do to stay competitive in the near term. Long term, Linux is open source. If you need explicit sync, create an explicit sync render path. It is that simple. Meanwhile, implicit sync will rule the roost for the foreseeable future.

                Comment


                • Originally posted by wertigon View Post

                  Yes, I did reread. I think this is pretty much monolithic vs microkernel debate all over again. Linux is still a monolithic kernel and that proved to be the superior option ,though there are still Microkernel enthusiasts out there. The argument is pretty much philosophical tech wankery, and while interesting academically, it does nothing to solve any problems.
                  Implicit sync by definition is shit and outdated, this is a fact that even Linux kernel developers admitted (hence the lkml article created by a mesa developer). There is now an ongoing effort to fix this problem and NVidia said they will provide full support when Linux properly supports explicit sync (which actually some Linux kernel devs begrudgingly agreed).

                  BTW this implicit sync is causing lots of issues in the Linux kernel that is completely unrelated to NVidia, for example Linux is having issues with Vulkan because just like the NVidia driver it has a proper modern design that only supports explicit sync.

                  Originally posted by wertigon View Post
                  You, on the other hand, have yet to supply a sufficient reason why Nvidia cannot create a mesa driver since this is obviously what they have to do to stay competitive in the near term. Long term, Linux is open source. If you need explicit sync, create an explicit sync render path. It is that simple. Meanwhile, implicit sync will rule the roost for the foreseeable future.
                  I wouldn't create a driver in Nvidia's position either, they are being told to rewrite their entire driver to an outdated design just to satisfy stubborn Linux dev's that can't admit their they are using an outdated design from the 90's.

                  Comment


                  • Originally posted by mdedetrich View Post
                    Implicit sync by definition is shit and outdated, this is a fact that even Linux kernel developers admitted (hence the lkml article created by a mesa developer). There is now an ongoing effort to fix this problem and NVidia said they will provide full support when Linux properly supports explicit sync (which actually some Linux kernel devs begrudgingly agreed).

                    BTW this implicit sync is causing lots of issues in the Linux kernel that is completely unrelated to NVidia, for example Linux is having issues with Vulkan because just like the NVidia driver it has a proper modern design that only supports explicit sync.
                    If it is creating all of these problems... Where is the drive for change?

                    What is preventing a group of disgruntled developers from developing a render path that does support explicit sync, like, say, the PREEMPT_RT team did? Sure, a big task but not impossible by any stretch of the imagination. This can be developed in parallell.

                    Since implicit sync is so easy to implement over explicit sync, what is the problem with creating an implicit sync layer over Vulkan for Wayland, and then use direct render for direct vulkan rendering? (What Kopper is doing if I've understood it correctly)

                    Complaining about it here isn't likely to change anything, and 99% of all Linux users are unlikely to care about the finer tech wankery of Implicit vs Explicit sync.

                    Originally posted by mdedetrich View Post
                    I wouldn't create a driver in Nvidia's position either, they are being told to rewrite their entire driver to an outdated design just to satisfy stubborn Linux dev's that can't admit their they are using an outdated design from the 90's.
                    Bottom line, end user doesn't care about Implicit vs Explicit sync. To them that is just technobabble. What the end user *does* care about is why AMD and Intel runs Wayland just fine, and Nvidia does not. It is likely system integrators and OEMs like System 76 and Dell will prefer Intel and AMD GPUs over Nvidia for their Linux desktops and laptops, since that will just work.

                    Nvidia needs to write a Mesa driver or risk losing a ton of market share. Their 1990s style of providing drivers just isn't sufficient any more.

                    Comment


                    • Originally posted by wertigon View Post
                      If it is creating all of these problems... Where is the drive for change?

                      What is preventing a group of disgruntled developers from developing a render path that does support explicit sync, like, say, the PREEMPT_RT team did? Sure, a big task but not impossible by any stretch of the imagination. This can be developed in parallell.
                      That's what's going on. Hence Ekstrand: Plumbing explicit synchronization through the Linux ecosystem getting linked earlier.

                      Originally posted by wertigon View Post
                      Complaining about it here isn't likely to change anything, and 99% of all Linux users are unlikely to care about the finer tech wankery of Implicit vs Explicit sync.
                      People are complaining that nVidia doesn't want to throw a ton of effort into something akin to developing a new X11 driver now that Wayland exists.
                      Last edited by ssokolow; 27 April 2022, 08:53 PM.

                      Comment

                      Working...
                      X