Slashdot Mirror


SLI Performance Reviewed at Anandtech

DarkSarin writes "Anandtech has a review of dual gpu's using nvidia's nforce4 SLI chipset. They are showing ~66 fps on doom3 at 1600x1200 with 4xAA turned on (a 75% increase in performance over a single gpu). It's not available yet, but it should be fairly soon."

45 comments

  1. Nice! But... by glowimperial · · Score: 2, Interesting

    I wonder how many of us are actually going to shell out for new graphics hardware. We are kind of hitting the wall in performace here already.

  2. Expensive..... by spicy+salsa · · Score: 1, Insightful
    I wonder how many people will actually use this. Two 6800 GT's at 400 dollars each equals 800 dollars. Plus your going to need a massive power supply. One 6800 GT requires a 350 watt power supply so would two require a 700 watt power supply?! Yeeeesh....

    Free Flat Screen HERE!

    1. Re:Expensive..... by justforaday · · Score: 2, Interesting

      Let's not forget another 100-200 bucks or so for the mobo to accomodate the two gfx cards...

      --
      I'll turn into a supernova and burn up everything. Well I'll turn into a black little hole and you'll turn into string.
    2. Re:Expensive..... by chrismcdirty · · Score: 3, Informative

      It requires a 350 watt when used with a standard system. So it shouldn't require 700, but still probably considerably more than 350.

      --
      It's like sex, except I'm having it!
    3. Re:Expensive..... by Shinglor · · Score: 0, Offtopic

      Mods should moderate any posts with lame referrer links as "Offtopic".

  3. Where are the Quake III Arena results? by Anonymous Coward · · Score: 2, Funny


    I want to see 700fps in Q3A. Not sure why, but I do.

  4. Math doesn't add up by chrismcdirty · · Score: 2, Informative

    So I'm expected to shell out 2x$$$ for a video card, extra $$$ for the mobo with 2xPCI-E, extra $$$ for the huge powersupply, but only get a 75% performace gain?

    --
    It's like sex, except I'm having it!
    1. Re:Math doesn't add up by Anonymous Coward · · Score: 1, Funny

      Yes...And you'll do as you're told. Now go clean your room!

    2. Re:Math doesn't add up by DumbRedGuy · · Score: 1

      I agree with you, but having the cutting edge equipment never scales in "bang-for-the-buck."

    3. Re:Math doesn't add up by Slime-dogg · · Score: 1

      This is a bit more insightful than just funny.

      No one is telling the parent poster to go out and buy these cards, and buy this hardware. Yes, it is being marketed (since "reviews" are really just a cheaper way of marketing for these people), but no one is making up your mind for you to plunk down two wads of cash for two graphics cards.

      This is the typical outburst that I hear from the lemmings that line up for plastic, and buy what the television tells them to buy. :ugh:

      --
      You need to restart your computer. Hold down the Power button for several seconds or press the Restart button.
    4. Re:Math doesn't add up by jensen404 · · Score: 1

      2.5 times the price for 1.75 times the speed is a quite good ratio for a very high end upgrade.

      An athlon 4000+ costs almost 3 times as much as a 3400+ with a less than 20% increase in speed. (and a dual cpu upgrade won't help in any game)

    5. Re:Math doesn't add up by UWC · · Score: 2, Informative

      The point made at the end of the article is that even if you can't afford two now, you can get one, and in 6 or 12 months not only will the current top of the line be cheaper, but you can get a huge performance increase by just buying an extra one (and maybe a bigger power supply) instead of replacing your current card entirely.

    6. Re:Math doesn't add up by NanoGator · · Score: 1

      a.) It's not NVidia's fault you'd have to buy a new mobo.

      b.) if you really need the performance that bad, you'll be hard pressed to spend 2x on the card and find a >75% faster card that uses only one slot.

      c.) It's optional, not expected.

      --
      "Derp de derp."
    7. Re:Math doesn't add up by HoneyBunchesOfGoats · · Score: 1

      A lot of the tests are limited by the CPU, which becomes very apparent with such a ridiculous amount of graphics power. The amount gained depended upon the game (notice that in Far Cry, for example, there was a 105% gain for the 6800GT at 1600x1200, 4xAA).

    8. Re:Math doesn't add up by Bios_Hakr · · Score: 1

      We've been talking about this at my local LAN group. I just don't buy into it. In 6 months, the top of the line game will be optimized for UberPixelSlamin 3.75. Your card will only support 3.70. So, now you have a choice: spend more money to do SLI and still not have the features you want; scrap the card and buy one that does have features you want.

