Nvidia G71 - rumours, questions and whatnot

Discussion in 'Pre-release GPU Speculation' started by ToxicTaZ, Dec 4, 2005.

Thread Status:
Not open for further replies.
  1. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,928
    Likes Received:
    230
    Location:
    Seattle, WA
    Sure, but at least closing the case nearly eliminates the sound. I know it does in my case, at least (and I have two GT's). With a dual-slot cooler this is not an option, and every review I've read notes that the 7800 GTX 512MB is quieter. And it's a better cooler.
     
  2. Skrying

    Skrying S K R Y I N G
    Veteran

    Joined:
    Jul 8, 2005
    Messages:
    4,815
    Likes Received:
    61
    ATi didnt put nearly the amount of effort into 0.11u so I dont know how you could really compare it to Nv in that respect. I think ATi did jump the gun a little with 90nm, but oh well. I think the second time around we'll see them get a lot out of it.
     
  3. DegustatoR

    Veteran

    Joined:
    Mar 12, 2002
    Messages:
    1,459
    Likes Received:
    202
    Location:
    msk.ru/spb.ru
    That doesn't mean that NV won't get a lot out of it right for the first time ;)

    I still think that there won't be any "killings" until DX10 hardware -- DX9 is too thouroughly known at the moment for any vendor to make some stupid architectural mistake (like NV3x). So there won't be much difference in quality and speed between G7-whatever and R5-whatever-0.

    I'm waiting for any G80 information. That's where all the bells and whistles are.
     
  4. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    10,536
    Likes Received:
    583
    Location:
    New York
    While you're bringing the single-slot cooled 7800GT into it, why not bring up the even louder X1600XT? I was just reporting what reviewers have said, YMMV.
     
  5. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    10,536
    Likes Received:
    583
    Location:
    New York
    Is there any game where the XT pulls away from the GTX256 without AA? I did a quick check of the bigger sites but couldnt find any. It seems people are letting AA numbers color their impression of the XT's shading prowess vs the GTX.

    How, so? I think the odds are in ATi's favor there if you're talking 32x650 vs 48x650. Right now, without AA the 256GTX and the XT are neck-and-neck in line with their theoretical parity. All irrelevant anyway as it's going to come down to AA and bandwidth because that's where most of the bottleneck is.
     
    #365 trinibwoy, Dec 24, 2005
    Last edited by a moderator: Dec 24, 2005
  6. Pete

    Pete Moderate Nuisance
    Moderator Legend Veteran

    Joined:
    Feb 7, 2002
    Messages:
    5,032
    Likes Received:
    465
    DriverHeaven's article should settle that, then. Clock for clock, NV's fragment shaders were more capable, IIRC--at least, in the tests DH used. On that note, it may be that the games tested are coloring our impressions of either card's shader prowess.

    Well, that doesn't take into account the different capabilities of each shader unit, as well as ATI's separated texture units (which, though they be only 16, may prove sufficient if newer titles are as shader-heavy as everyone says). Actually, 48:16 is exactly the 3:1 ratio both NV and ATI have publicly anticipated.

    Edit: Duh-urr, I just repeated the pts that two other ppl made on this very page. I go now.
     
    #366 Pete, Dec 24, 2005
    Last edited by a moderator: Dec 24, 2005
  7. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,928
    Likes Received:
    230
    Location:
    Seattle, WA
    Ah, but the R580 only increases one aspect of its architecture's performance, so you just won't see all of those new math units used all the time.
     
  8. ToxicTaZ

    Newcomer

    Joined:
    Mar 6, 2003
    Messages:
    52
    Likes Received:
    2
    Location:
    Edmonton
    90nm GPU from Nvidia

    I gess Nvidia will show off all there new 90nm GPUs here?

    http://www.nvidia.com/object/IO_26082.html

    Catch NVIDIA at the Consumer Electronics Show 2006!

    CES 2006: January 5-8 in Las Vegas, Nevada

    Stop by the NVIDIA booth (#36200) located on the upper level of the Las Vegas Convention Center, South Hall. This year NVIDIA will showcase the latest technologies which enable users to experience an unparalleled High Definition experience – in PC games, video, and DVD playback.


    More info on the G71

    G71/RSX, G72, G73

    and there new driver too!

    Forceware 90xx
     
    Geo likes this.
  9. Razor1

    Veteran

    Joined:
    Jul 24, 2004
    Messages:
    4,232
    Likes Received:
    749
    Location:
    NY, NY

    Well isn't it more like 32x2x650 vs 16x3x650? But there is a possibility the g71 only has 24 pipes so 24x2x650, which will be the same as ATi's 16x3x650, so I'm guessing its going to go in favor of ATi, but not by much, if nV stays at 24 pipes and goes infavor to nV if they move up to 32 pipes.
     
