1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA GPU Lineup for 2009 Tabled

Discussion in 'News' started by btarunr, Dec 27, 2008.

  1. DarkMatter New Member

    Oct 5, 2007
    1,714 (0.47/day)
    Thanks Received:
    Those are not the names of the cards, but the codenames of the chips. Just like G92 > G94 > G96 > G98. OR G71 > G73. NV40 > NV43.

    It's been always like that and IMO it just has to do with the number of the project that was chosen to continue into production. Pretty much like with US fighters, the F-22 could have been the F-21 if that was the project that'd won between the two. And maybe I'm mistaken but the F-15 and the F-16 projects started at the same time and F-15 > F-16.

    In the case of GPU tesselation and the one that Ati has right now is just SUBDIVISION, that is subdivide the polygons into more polygons. It has nothing to do with textures nor with repeating paterns, etc.

    It does not really save computing power either, at pure rendering end I mean, because resulting triangles have to be calculated, textured, lit, etc. It does save a lot of memory, something that could make data transfers easier. And it definately makes animation easier: as long as you have enough vertex to make detailed animations, the less vertex the better, so that's something that is potentially keeping game models at "relatively low" poly count. The idea behind using tesselation is to have 20.000-50.000 poly models and make them look as if they had 100.000++ or millions of them. No more spikes in the heads. :)

    Tesselation could essentially be used to make models LESS detailed too and get rid of LOD models, something that could be very useful too.

    The end goal is to have models that vary on detail depending on the distance or simply the power of the GPU or the settings you use, without the need of LODs and the jumps between them that are so apparent sometimes.

    In any case the simple use of tesselation has like a constant performance penalty, much like DX10 has, but more pronounce probably, so the use of it in current games is questionable. See it as phone contracts with a high constant component and a small price per spoken minute. They are not good unless you will speak a lot AND you can pay that much. Current GPUs probably can't "pay" as much still.

    EDIT: Oh and on-topic I don't take that chart seriously. That roadmap is VERY different from other expectations/rumors and suggested roadmaps and without any proof to make a decision, which one is to believe...
    Last edited: Dec 28, 2008
    10 Year Member at TPU
  2. 3870x2


    Feb 26, 2008
    4,870 (1.38/day)
    Thanks Received:
    Joplin, Mo
    it DOES pull less juice, that would be common sense.
  3. eidairaman1

    eidairaman1 The Exiled Airman

    Jul 2, 2007
    18,503 (4.92/day)
    Thanks Received:
    so to me that would be an advantage when you can run 2 or 3 of them and not have to get a higher level PSU.
    10 Year Member at TPU

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)