      Look at it this way. I just bought a FX5900 in January. I can't remember what I paid, but I know I saved up for a while. Now they have these 68oo thingies. They are only about a cunt-hair faster than mine, but they support more cool stuff like pixel shaders and bump mapping. What good will two crappy cards do me?

      So, the Uberhardcore will buy two right off the bat. In 10 months, they'll get 2 more of the next best thing. The semihardcore will always get the next best thing every 10 months. The hardcore will upgrade every 12 to 16 months; usually to one step below the next best thing. The casual gamers don't really give a shit.

      No one will care about SLI except as a dream. It didn't work in 1997 and it won't work today.

      --
      I'd rather you do it wrong, than for me to have to do it at all.
    9. Re:Math doesn't add up by Kris_J · · Score: 1

      Nah, that's not going to work. Pairing any two components that aren't bought at the same time never works well. SLI is about having better performance than you can get from one of the top of the line cards.

    10. Re:Math doesn't add up by Kris_J · · Score: 1

      Yes. SLI allows you to get better performance than is possible with the best single card. If you want the best 3D framerate you can get, the solution is SLI. It doesn't matter that the overhead means you don't double your performance, there's no other way to get performance this good.

    11. Re:Math doesn't add up by megan_of_wutai · · Score: 1
      Look at it this way. I just bought a FX5900 in January. I can't remember what I paid, but I know I saved up for a while. Now they have these 68oo thingies. They are only about a cunt-hair faster than mine, but they support more cool stuff like pixel shaders and bump mapping.
      That's a pretty big cunt hair you've got there.

      Also, Nvidia must have really taken a step backwards to have knocked bump mapping out of the FX series, considering that environment bump mapping has been in cards since the Matrox G400, and embossed bump mapping was basically there since x86 3D cards existed. The lack of pixel shaders in the GeForce FX series surprises me, but considering how awful the reviews were I'm not surprised.

    12. Re:Math doesn't add up by Bios_Hakr · · Score: 1

      Selective benchmarking? You picked the biggest difference you could find.

      NWN has less than a 10fps differnece.

      FFXI is only about 1000 whatevers slower.

      Wolf:ET is about 10fps slower.

      UT2k4 is also about 10fps slower.

      Also, all of these were run at 1280*1024. Running at the more common 1024*768 would close that 10fps gap quickly.

      Yes, my card does have bumpmapping. I was just being an asshole about features like PixelShaders and "full Dx9 compatible" which seem to be artifically gimped on older generation cards.

      --
      I'd rather you do it wrong, than for me to have to do it at all.
    13. Re:Math doesn't add up by megan_of_wutai · · Score: 1

      Halo is a shader/video card bound game, just like almost all future graphically impressive games will be, like doom 3 is, like HL2 will be. NWN, all the Q3 engined games, and UT2K4 are far less video card bound, relying much more on the CPU.

      They test at that res to try and get away from games being CPU bound and actually test the performance of the *card*, not the system it's in, testing the CPU in video card reviews offers less differentiated graphs, for some reason :-P.

      You basically want benchmarks that make your card look better, that's fine, I want the same, I have a 9700 pro @9800 pro speeds and I dislike watching my card get beaten into the ground by the 6800, but my next card will be a GF6 or above (incidentally PCI-e too :) ).

      It used to be that cards were mainly limited by the memory bandwidth they had, nowadays that's changed to how fast they can run fragment programs.

      The GFFX was a really bad design, considering how games have and are going, all of the sites have been saying this since the 5800 came out and benched worse than the 9700 pro that had been out for 6 months. Look at the GeforceFX post-mortem in the video section at anandtech.

    14. Re:Math doesn't add up by Anonymous Coward · · Score: 0

      Somehow I doubt that. It's more likely that it's limited by memory technology.

  5. Some of you guys are missing the point by Anonymous Coward · · Score: 4, Insightful

    Yeah, it's expensive, but cost has never really ever been the point. Having the fastest computer is.

    I totally undestand how some of you guys wouldn't spend that kind of cash on an SLI system, you just want a sensible computer for everyday use. Well, guess what? I don't. I want the highest framerates possible, so when I play my brother online in Halo or Doom 3 deathmatch, I can SPANK HIS ASS.

    I think that nvidia bringing back the SLI upgrade path is not only sensible, it's FUN. And that's the point.

