Announcement

Collapse
No announcement yet.

One Million Dollars For A Shader-Based LGPLv3 GPU

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

  • #31
    Originally posted by fbruno View Post

    (...)

    3) We didn't try to do hardware for a few reasons. I don't think we could have a one size fits all approach. The 2D part will fit in a small FPGA. The 3D part needs more. I think hackers might want a Cyclone V SOC or Xilinx Zynq to play with. I personally would love the time to take the Sparc T1/T2 and pair it with this and see if I could squeeze it into a Stratix V or Vertex 7, but they would cost 10K a piece.

    (...)
    As for the Xilinx ZYNQ, I backed the Red Pitaya project
    http://www.kickstarter.com/projects/...s-for-everyone
    which comes with a ZYNQ 7010 AFAIK.
    Do you expect the 2D core to fit on that FPGA?

    Comment


    • #32
      Originally posted by Kivada View Post
      Opensource GPU using 2001 tech that nobody will buy or hire more devs to make better drivers for existing hardware that people are already buying?
      Why do you keep on insisting on that comparison? Do you think a driver is somehow exchangeable with a hardware design?

      Comment


      • #33
        Originally posted by Kivada View Post
        The patents expired on i586, there are now unlicensed Intel compatible i586 chips on the market. The same can be done for anything once it's out of patent.
        No one is saying otherwise. Please go back and reread both bridgman and my comments. The point its not that you can't release a compatible chip. The point is that anyone releasing i586 chips has either created the chip design from scratch (spending lots on money in the process) or more likely has paid someone for the chip design.

        This Kickstarter means anyone could reuse the GPU design for free and contribute to the GPU's improvements.

        Originally posted by Kivada View Post
        This GPU, open as it may be has no potential market, especially when you se the limits of where they say they can take it's tech on their Kickstarter page. Even ARM GPUs have complete modern features unlike this project's high end goals that will only be 2001 era OpenGL1.4/DirectX8.1 features that are not even enough to run a composited desktop on anymore since those require OpenGL 2.*.

        Opensource GPU using 2001 tech that nobody will buy or hire more devs to make better drivers for existing hardware that people are already buying?
        Again you don't seem to be reading (or understanding) things properly. The current 3D (DirectX7/8) option is only stretch goal 1 and is about releasing existing tech. The high end goal states:
        This is our ultimate stretch goal and requires a complete redesign. It's something we have been wanting to do for years, but didn't have the resources. This would allow us to create a complete open source implementation of a modern day graphics accelerator.

        If we receive more than the above, it will allow us to devote more time and effort to the project and we'll be able to release code sooner.

        This is new design work and our anticipated delivery would be Q2 2015.
        All in all $1 million for all this seems like a good deal to me.
        Last edited by tarceri; 10-11-2013, 06:23 AM.

        Comment


        • #34
          I would like that but it must be good enough to Steam

          OS GPU, thank you.
          Imagine Game developers asking for features and contrubuting to improve drivers.
          Imagine prices like ARDUINO or RASBERRY PI
          Imagine also a great OS CPU + GPU project

          But it must run Steam I mean GAMES at least as an actual 150 USD AMD or Nvidia GPU

          And it must have INCOME, donations to the project from brands that use it in the form of eng time and money.

          If some companies can add workers to OS GPUs, the actual oligopoly will break.

          Comment


          • #35
            Originally posted by Kivada View Post
            The patents expired on i586, there are now unlicensed Intel compatible i586 chips on the market. The same can be done for anything once it's out of patent.
            Just as everyone else has been saying, it seems you miss they're not clones. They are reimplementations from scratch - huge difference in work and cost!

            There is no open by-Intel 586 implementation out there, and the patents expiring merely let others use the 586 _instruction set_.

            Comment


            • #36
              Originally posted by mitcoes View Post
              OS GPU, thank you.
              Imagine Game developers asking for features and contrubuting to improve drivers.
              Imagine prices like ARDUINO or RASBERRY PI
              Imagine also a great OS CPU + GPU project

              But it must run Steam I mean GAMES at least as an actual 150 USD AMD or Nvidia GPU

              And it must have INCOME, donations to the project from brands that use it in the form of eng time and money.

