Nvidia/Ati - GPU roadmap

Discussion in 'Architecture and Products' started by overclocked, Jun 9, 2003.

  1. overclocked

    Veteran

    Joined:
    Oct 25, 2002
    Messages:
    1,317
    Likes Received:
    6
    Location:
    Sweden
  2. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,903
    Likes Received:
    218
    Location:
    Seattle, WA
    Is PCI Express really set to be ready by early next year?

    It will be interesting to see how video card manufacturers decide to handle the switch to PCI express....
     
  3. Freak'n Big Panda

    Regular

    Joined:
    Sep 28, 2002
    Messages:
    898
    Likes Received:
    4
    Location:
    Waterloo Ontario
    I really doubt .11 will by rdy at TSMC by mid '04.
     
  4. indio

    Newcomer

    Joined:
    Feb 1, 2003
    Messages:
    221
    Likes Received:
    0
    Location:
    Berkshire County , Ma USA
    i still have my doubts that .13 is ready. It's still not available on a large scale.
     
  5. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,903
    Likes Received:
    218
    Location:
    Seattle, WA
    Well, nVidia does have the IBM option open now.

    However, from what I've heard, the NV40 will still be on the .13 micron process.
     
  6. RussSchultz

    RussSchultz Professional Malcontent
    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    2,855
    Likes Received:
    55
    Location:
    HTTP 404
    TSMC is saying different. (Though, of course they would)
     
  7. Arun

    Arun Unknown.
    Moderator Legend Veteran

    Joined:
    Aug 28, 2002
    Messages:
    5,023
    Likes Received:
    299
    Location:
    UK
    I'm going to have fun quoting myself here :D

    From http://www.notforidiots.com/GPURW.php :

    Uttar
     
  8. 3dcgi

    Veteran Subscriber

    Joined:
    Feb 7, 2002
    Messages:
    2,436
    Likes Received:
    264
    Who's manufacturing the 9600's? They're available.
     
  9. LeStoffer

    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    1,253
    Likes Received:
    13
    Location:
    Land of the 25% VAT
    You have good reasons to be in doubt. Neither TSMC nor IBM will [ever] offer you any 0.11 process! :lol:
     
  10. Malo

    Malo Yak Mechanicum
    Legend Veteran Subscriber

    Joined:
    Feb 9, 2002
    Messages:
    7,087
    Likes Received:
    3,159
    Location:
    Pennsylvania
    one thing I would like to ask regarding the supposed move to PCI-Express boards in the near future according to the comparison chart. The limitation of AGP currently means we can only have one high-bandwidth video card in the PC at one time. With PCI-Express, I assume this means we can have mutliple display adapters running concurrently?

    With this in mind, could we possibly see a trend towards support for multiple monitors in games? Of course Matrox tried this but given the lack of cards in the market, developers basically either didn't care or couldn't justify the added development costs for such a small market, thus such a small amount of titles with the capability.

    But with the promise of multiple high-bandwidth GPU's in future systems, hopefully we will see more developement to multiple display support once PCI-Express becomes more mainstream?

    kalbaz
     
  11. Hellbinder

    Banned

    Joined:
    Feb 8, 2002
    Messages:
    1,444
    Likes Received:
    12
    Yes and ATi could do the exact same thing. There are other options as well.
     
  12. Basic

    Regular

    Joined:
    Feb 8, 2002
    Messages:
    846
    Likes Received:
    13
    Location:
    Linköping, Sweden
    Kalbaz:
    If I haven't got it completely wrong, the typical PCI-Express motherboard would still only have one AGP-like slot. The typical configuration would be a couple of single channel slots (PCI-replacers), and one 16 channel slot (AGP-replacer). The market is probably too small to make chipsets with two 16 channel ports.

    The best I can hope for (and I'm only dreaming here) is a chipset that can split its "AGP-replacer" into two smaller ones. How about this idea: motherboard has two 16-channel slots, but there's only two 8-channel ports from the chipset (one to each slot). The remaining 8 channels are connected between the two slots.