    1. Re:Some of you guys are missing the point by spicy+salsa · · Score: 1
      Well, anything over 60 FPS is not going to help you anyway. The eye can't see much over 60 FPS. A dual card setup will allow you to run Doom III at very high detail levels though. If you want to run a game like Doom III with everything turned on then by all means buy two video cards. Graphics just don't mean that much to me, I don't mind turning down the detail levels a little bit.

      Free Flat Screen HERE!

    2. Re:Some of you guys are missing the point by Anonymous Coward · · Score: 0

      1st of all, our eyes dont see in "fps"

      using vsynch at 60mhz cause serious eyestrain for most people. personally, anything less than 100mhz gives me a headache.

    3. Re:Some of you guys are missing the point by justforaday · · Score: 1

      1st of all, our eyes dont see in "fps"

      Excuse me?!?!? I always use my eyes when playing a first person shooter... : p

      --
      I'll turn into a supernova and burn up everything. Well I'll turn into a black little hole and you'll turn into string.
    4. Re:Some of you guys are missing the point by scot4875 · · Score: 1

      That's because it's the same rate as ambient room lighting. Try setting a monitor to 60hz and looking at it with the lights on and off -- it looks worse with them on. It's especially pronounced with fluorescent lights. I can deal with 60Hz monitors as long as the room lighting is off.

      That said, I still prefer 75-85Hz on my monitors.

      --Jeremy

      --
      Jesus was a liberal
    5. Re:Some of you guys are missing the point by Anonymous Coward · · Score: 1, Insightful

      Well, anything over 60 FPS is not going to help you anyway.

      Uhh, do you actually play first person shooters competitively or on a regular basis at all? There's a tremendous difference in feel between a system doing 100+ frames per second and one that's just doing around 60.

      10 out of 10 gamers, people who do this all the time, who were going into a competition would pick the former machine over the latter, every time.

      Graphics just don't mean that much to me

      Then WTF are you posting here about? Are you just trolling then?

    6. Re:Some of you guys are missing the point by spicy+salsa · · Score: 1
      I am well aware of how some people who play FPS's competitively always try to get obscene FPS's. That does not mean they are right. If I ran a game at 100 FPS and asked someone "is this running at 60 or 100" they would not be able to tell me. Anything above 60 and it gets VERY difficult to see the difference. Monitor hertz is a little different. Most people can see the difference up to 72 hertz and some can see it up to 75. The difference in screen refresh is a little easier to detect though, then the FPS of on screen movement. Its easier to see "flicker" then the changes in FPS of on screen movement.

      Free Flat Screen HERE!

    7. Re:Some of you guys are missing the point by lsmeg · · Score: 1

      I think the general point is that any benchmark measuring frames per second is really giving you average frames per second. So that may mean a constant, smooth 60, or it could mean half the time it's 90, half the time it's 30, which would definitely be noticeable. The quest for frames per second is to get the average high enough so that your minimum meets or exceeds that sweet spot.

      --
      It's OK! I'm a limo driver!
    8. Re:Some of you guys are missing the point by Anonymous Coward · · Score: 0

      If that were true, benchmarks would report worst-case FPS. But they don't because it's really all about the size of your compu-penis.

      (Oh, and the fact that ridiclous FPS in Quake1/2/3 games allow you to jump farther.)

  6. It's called Lock-In by Alzheimers · · Score: 2, Interesting

    There's a very simple reason why this SLI is so desireable by the marketing types at NVidia -- Lock In. Sure, you may only have the cash to buy one card right now, but in a year when ATI releases their next new thing are you going to spend *another* $400 for the next new card, or are you going to buy the now "Budget" 6800 and get almost as good performance?

  7. Doom3 is capped at 60fps by fok · · Score: 3, Informative
    --
    \m/
  8. Wrong parent by Shinglor · · Score: 1

    I was referring to the link in the grand parent post, I accidently replied to the wrong comment

    1. Re:Wrong parent by realdpk · · Score: 1

      Should also "foe" them, and set foe as the lowest moderation in preferences. Then you won't see 'em again.

  9. Re:Nice! But... by aliens · · Score: 1

    Not until the hardware is pushing 100FPS with all the eye-candy on at the highest settings and at the highest resolution your monitor could possibly do.

    No walls, except those made out of cellophane.

    --
    -- taking over the world, we are.
  10. Re:Nice! But... by dbretton · · Score: 1

    We are in no way hitting the wall in performance.

    Perhaps you meant that we are hitting the wall in terms of the price/performance curve, with which I agree 100%. Most of us cannot justify the cost of these new cards, given the associated performance increase.