              If some companies can add workers to OS GPUs, the actual oligopoly will break.
              Sure. You can easily get Nvidia level GPU, you only need Nvidia level R&D budget.
              Which are around 800 MUSD a year, a perfect target to start with any project.

              Comment


              • #37
                Originally posted by entropy View Post
                As for the Xilinx ZYNQ, I backed the Red Pitaya project
                http://www.kickstarter.com/projects/...s-for-everyone
                which comes with a ZYNQ 7010 AFAIK.
                Do you expect the 2D core to fit on that FPGA?
                Sorry for taking so long to respond.
                2D without the VGA (Only needed for true VGA applications, i.e. booting legacy PCs) will fit nicely in a 7010. The VGA may fit also, but it would be tight and not really necessary as the 2D part is much higher performance.

                Comment


                • #38
                  Originally posted by erendorn View Post
                  Sure. You can easily get Nvidia level GPU, you only need Nvidia level R&D budget.
                  Which are around 800 MUSD a year, a perfect target to start with any project.
                  We have no intention of ever competing against AMD or Nvidia for the 800M/yr reasons you mention above.
                  My goal is to give back to the community and try to get the project to the next level if possible. We'll still work on a shader architecture. We are hardware hackers and digital designers at heart. The only question is if we get their via the open or closed source route. I'd much rather do it this way, be open about it and let the community benefit.

                  Probably a bad analogy, but Linus released the linux Kernel 20 years about almost. Look what it has become. Maybe a similar renaissance can happen in hardware.

                  Comment


                  • #39
                    Originally posted by fbruno View Post
                    Sorry for taking so long to respond.
                    2D without the VGA (Only needed for true VGA applications, i.e. booting legacy PCs) will fit nicely in a 7010.
                    The VGA may fit also, but it would be tight and not really necessary as the 2D part is much higher performance.
                    Thank you!

                    Comment


                    • #40
                      Originally posted by Kivada View Post
                      I'm saying the money would be better spent hiring more full time Gallium3D devs.
                      I think the money would be better spent on feeding starving children in 3rd world countries. However, I'm not sure how that relates in any way to this kickstarter campaign. It's unlikely people contributing would want their money going to some unrelated task, or they would have sent it there in the first place.

                      Comment


                      • #41
                        I don't see the point of this.

                        FPGA can be great for solving the problems that one can't solve with custom-made silicon, but once it has to compete with it 1:1, FPGA sorely looses.

                        So, since GPU is so area and power intensive and since it has great players on the field that are opening gradually their documentation ( AMD and nVidia, even Intel) I fail to see any advantage of this.

                        Even if you open the documentation and let's say you come performancewise sufficiently close for someone to choose your solution, how would s/he escape patent suit ?

                        main players already have bunch of patents and capable legal teams for battle if need arises.

                        What do you do if you get nice and cosy letter from their legal theams ?

                        Comment


                        • #42
                          [quote]main players already have bunch of patents and capable legal teams for battle if need arises.

                          What do you do if you get nice and cosy letter from their legal theams ?{/quote]

                          As long as it's in a FPGA, it's software, and thus hardware patents do not apply. So the free world doesn't need to care, SW patent countries like USSA need not apply.

                          Comment


                          • #43
                            I'm not so sure you could defend that on your own against legal team.

                            What exactly is a software ? A small blob that you download into FPGA ? But that blob is not SW that gets executed but bistream that configures HARDWARE.

                            Comment


                            • #44
                              I don't know if there are any precedents, quick googling didn't reveal any. But I would argue that since I can make it do other things tomorrow, like calculating pi, by typing on my computer, it is software.

                              Then there's the selling aspect too. If you only keep it in a repo somewhere, without even selling FPGAs (or selling empty FPGAs and telling users how to load it), you should be in the clear: see ffmpeg/mplayer which are affected by approx 10^7 software patents.

                              Comment


                              • #45
                                Suuestion

                                I would be more interested to see a project like Lucid Virtu MVP.
                                http://www.lucidlogix.com/product-virtu-mvp.shtml

                                I would donate $50 to $100 on such projects.

                                Comment

                                Working...
                                X