Announcement

Collapse
No announcement yet.

Open-Source "Nouveau" Driver Now Supports NVIDIA Ampere - But Without 3D Acceleration

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

  • #51
    Originally posted by mdedetrich View Post
    I do this as well, zero problems (if by text mode you mean TTY)
    There is a documented race condition still not fixed in the current day Nvidia drivers. So that just you being lucky and not having the right/wrong loads on the GPU when you do the switch to TTY. If a person workflow is the right pattern Nvidia is the worst thing on earth on Linux. Hopefully with DMA BUF support coming they will fix up the buffer sharing issue that causes this fail.

    Comment


    • #52
      Originally posted by oiaohm View Post

      There is a documented race condition still not fixed in the current day Nvidia drivers. So that just you being lucky and not having the right/wrong loads on the GPU when you do the switch to TTY. If a person workflow is the right pattern Nvidia is the worst thing on earth on Linux. Hopefully with DMA BUF support coming they will fix up the buffer sharing issue that causes this fail.
      I don't know if you are talking about https://bugs.launchpad.net/ubuntu/+s...s/+bug/1275760 but it was fixed 7 years ago and I also don't know whats the point of this nitpicking you are doing. Its not like AMD cards are free from bugs (far from it).

      Comment


      • #53
        Originally posted by mdedetrich View Post
        I don't know if you are talking about https://bugs.launchpad.net/ubuntu/+s...s/+bug/1275760 but it was fixed 7 years ago and I also don't know whats the point of this nitpicking you are doing. Its not like AMD cards are free from bugs (far from it).
        I am not talking about that bug I am in fact talking about faults that are not fixed.

        There are multi current bugs that people report that are basically all the same thing. Nvidia memory management screws up then it screws up the GPU that then nicely locks up in ways that the auto GPU reset does not fix.

        These are fatal system stopping bugs that the Nvidia closed source driver is suffering form. There are many people hitting these bugs mdedetrich I am not saying AMD is bug free but is also not correct to make out Nvidia is bug free. Nvidia closed source drivers on Linux have some very savage system stopping bugs. Yes it possible to have a workload where you are hitting the Nvidia bugs every 10 mins forcing a off at wall power reset of course if you are having that happen you cannot be productive at all.

        Comment


        • #54
          An amazing amount of hatred against the vendor no one must be using here in this thread. And if you're still using it you're not a Linux zealot enough, so please kindly f' off.

          In short Open Source fan-atics continue to demonstrate why companies which refuse to spend millions of dollars on a God forsaken OS (I'm talking about its desktop share) shouldn't really bother with supporting Linux more than nominally even through God forbidden blobs.

          NVIDIA is a very successful multi-billion business and most of you are no ones. Maybe step back a little and have some modesty.

          P.S. A happy and content NVIDIA user using a horrible blob under outdated X.org.

          Comment


          • #55
            Originally posted by oiaohm View Post
            These are fatal system stopping bugs that the Nvidia closed source driver is suffering form.
            By definition they are not fatal system stopping bugs otherwise literally no one would be using NVidia on Linux which is far from the case (ergo stop exaggerating). There are also plenty of similar bugs of same severity on AMD's side although there is no point in digging these up because the discussion is not going to get anywhere. Also from the bug reports you listed, it was a regression so its likely it will get fixed at some point in the future (welcome to software development, shit like this happens).

            Again what is your point here? No one is saying that Nvidia is bug free, but that a lot of people (including myself) have been using NVidia without any real problems for non trivial use cases (I have personally been using them for a decade). In fact the only thing that I can thing of that was major is the whole optimus/hybrid issue but then again at the time AMD didn't even have an equivalent solution (or it was a joke, can't remember now) and its working today anyways (I have optimus prime offload on my laptop).

            Comment


            • #56
              Originally posted by oiaohm View Post

              I am not talking about that bug I am in fact talking about faults that are not fixed.

              There are multi current bugs that people report that are basically all the same thing. Nvidia memory management screws up then it screws up the GPU that then nicely locks up in ways that the auto GPU reset does not fix.

              These are fatal system stopping bugs that the Nvidia closed source driver is suffering form. There are many people hitting these bugs mdedetrich I am not saying AMD is bug free but is also not correct to make out Nvidia is bug free. Nvidia closed source drivers on Linux have some very savage system stopping bugs. Yes it possible to have a workload where you are hitting the Nvidia bugs every 10 mins forcing a off at wall power reset of course if you are having that happen you cannot be productive at all.
              I used a GTX 1060 for about 5-6 months last year and never experienced what was described in the Reddit thread.

              I'd be confident that all drivers have some sort of "fatal system stopping bugs" that pop-up in different scenarios. I know amdgpu isn't perfect (personally had a few random lock-ups last year with 2 RX 580s), radeonsi isn't (had half a year where I couldn't play videos without random complete system freezes on my HD7850), and neither is i915 (had to redo a relative's laptop and put Windows back on it a few days ago because of instability with a kernel update; as bad as rep as Windows might get, I've never seen it randomly freeze without an obvious reason like OC'ing). I haven't had NVIDIA graphics for a large amount of time though, but my time with using them has been flawless.

              If NVIDIA's proprietary driver was really as unusable as people here are making it out to be, it likely wouldn't exist (at least for general users and gaming), and definitely wouldn't be supported on mainstream distros. If open-source was as-beneficial as people are also making it out to be, NVIDIA would have done it. I'm willing to bet NVIDIA knows more about what they're doing and why than some angry forum users though

              (offer still stands if someone wants to trade me a comparable NVIDIA GPU for a RX 580)

              Comment


              • #57
                Originally posted by mdedetrich View Post
                By definition they are not fatal system stopping bugs otherwise literally no one would be using NVidia on Linux which is far from the case (ergo stop exaggerating). There are also plenty of similar bugs of same severity on AMD's side although there is no point in digging these up because the discussion is not going to get anywhere. Also from the bug reports you listed, it was a regression so its likely it will get fixed at some point in the future (welcome to software development, shit like this happens).
                Really did you not look closely. That fail is the 7 year old bug you point to in Ubuntu. 7 years ago that fault was fixed by altering the work flow not fixing the memory management issue in the driver. Also its not a regression either.

                Originally posted by Espionage724 View Post
                I used a GTX 1060 for about 5-6 months last year and never experienced what was described in the Reddit thread.
                With these race condition class bugs unless you do the right set of actions they don't happen. You can use something for years and hot hit them. Problem is some users will have a workflow that doing the right conditions over and over again and have it going splat.

                Horrible part here is these faults with the Nvidia driver go back 18 years and its never been properly fixed. The result of people having a workflow hit this bug is absolute hate of Nvidia. I hope that with the work to introduce DMA BUF into the Nvidia driver that Nvidia finally fixes up the memory management problem. Its how buffers are shared between things is what allows the race condition that leads to complete GPU processing death.

                Yes there are tones of similar bugs on the AMD side but on Linux they are fixed by fixing the driver properly not by altering the workflow to avoid them. There are quite a few core bugs in the Nvidia binary driver that have been masked over that finally need to be fixed properly.

                Comment


                • #58
                  NVidia fuck you !! still stands.

                  Comment

                  Working...
                  X