    If you have two gfx cards, you have 2.5 GB/s in each direction to each card, and the cards can talk to each other with 2.5 GB/s in each direction (good for transfering results from render-to-texture).

    If you have one gfx card, you put a dummy in the other slot, that connects the two 8-channel ports. That way you get a 16-channel port to your single card (5 GB/s in each direction).

    Warning: that was just brainstorming, it could be impossible to do it that way, and/or there could be much better ways.
     
  13. T2k

    T2k
    Veteran

    Joined:
    Jun 12, 2002
    Messages:
    2,004
    Likes Received:
    0
    Location:
    The Slope & TriBeCa (NYC)
    No way, it's BS.

    First of all, it's still in paper-phase: http://www.pcisig.com/specifications/pciexpress/
     
  14. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,903
    Likes Received:
    218
    Location:
    Seattle, WA
    Yeah, that's basically what I thought.

    From what I've heard previously with PCI Express, here's what I basically expect:

    1. First step will be to integrate the lowest-bandwidth form of PCI Express. From what I read, this could be implemented as a small bracket between the PCI slot and the edge of the motherboard, allowing the motherboard to support both types of cards.

    2. Once adoption of PCI Express is high enough, start moving toward higher-bandwidth PCI Express forms.

    3. In the meantime, motherboards may contain one PCI Express high bandwidth slot below the AGP slot, primarily for the graphics cards.

    4. How high-end graphics cards will handle the change I still don't know. I expect that at first, most will be available in both formats, but within a year after adoption, most of the brand-new graphics cards will only be available in PCI Express format.
     
  15. overclocked

    Veteran

    Joined:
    Oct 25, 2002
    Messages:
    1,317
    Likes Received:
    6
    Location:
    Sweden
    Are you sure there "exist" no 0,11nm?
    I don´t know how the manufactoring works but can´t you compare it to that it´s much smarter too learn chemistry A(0,13nm) and then B(0,11nm) instead if jumping direct to the C-grade(0,09nm).
    It sounds smarter for someone like me that not know the hardcore facts.

    edit spelling
     
  16. RussSchultz

    RussSchultz Professional Malcontent
    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    2,855
    Likes Received:
    55
    Location:
    HTTP 404
    There was an intermediate step between .25u and .18u (.22u) and between .18u and .13u (.15u). However, there really hasn't been too much talk about an intermediate step between .13u and .09u. Doesn't mean it isn't happening, but its not listed on any of the fabs' websites.
     
  17. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    648
    Location:
    O Canada!
    Aren't there a number of memory modules being produced in 110nm?
     
  18. Entropy

    Veteran

    Joined:
    Feb 8, 2002
    Messages:
    3,069
    Likes Received:
    1,028
    Samsung.
    Process level/quality/sophistication does not really let itself be described by a single figure of merit.

    Entropy
     
  19. Natoma

    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    1,913
    Likes Received:
    84
    Huh. Then how do you explain PCI-Express being in Intel's Grantsdale chipset for 1H'04? Actually specifically 2Q'04, but there are rumors that if AMD's athlon64 garners some heavy competition in the desktop space, Intel will move Grantsdale and Tejas up to 1Q'04. But that's hearsay.

    1066-FSB (eventually 1200-FSB), Serial ATA/USB2.0/Firewire ports out the arse, PCI-Express x16 (aggregate bandwidth of 32Gbit/sec. 2Gbit/sec per channel), DDR-I and DDR-II support, plus a few more nifty things.

    That's what I'm waiting for. :)
     
  20. gordon

    Newcomer

    Joined:
    May 22, 2003
    Messages:
    76
    Likes Received:
    1
    IBM already has the capacity to produce 110nm devices and currently they are too. Dunno how that would play into Nvidias plans.

    Inq reports that intels next platform for the next generation Prescott and Tejas processors, to be released 2Q'04, will use it.
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...