    #369 Razor1, Dec 26, 2005
    Last edited by a moderator: Dec 26, 2005
  10. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,081
    Likes Received:
    651
    Location:
    O Canada!
    The "3" would refer to parallel pixel pipelines, each of which are still both "2" (although the second is only ADD/Modifer capable).
     
  11. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,081
    Likes Received:
    651
    Location:
    O Canada!
    Although, anytime the current architecture is bottlenecked by the shaders means that other elements will be underutilised, so those dynamics will likely change as well.
     
  12. Shadowmage

    Newcomer

    Joined:
    Sep 30, 2005
    Messages:
    60
    Likes Received:
    3
    I'll be sure to find out when I'm there, thanks for the heads up :)

    ----

    I heard rumors of G71 being 700mhz+ core speeds. Any truth in those, or complete bull?
     
  13. _xxx_

    Banned

    Joined:
    Aug 3, 2004
    Messages:
    5,008
    Likes Received:
    86
    Location:
    Stuttgart, Germany
    Corrected ;)
     
  14. Bad_Boy

    Bad_Boy god of war.
    Veteran

    Joined:
    Apr 15, 2004
    Messages:
    3,355
    Likes Received:
    25
    So the RSX is based off the G71 and not the G70? :) nice
     
  15. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    10,536
    Likes Received:
    583
    Location:
    New York
    Well like Dave said it's more like 32x2x650 vs 16x3x1.5x650. I don't think you can equate two ALU's per shader with two individual shaders with 1.5 ALU's each. I think you'll get better utilization out of the latter, especially with one ALU being borrowed for texture lookup duties on G7x.
     
  16. Razor1

    Veteran

    Joined:
    Jul 24, 2004
    Messages:
    4,232
    Likes Received:
    749
    Location:
    NY, NY

    Oh yeah I see that may make the r580 quite a bit more efficient over the the g70 tech possibly.
     
  17. walterman

    Newcomer

    Joined:
    Dec 27, 2005
    Messages:
    8
    Likes Received:
    0
    I still don't understand this R580 vs G71 PS/ALU config.

    Pixel Shader Unit -> A group of ALUs (full or mini or shared or scalar or vectorial)

    So, talking about the number of ALUs: How many ALUs could have each chip ?

    http://techreport.com/reviews/2005q2/geforce-7800gtx/pixel-block-small.gif
    So, G71 has 2 full ALUs (1 shared) & 2 mini ALUs per PS unit ???

    http://www.beyond3d.com/reviews/ati/r520/images/ps.gif
    & R580 has 2 full ALUs & 2 mini ALUs per PS unit. ???

    Please, somebody, fix me, cause i'm messed with these PS setups.

    So, lets say that 2 Full ALUs per PS unit in G71 & R580.

    G71 -> 2 ALUs per PS unit * 1 PS Unit per pipe * 32 pipes = 2 * 1 * 32 = 64 ALUs
    R580 ->2 ALUs per PS unit * 3 PS Units per pipe * 16 pipes = 2 * 3 * 16 = 96 ALUs

    So, if both chips will work at 700MHz +-, R580 should have a 50% advantage. But, also, how did ATI manage to put 3 times more ALUs than in R520 ?

    What's wrong ?

    Please, help !
     
  18. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
    since you can't compare R580 to R520, remember? R580 = RV530 on steroids .. now the only thing we need to find is a spot where the 1600XT shines..

    Or you can take a XL or a 256mb XT, downclock it to the 1600xt speeds and disable the 4 pipelines.. then you'd have a good index on r520/r580 performance..
     
  19. no-X

    Veteran

    Joined:
    May 28, 2005
    Messages:
    2,333
    Likes Received:
    290
    walterman: I think G70 has 2 PSPs (not parallel) per pipeline, each PSP consists of one full-ALU and one mini-ALU. One full-ALU is shared.

    R520 (R3xx, R4xx) has one PSP per pipeline, each PSP consists of one full-ALU (vec3 + scalar MUL or ADD or MAD) and one "uber" mini-ALU (vec3 + scalar ADD). And a separated texture-address ALU.

    R520 has 4 quads of these pipelines; R580 will have 3-times more of them.

    But I can be completely wrong, of course :)
     
  20. ninelven

    Veteran

    Joined:
    Dec 27, 2002
    Messages:
    1,710
    Likes Received:
    131
    Walterman, you should have:


    G71 -> 2 * 1 * 32 = 64 MADDS
    R580 ->1 * 3 * 16 = 48 MADDS

    Of course, 64 - 16tex = 48.

    EDIT: "ALUs" is rather vague... so replaced with madds... of course different operations will paint different pictures.
     
    #380 ninelven, Dec 27, 2005
    Last edited by a moderator: Dec 27, 2005
Loading...
Thread Status:
Not open for further replies.

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...