    Then again, this really isn't anything we all don't already know.
    Moderator Points: -1, Obvious

  11. nVidia, best drivers bar none by Lisandro · · Score: 2, Informative

    It surprised me how painless the SLI setup process was described. Writting GPU drivers is complex enough without having to handle multiple accelerators, never mind having optimal performance AND making the install painless.

    nVidia does their homework with their drivers, and it shows. GPU drivers are 50% of their perfomance. It sucks for ATI, because they have the better hardware, but their drivers suck (and pretty much always did).

    1. Re:nVidia, best drivers bar none by PoderOmega · · Score: 1

      Dude, you clearly have no idea what you are talking about.

    2. Re:nVidia, best drivers bar none by Lisandro · · Score: 1

      And that would be because....?

  12. Wonder how Oblivion will look on this... by realdpk · · Score: 1

    The screenshots of Oblivion are *incredible*, would love to see how it'd perform with two SLI cards. If you haven't yet, do yourself a favor and look up the Elder Scrolls: Oblivion screenshots. The company pulled some bullshit marketing thing giving Game Informer some exclusives, but some people have scanned them.

    IMO, it's like having an interactive Myst, it's so pretty.

    1. Re:Wonder how Oblivion will look on this... by Fryed · · Score: 1

      They gave an exclusive to Game Informer?! Wow. That's really rather ridiculous. I mean, at least with Game Informer, you can buy your game a damn good review, but no one I know reads the reviews in there because they're generally a complete pile of crap. At least the last time I read the magazine, a year ago, it was like this...and I can't imagine they've changed a whole lot. Some of the reviews it was pretty obvious had simply been paid for, and it was also obvious when you were reading a review that wasn't paid for, because they would make up stuff about how bad it was.

      Giving an exclusive to Game Informer sounds pretty fishy to me, then...almost like they have something to hide, and want to make sure they can guarantee a good review.

    2. Re:Wonder how Oblivion will look on this... by realdpk · · Score: 1

      It was a preview in this case, they're not releasing yet. In this case, Game Informer paid to host the exclusive screenshots/preview in their magazine.

      I agree, they aren't very trustworthy for reviews (almost none of the sites/magazines are).

      I'll be getting the game based on how much I got out of Morrowind, and now because it's just so beautiful.

  13. Re:Nice! But... by Deliveranc3 · · Score: 1

    I was thinking about that, what would you need for the perfect gaming machine. Entirely theoretical :P

    Basically to get past the point where software is at now (remember it will take a long LONG time for software to be able to model really complex systems, we need to scale the nuclear model). However there are graphics right now which they are attempting to render in real time, TOY STORY!

    All the CGI animations coming out of say PIXAR are lightyears ahead of what we see in games right now, however it is all computer modeled and ready to go as soon as we have the hardware to push it.

    I remember this one time I was making a quake 1 map.

    I decided to make a light sphere, just a sphere that was a light. (Some of you already know where this is going). Anyway I decided that 1000 sides was pretty good :P Then 200, then 100, then 50, well that one was compilable, it took 3 hours on my p3 550 with a TNT2 in it. Ah getting 1 frame a second those were the days.

    Conclusion: We still have a LONG WAY to go.

  14. maybe it will be available late next year? by MarcoAtWork · · Score: 2

    I mean, come on, NONE and I mean NONE of my local computer stores have in stock ANY 6800GT (you have to preorder, takes quite a while to get it) or ANY 6800Ultra (don't think they *ever* had one in stock period) AGP. The only cards that seem to be fairly available are vanilla 6800s and even then they have been available for a month tops.

    Given that the GT/Ultra AGP cards have been announced *several months* ago, before their PCIe versions will be available likely NVidia will have already announced the 8xxxx series or something.

    I fondly remember being able to buy video cards max. 2 weeks after they were announced, then it was a month, then two, and now it seems cards become available together with the paper launch of the new models :(

    I really feel for the stores, though, it must suck being able to stock only 'previous generation' parts.

    --
    -- the cake is a lie
  15. Suspect some performance selection, /.ers by lukestuts · · Score: 2, Interesting

    They are showing ~66 fps on doom3 at 1600x1200 with 4xAA turned on

    Were there any monsters on screen at the time? Was there any discarge of firearms? In my Doom 3 experience, you could go for a good 10 minutes with nothing worth shooting at - this would surely throw the fps up?

    Oh no! A teleporting cyberdemon! No wait, it's just doing the Doom 3 stutter-shuffle.