Announcement

Collapse
No announcement yet.

NVIDIA 470 Series Driver Looks Like It Will Bring OpenCL 3.0 Support

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

  • #11
    Originally posted by entropy View Post
    Now that CUDA is dominant due to dubious business practices by Nvidia they now finally offer OpenCL 3 when it's no threat anymore to their strategy.

    (No, I don't believe there were real showstoppers due to technical issues)
    What's that "finally" about? OpenCL 3.0 spec was released in september 2020 (which is btw mentioned in the article). Dubious business practices of not supporting non existent specification? How's AMD doing on that front?

    Or did you mean OpenCL 2.x which was barely supported by anyone and ultimately scrapped by khronos themselves?

    Comment


    • #12
      Originally posted by RamblingMadMan View Post
      That's cool and all, but can my dGPU turn off when I'm not using it yet?
      If you've got an AMD card, yeah

      Comment


      • #13
        Originally posted by FireBurn View Post

        If you've got an AMD card, yeah
        I wish

        Comment


        • #14
          Originally posted by RamblingMadMan View Post

          Well that's great for you, but for everybody running a Pascal or earlier GPU it's a different story. You would think that an earlier generation of card would have better support than a new card, not worse; but I guess I shouldn't expect any better from NVIDIA's linux offering.
          Not to sound like a smart ass, but not really. Not with NVIDIA. Their GPUs and drivers are known for petering out in performance. We'll get a year or three out of one and then they do enough to keep it working and expect us to buy a newer model to get better performance.

          And those are old school Windows user anecdotes. The stuff I used to see people speculating about back in the days when I had NVIDIA GPUs and kept coming across Windows fixes for my Debian & Arch problems . I wouldn't know. I knew my shit sucked to begin with . MX 4s and Geforce 8s. Real top of the line cards there

          Comment


          • #15
            Originally posted by bofh80

            Are using Xorg or Wayland? I have a 1050 in my laptop i think it's powered down currently. (I think).

            I'm not really paying that much attention to it though. How do you know it's ON?

            I assume from the tone of the post that you understand to tell Xorg to use the Intel onboard or whatever etc.

            I had problem up till now unaware they were caused by my stupid dell dock with displaylink gpu that makes the nvidia one not work
            In my case I know the dGPU is still on because there is an indicator light for it at the hardware/firmware level (MSI GP62 6QF, i7-6700HQ, GTX 960M 4GB). Also, on my laptop running PowerTOP with bbswitch properly disabling the dGPU shows 7.74W power consumption, while letting the NVIDIA proprietary driver handle it (which it doesn't, it just leaves the dGPU running at the minimum frequencies) by itself shows 9.26W. In both cases I set PowerTOP to measure the power consumption for 4 minutes, and in both cases the desktop was running on the iGPU, with nothing else other than the terminal open.

            The real frustrating thing about this is that even nouveau can successfully handle turning off the dGPU on my laptop with a new enough kernel, hell even manual reclocking works since it's Maxwell 1 (so at least it's noticeably faster than the iGPU, yay), so I don't see any excuse for NVIDIA not being able to get proper power management working for the older cards, especially given that they do in fact have a generic driver for Windows that takes care of the power management for most laptops so they know how it generally works.

            Comment


            • #16
              Wow, cutting edge support from nvidia, finally they support the latest and greatest...

              Oh wait, wasn't v3 actually a step back to basically v1, forced exactly by that lousy greedy company, that made no efforts whatsoever to support v2, in an attempt to stifle competition to its own proprietary solution?

              Comment


              • #17
                Originally posted by RamblingMadMan View Post
                That's cool and all, but can my dGPU turn off when I'm not using it yet?
                You should try bumblebee! Most will tell you it's a hack, but that's the only think that works on my 540MGT.

                Comment


                • #18
                  Originally posted by entropy View Post
                  Now that CUDA is dominant due to dubious business practices by Nvidia they now finally offer OpenCL 3 when it's no threat anymore to their strategy.

                  (No, I don't believe there were real showstoppers due to technical issues)
                  The main reason that CUDA is so dominant is that the tooling and ecosystem is MUCH MUCH better than OpenCL (which is a joke in comparison). It also doesn't help that OpenCL 2 was a massive flop.

                  Of course "shady business practices" helped but it wouldn't have worked if the actual product wasn't better.


                  Originally posted by ddriver View Post
                  Wow, cutting edge support from nvidia, finally they support the latest and greatest...

                  Oh wait, wasn't v3 actually a step back to basically v1, forced exactly by that lousy greedy company, that made no efforts whatsoever to support v2, in an attempt to stifle competition to its own proprietary solution?
                  Don't blame this on NVidia, pretty much no one (even AMD) supported V2 because it was terrible in design.
                  Last edited by mdedetrich; 10 March 2021, 04:15 AM.

                  Comment


                  • #19
                    For now, Nvidia only supports OpenCL (v1.2) one linux x86. There is no support on ARM or PowerPC*. I hope this will change, since PoCL is often used on those platforms to use OpenCL code on Nvidia GPUs. By the way, kudos to the PoCL team !

                    I believe Nvidia will activate or make public some features they kept secret for some years now, like workgroup function (reduction, ...) from OpenCL2. OpenCL3 offers the ability to support only certain features and not everything ... since shared memory is something too complicated for them (AMD managed to do it :-)

                    * There is an OpenCL driver, it just lacks the compiler !

                    Comment


                    • #20
                      Originally posted by kieffer View Post
                      For now, Nvidia only supports OpenCL (v1.2) one linux x86. There is no support on ARM or PowerPC*. I hope this will change, since PoCL is often used on those platforms to use OpenCL code on Nvidia GPUs. By the way, kudos to the PoCL team !

                      I believe Nvidia will activate or make public some features they kept secret for some years now, like workgroup function (reduction, ...) from OpenCL2. OpenCL3 offers the ability to support only certain features and not everything ... since shared memory is something too complicated for them (AMD managed to do it :-)

                      * There is an OpenCL driver, it just lacks the compiler !
                      Nvidia did support SVM. Whole thing is that they only supported it on 64 bit mode in wierd "opencl 2.0 evaluation mode".

                      Also AMD implementation of SVM was also plagued with bugs (google search gives a lot of interesting results) pretty much only leaving Intel as only company supporting SVM correctly.

                      Comment

                      Working...
                      X