Quad SLI vs Quad Crossfire Heavyweight Showdown GTX 590 & HD 6990 Linus Tech Tips

Linus Tech Tips ·Linus Tech Tips ·2011-05-08 · 1,064 words · ~5 min read
Floatplane YouTube

Transcript

JSON SRT VTT 80
0:00 so the purpose of today's video is going to be to evaluate quad crossfire with
0:05 dual Radeon 6990s which you can see on my bench right now
0:09 against quad sli with dual dual jewel dual gtx
0:15 590 so those are
0:18 about to go on the bench so i'm going to be using a different test platform from
0:22 what i normally use this is a Gigabyte g1 sniper motherboard
0:28 and it is running with an Intel core i7 980 x at 4.26 gigahertz or something
0:35 like that so i'm well over 4 gigahertz i have 12 gigs of ddr3 1600 RAM i am using
0:43 a different power supply as well but i've got my regular testing hard drive
0:47 and then because my external drive broke i'm using a wd green drive for my steam
0:52 files so that is the test bench i will be using to determine which of these
0:56 configurations runs better you may have noticed already that it's a little on
1:01 the warm side oh no not warm side it's a little on the loud side the reason for
1:04 that is that you can see these rubber spacers that come on the 6990s
1:10 here yeah now you can see it that is to give it see you can see at
1:14 the bottom it's very close together at the top it's
1:17 slightly separated that gives it a little bit of a gap for that fan to
1:21 intake air but it still has to work pretty hard to cool those two gpus
1:26 so it is a little on the loud side here you
1:30 can see my lost planet two results i'm going to be comparing
1:33 the quad crossfire against the quad sli and then if i have time i will also run
1:38 individual 6990 and individual gtx 590
1:41 numbers to compare against their dual
1:45 graphics equivalents although it's not equivalent
1:48 at all so i've got most of my testing done i'm
1:53 just running the lost planet 2 bench on the gtx 590 quad sli configuration so of
1:59 note about this particular video is the fact that i will be using a higher
2:04 resolution than normal you can see my 1920x1080 monitor is actually back there
2:09 behind my 30 inch 2560x1600 monitor also because we are
2:14 using very high-end GPU configurations i will be using higher anti-aliasing in
2:18 isotropic filtering settings than i normally do so i'll be cranking up
2:22 as high as they go for all of the games in my testing suite which will be
2:26 battlefield bad company 2 crisis 2 metro 2033 and lost planet 2. so for the first
2:31 three games i'll be doing my own custom run-throughs recorded with fraps and for
2:35 lost planet 2 i'm using test b you can see here i have the gtx 590s on
2:40 the test bench they're not nearly as loud as the 6990s
2:44 although they do get toasty and they do get loud running in a configuration like
2:47 this especially that top card although you can see the spacing is significantly
2:51 more than we were seeing previously with
2:55 our other config you can actually see all the way through it and both cards
2:59 are able to sit straight in their slots
3:02 you may ask why i chose to use the slots
3:05 that are right next to each other on this board versus this one right here
3:09 and the reason as you can see in the PCIe slot there is that
3:13 this bottom slot is only wired for 8x so if i were to use these two high-end gpus
3:18 uh in the top and the bottom PCIe 16x slots on the g1 sniper i'd be running
3:23 them at 16x and 8x mode respectively and
3:26 based on my recent testing with PCIe bandwidth i felt that was not really
3:30 appropriate so i would have liked to use the two 16x slots which is what i've
3:34 decided to do here so i will be back with the results once i'm finished
3:37 running my single gtx 590 tests all right guys so i'm using the latest
3:42 drivers available at the time of filming and benchmarking which is all on the
3:47 same night tonight and uh here are the settings that i was
3:50 using so crisis 2 is obviously at maximum so the game that scaled the best
3:56 out of the ones in my test suite was obviously battlefield bad company too uh
4:00 the 6990s um actually scale more than a hundred
4:05 percent which makes no sense but uh we'll just consider that within the
4:09 margin of error and we'll call it 100 scaling and then for the 590s we saw
4:14 almost 100 scaling something like 90 scaling so excellent scaling on that
4:19 particular game 590 scaled very very well in crisis 2. 69.90 besides the
4:24 flickering bug which made the game pretty much unplayable
4:28 also didn't scale very well AMD has some work to do on the crisis 2 performance
4:33 of their 6990 but i'm sure that's something they'll resolve with a driver
4:36 issue in the future metro 2033
4:40 6990 scaled quite well uh adding another
4:43 50 performance to itself as well as the 590 590 actually scaled quite well as
4:48 well but still falls short of the quad crossfire setup in terms of overall
4:52 performance on lost planet 2 we see very
4:56 poor scaling across the board and the 590 sli beats a little crossfire
5:01 as well as on battlefield bad company 2
5:05 the 590 quad sli wins so basically the
5:08 objective today was not to determine a clear winner it was more to look at the
5:12 state of quad GPU gaming as a whole i'm not going to say
5:16 who but i well actually okay i got blue screens with both solutions during my
5:21 testing among other sort of random issues that
5:24 did crop up so i would go as far as to say that there's definitely some work to
5:28 do on the drivers for uh both of these cards and uh but once that's resolved
5:32 they definitely deliver some crazy performance numbers in the
5:37 right scenarios remember these are all benchmarked at 2560 by 1600 so we're
5:42 looking at battlefield bad company 2 numbers well over 100 FPS