Half Life 2 current generation GPU benchmarks @ 3Digest

Discussion in '3D Hardware, Software & Output Devices' started by Sho, Jul 22, 2004.

  1. Sho

    Sho
    Newcomer

    Joined:
    May 28, 2003
    Messages:
    32
    Likes Received:
    0
    Well, take a look:

    http://www.ixbt-labs.com/articles2/digest3d/0604/itogi-video-hl2-wxp.html
    http://www.ixbt-labs.com/articles2/digest3d/0604/itogi-video-hl2-wxp-aaa.html

    Now, how conclusive are those results? Back in the 59xx days, HL2 used a special nVidia path with lowered precision and other optimizations at the cost of image quality due to nVidia's less-than-stellar shader performance. If above benchmark still used the same path for the 6800 class cards, the comparision wouldn't be very fair to ATi. On the other hand, I'd fully expect nVidia's latest offerings to run the high-precision/quality path with reasonable performance. Not sure what to make of this.

    Comments, speculation, facts?
     
  2. jvd

    jvd
    Banned

    Joined:
    Feb 13, 2002
    Messages:
    12,724
    Likes Received:
    9
    Location:
    new jersey
    This is the same old beta . It most likely regards the 6800ultra as a nv3x part and may be running the mix mode or dx 8.1 path. I wouldn't put any stock into these numbers .

    Most likely another month or two and we will learn more about half life 2
     
  3. ZoinKs!

    Regular

    Joined:
    Nov 23, 2002
    Messages:
    782
    Likes Received:
    13
    Location:
    Waiting for Oblivion
    Current rumours put hl 2 gone gold date at the first week of August. That'll leave just enough time to play through d3 once and play a few rounds of deathmatch.

    It's a good time to be a gamer. :D
     
  4. solobird

    Newcomer

    Joined:
    Apr 18, 2004
    Messages:
    32
    Likes Received:
    0
    FYI... Since the architecture difference between 6800 series and 5950 series is significantly huge, the shortage in shader performance lives nowhere in 6800. Therefore, if the mixed precision can do good to 5900 series, then this trick can cause no benifit to 6800 any more, due to the full speed in full precision processing in 6800 PipeLine. As a reference of shader performace the Shadermark reveals that the 6800 series runs almostly the same speed on FP16 and FP32, and sometime the FP32 is even faster than FP16! OMG! :D
     
  5. Ruined

    Regular

    Joined:
    Sep 20, 2003
    Messages:
    317
    Likes Received:
    4
    Nothing wrong with Nvidia using FP16pp in HL2, in fact I hope they use it. Crytek has shown that FP16 can be used without quality loss in the FarCry 1.2 patch. We can also see from FarCry that the 6800 didn't gain any performance from running the NV3x path - in fact, sm2.0 in the 1.2 patch is faster than the nv3x path in 1.1.

    As long as Valve doesn't convert shaders to 1.1 like Crytek did for the NV3x path, there should be no quality loss with FP16pp assuming Valve uses it properly.

    I expect R520/NV50 to both use FP16/FP32 as well. Don't think either company would give up free performance with no IQ loss.
     
  6. jb

    jb
    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    1,636
    Likes Received:
    7
    I agree nothing wrong. However next time you use the ATI is only doing 24 bit while NV is doing 32 here or at nVnews, then I will proceed to flogg you using your reply. If its ok for one to get by on a lower precesion, then you can not say its not in another case :)
     
  7. hovz

    Regular

    Joined:
    May 10, 2004
    Messages:
    920
    Likes Received:
    0
    ya but ruined doesnt think with logic, he thinks with fanboyism
     
  8. StarGazer

    Newcomer

    Joined:
    Apr 13, 2004
    Messages:
    48
    Likes Received:
    0
    I'd really like to see nVidia cards playing games at FP32... oh wait... even 6800 cards are too slow to render everything at FP32.
     
  9. jb

    jb
    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    1,636
    Likes Received:
    7
    Hey now play nice! I made that statement to be more of a wise crack that had a very small point. I should have used more smilies to indicate it was more of a jest. However I do not like it when I see people being hypocritical. Its find to make a statement then over time change your view (ie they found out some more info that changed their minds, they did not think it through the first time, ect) for what ever reasons you have. But when you flip around weekly just to support something other than a logical stance. And I am not saying Runied did that...just a joking warning that if he does I will have to call him on it :)
     
  10. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    10,430
    Likes Received:
    433
    Location:
    New York
    Huh? Based on what? From what I've read FP32 is much better on NV40 compared to NV3X but FP16 is still used wherever there is no IQ degradation because FP16 is still faster than FP32.
     
  11. hovz

    Regular

    Joined:
    May 10, 2004
    Messages:
    920
    Likes Received:
    0
    u should visit nvnews forums then, cuz trust me he does.
     
  12. Star_Hunter

    Newcomer

    Joined:
    Jan 21, 2003
    Messages:
    77
    Likes Received:
    0
    Location:
    Michigan
  13. StarGazer

    Newcomer

    Joined:
    Apr 13, 2004
    Messages:
    48
    Likes Received:
    0
    Of course it's better, but I don't think nVidia would like to trade FP32 for FP16 as, again, I believe the card can't render the same scene at FP32 with enough speed.
     
  14. Ostsol

    Veteran

    Joined:
    Nov 19, 2002
    Messages:
    1,765
    Likes Received:
    0
    Location:
    Edmonton, Alberta, Canada
    FP16 vs FP32 only becomes an issue when a large number of registers are being used. This is especially so on the NV40, where the register limitation has been significantly reduced from the horror that was the NV30. Using a handful of registers, either precision results in the same performance.
     
  15. Sho

    Sho
    Newcomer

    Joined:
    May 28, 2003
    Messages:
    32
    Likes Received:
    0
    Meaning above benchmark comparision would be reasonable if it used the NV3x branch on the NV40-based cards?
     
  16. Graphics_Krazy

    Newcomer

    Joined:
    Oct 15, 2003
    Messages:
    39
    Likes Received:
    1
  17. Sandwich

    Regular

    Joined:
    Mar 25, 2004
    Messages:
    288
    Likes Received:
    1
  18. Ostsol

    Veteran

    Joined:
    Nov 19, 2002
    Messages:
    1,765
    Likes Received:
    0
    Location:
    Edmonton, Alberta, Canada
    It -is- an issue, but not as much. Shader writers have much more leeway before they really have to worry.
     
  19. Bouncing Zabaglione Bros.

    Legend

    Joined:
    Jun 24, 2003
    Messages:
    6,363
    Likes Received:
    82
    Nvidia makes a big deal of FP32 and rubbishes FP24. They finally produce a card that can do FP32, and then make sure that all the customer sees is FP16 - worse than what they criticise ATI for. For all intents and purposes, NV40 becomes a FP16 card if that's all you can get out of it, no matter how capable the hardware is while running FP32.

    The consumer suffers because of marketing wanting to get a few more frames on the benchmarks.
     
  20. Razor1

    Veteran

    Joined:
    Jul 24, 2004
    Messages:
    4,232
    Likes Received:
    749
    Location:
    NY, NY
    Iq wise fp 16 looks just as good as fp 24 now, so why is everyone complaining? If there was an Iq difference yes thats something to be noted. fp 16 can do somethings that fp 32 can't yet. Like blending. Fp 24 can't do blending either if I got my information correct. Speed wise as someone stated early is not much of a difference until number of registries are maxed out.
     
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...