Slashdot Mirror


NVIDIA Sues 3dfx For Patent Infringement

David D writes: "Apparently NVIDIA has filed a patent infringement lawsuit against 3dfx. The CEO of NVIDIA commented, ``We have always been on the forefront of innovation in 3D graphics technology and visual computing...''. The competition has been pretty even, with 3dfx having no apparent advantage over NVIDIA. Where will this lead?"

3 of 187 comments (clear)

  1. More bad patents or a real case of IP theft? by tjwhaynes · · Score: 5

    The patents that NVIDIA is suing boil down to methods for efficient I/O - in the words of Derek Perez of NVIDIA

    3dfx infringes on at least 5 patents dating back as far as NV1. All 5 patents are essentially I/O patents relating to efficiency of the interaction between the graphics processor and the core logic, memory or CPU.

    But like all patents, these are't easy reading. Trying to get to the essence of the method isn't easy. But here goes anyway:

    Patent 6,092,124 can be sumarised as being very similar to a local cache - a DMA sits next to the I/O bus and acts as a buffer for passing information over the I/O bridge or back down to system memory depending on the value of pointers held in the DMA. To me this does not sound very original - it sounds like a primitive level 2 cache.

    Pate nt 5,758,182 This one is an autonomous (of the OS) memory manager - it maps virtual addresses to physical ones. It uses pages to map memory and holds structures keeping tabs on that memory. Hardly mind blowing stuff. The main swansong of this patent is that it does this without the need for the OS to be involved, but I strongly suspect this is a commonly used technique and hardly worthy of patent protection - indeed for an autonomous device like a graphics card I think it would be difficult to avoid coming up with something like this regardless of your prior knowledge - you have to have something managing the memory on the card and it has to live with getting it's info from the application because most OS's won't necessarily be aware of the memory configuration on the board. In fact, the only work around for this patent as far as I can see is to expose the memory to the OS and let it use it as it sees fit. I have used one system where the VRAM could be used as system memory (Acorn RiscPC) but there was no hardware acceleration on that system. As soon as the GPU does any work on the memory at its fast IO busses to that graphics card RAM, there would have to be negotiation between the OS and the card to update the page tables on memory and that would hamstring a GPU card.

    I could go on but there are other people here on Slashdot who can do a better hatchett job on these patents. But these patents strike me as being 'obvious'. And I'm an NVIDIA card owner too so I'm not some disgruntled 3dfx owner with an axe to grind.

    Cheers,

    Toby Haynes

    --
    Anything I post is strictly my own thoughts and doesn't necessarily have anything to do with the opinions of IBM.
  2. Just as I expected from Slashdot by Temporal · · Score: 5

    This message is addressed to the Slashdot collective. That is, the large group of vocal Slashdotters who all think alike and act alike, and all carry the same extreme and often incorrect biases and predjudices. If the reader is not one of these people, then please do not take this personally.

    Yep, yep, yep. There they go again. The Slashdot collective. You guys are just too damned focused on politics. Why? Politics are stupid and futile, as I have discovered. (Case and point, the coming US election...)

    Look, my point is, you all see "NVidia sues 3dfx" and instantly most of the people on Slashdot assume NVidia is evil. Most of you have considered NVidia to be evil for quite some time (since they decided not to give their extremely-high-quality drivers to their competitors for free (GL drivers have a lot of hardware-independent code it them!)) and more evidence is just what you want. Woohoo, rally behind 3dfx, they're saints!

    Well, what if I told you that two years ago 3dfx sued NVidia over a patent the held on multitexturing? That's right: The act of applying more than one texture to a surface. 3dfx patented it. Then they sued NVidia. I would have a link to a news article about it, but my internet connection is on the fritz. Please see Linuxgames for a link to such an article.

    So, now who is the evil one? Answer: They BOTH are! Almost every large corporation on this planet has done something evil. Those that don't go out of business.

    If you try to judge a corporation on any sort of principles, you are likely to judge incorrectly. Personally, I have given up on judging such things. It is futile. All that matters to me now is who makes the best hardware. And for me, a 3D game engine writer, that is NVidia. If you want to buy 3dfx, fine. I don't care anymore. But buy for the hardware, or buy for the software. Don't buy on principle. It doesn't work.

    Obviously, I expect this to be moderated down as flaimbait. Do your worst. My karma has been maxed out for quite some time.

    ------

  3. What are these patents? by Chas · · Score: 5

    The patents in question are:

    • Register array for utilizing burst mode transfer on local bus (5,687,357)
    • Apparatus adapted to be joined between the system I/O bus and I/O devices which translates addresses furnished directly by an application program (5,721,947)
    • DMA controller translates virtual I/O device address received directly from application program command to physical i/o device address of I/O device on device bus (5,758,182)
    • Method and apparatus for accelerating the transfer of graphical images (6,023,738)
    • Method and apparatus for accelerating the rendering of images (6,092,124)

    I'm surprised that nobody ever considers the possibility of parallel development tracks.


    Chas - The one, the only.
    THANK GOD!!!

    --


    Chas - The one, the only.
    THANK GOD!!!