Announcement

Collapse
No announcement yet.

NVIDIA Announces MONAI Open-Source AI Project

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

  • #11
    The reality is that years ago Nvidia made a very significant gamble, they started making a huge investment into developing GPU Compute technology. They were told they were crazy... who would want to use a GPU to compute things when you had CPUs... and the GPU could never run an OS so what good was it anyway for computing?

    Now they own the GPU computer ecosystem. AMD is making several half hearted attempts to counter Nvidia's hold on the high ground, but AMD cant even come out with a coherent platform (ex. OpenCL, Rocm, etc.)!

    Team green made a long bomb pass, and they caught it for the touch down now. While I would love for them to open source their driver, its to tightly integrated into what makes them huge sums of money now. They took the risk, they are going to take the reward, and by all rights of reason they should.

    AMD really should make an actual investment in GPU compute rather than these half baked plugs to say they are, but they never work.

    I would love Nvidia to open up... but I would also love Microsoft to do the same with Windows! Neither is going to happen for a long long time.

    I find it ironic though that everyone wines about Nvidia on Linux and demand that they open source... if Nvidia didnt have the best performance all this wining wouldn't happen I dont think :P

    Comment


    • #12
      It would be great if we could figure out how to build an army of computers that use this to reverse engineer the nvidia driver.

      Comment


      • #13
        Originally posted by onlyLinuxLuvUBack View Post
        It would be great if we could figure out how to build an army of computers that use this to reverse engineer the nvidia driver.
        Just a matter of time. AI algorithms are getting better and better. The irony would be using nvidia cards to crack it

        Comment


        • #14
          Originally posted by zexelon View Post
          ...
          Now they own the GPU computer ecosystem. AMD is making several half hearted attempts to counter Nvidia's hold on the high ground, but AMD cant even come out with a coherent platform (ex. OpenCL, Rocm, etc.)!...
          From my understanding, for quite a while AMD Rdeon were the preferred hardware, even though their software was circumspect. GCN worked well with compute.

          Comment


          • #15
            Originally posted by bridgman View Post

            AFAIK PyTorch has been running on ROCm/HIP for quite a while now.
            Still preliminary/experimental.

            Comment


            • #16
              Originally posted by thelongdivider View Post
              You can't hate on nvidia for developing a highly sophisticated API for parallel linear algebra.
              Are you kidding? This is Phoronix. The mob will work themselves into a frenzy as soon as they read "Nv", whether there's a good reason or not.

              Comment


              • #17
                Originally posted by tildearrow View Post
                The good news: It's open-source
                The bad news: Uses PyTorch (therefore CUDA) and requires an NVIDIA card.

                NVIDIA - fooling the open-source community since 1993
                By the same logic, you shouldn't be using a PC, because, no matter the license your software carries, there's still a blob in the microcode/firmware.

                Plus, CUDA is not a hard requirement, but look how many likes your FUD earned you!

                Comment


                • #18
                  Originally posted by DanL View Post

                  Are you kidding? This is Phoronix. The mob will work themselves into a frenzy as soon as they read "Nv", whether there's a good reason or not.
                  LOL so very true! The only thing that could be more contentious is if Steve Ballmer started helping Michael do OS benchmark comparisons in his spare time.

                  Comment


                  • #19
                    Originally posted by zexelon View Post
                    The reality is that years ago Nvidia made a very significant gamble, they started making a huge investment into developing GPU Compute technology. They were told they were crazy... who would want to use a GPU to compute things when you had CPUs... and the GPU could never run an OS so what good was it anyway for computing?

                    Now they own the GPU computer ecosystem. AMD is making several half hearted attempts to counter Nvidia's hold on the high ground, but AMD cant even come out with a coherent platform (ex. OpenCL, Rocm, etc.)!

                    Team green made a long bomb pass, and they caught it for the touch down now. While I would love for them to open source their driver, its to tightly integrated into what makes them huge sums of money now. They took the risk, they are going to take the reward, and by all rights of reason they should.

                    AMD really should make an actual investment in GPU compute rather than these half baked plugs to say they are, but they never work.

                    I would love Nvidia to open up... but I would also love Microsoft to do the same with Windows! Neither is going to happen for a long long time.

                    I find it ironic though that everyone wines about Nvidia on Linux and demand that they open source... if Nvidia didnt have the best performance all this wining wouldn't happen I dont think :P
                    1) As per usual with Nvidia fanbois and trolls, you are talking out of your behind since you have no actual tech knowledge other than the pop-tech-news you read here and there on the internet. For example you ignore that AMD had also a gpgpu framework, announced almost a year before CUDA was introduced, it was named CTM (Close to Metal). So no, Nvidia wasn't (yet again) innovative, in fact they were late to the party, but they (yet again) had a stranglehold on media and brand loyalty so their solution got much more support. CTM was then depreceted in favor of OpenCL, which by the way is not AMD's, it is Apple's....

                    2) CUDA actually sucks. AMD's Heterogenous architecture is better. The only reason CUDA dominates is because CUDA dominates. AKA the "windows problem", and any other established product, it is hard for a competing tech to displace it, because so many people are using it. The only thing that CUDA has that is actually better is documentation and dev tools, but again, those are better because the ecosystem is more mature because -you guessed it- more people are using it.

                    3) People do not whine about NVIDIA. Most non-Nvidia users don't give a single flying fsck about nvidia. It is nvidia users who complain about Wayland all the time and spread FUD that "wayland isn't ready" because nvidia is trash and don't want to properly support it. If performance was the issue, we would be buying nvidia, but we don't. We are buying intel and amd because they are supporting open source drivers and a superior Wayland experience.

                    Comment


                    • #20
                      Originally posted by DanL View Post

                      Are you kidding? This is Phoronix. The mob will work themselves into a frenzy as soon as they read "Nv", whether there's a good reason or not.
                      Dunno about you but have you noticed that video card prices have tripled at the high end, and that Nvidia doesn't balk at selling you a "Titan" card for 3 grand? That is gimps fp64 or fp16 on "consumer" cards that have exactly the same chips in them than Teslas and Quadros that cost 5-10x more? That it disallows you from using your own hardware that you paid for, in the way you want?

                      See for me, if the mob helps to break the monopoly, and takes down Intel-style oversegmentation, then yay the mob.
                      Last edited by vegabook; 21 April 2020, 07:42 PM.

                      Comment

                      Working...
                      X