gtx 580 3 way sli multi monitor benchmarks?

[SEI]

all the reviews are usually just for one screen. and they all run into cpu limitations with the third card. i only found one review with 3 way sli multi monitor and results seem promising.. (im asking because i just switched from a 5850 to 2x gtx 580 and was wondering if i should add a third one because i still cant run some games at max with just two)

[thales100]

Well i dont have three GTX 580, but 3 x GTX 480, unfortunately the third GPU seems to be pretty useless, plus when i run tri-sli and surround i get occasionally some BSoD and CTD, but no issues at all when running 2 x GTX 480.

[SEI]

thanks for that. gtx 480 surround benchmarks would be good aswell basically the same. just wondering since it seems 3x is only usefull for multi monitor. so, in your experience you think a third card wont make a significant difference for us surround gamers?

[thales100]

in your experience you think a third card wont make a significant difference for us surround gamers?

Yes thats it, maybe in the future, who knows, but not now.

[SEI]

thanks :) saved me a 500 dollar experiment

[scavvenjahh]

Three 580s here, and GPU usage is even on all cards in SLI games, so I do hope it makes a difference :) No BSOD/CTD so far.
"Remove one card and check multi-mon performance" > added to my to-do list, I will post results here.

[thales100]

Three 580s here, and GPU usage is even on all cards in SLI games, so I do hope it makes a difference :)

Dont think so, its same here with GPU utilization, but ive tested 3 x GTX 480 and 2 x GTX 480, the third card is pretty useless, and in some (and not few) games you get better FPS with two crads than three. :lol:

[SEI]

ooh.. getting interesting.. so i still might get that third card :P i think i need to change motherboard i wanna add a pcie ssd and my p6x wont have enough slots now

[scavvenjahh]

Don't get too excited :oops:
I've been running benchmarks on 1/2/3-card configurations, and clearly, there are SLI issues on my system. I'll have data to post later on; suffice to say that only 2 out of 8 games I've tested fare better in 3-way than in 2-way SLI. :shock:
I'm more and more convinced that my mainboard's 16x/8x/8x PCI-E config is the culprit, but it's mostly because I can't think of anything else...

It's not all a waste though... 3 cards can be used in other ways than triple-SLI: Mafia 2 loves the "two in SLI, the third for PhysX" arrangement :mrgreen:

More on all that later.

[thales100]

Don't get too excited :oops:
I'm more and more convinced that my mainboard's 16x/8x/8x PCI-E config is the culprit, but it's mostly because I can't think of anything else...

No way, its a driver issue, dont blame your mobo for nothing. Btw im runninG three 16x PCI-e. :wink:

[scavvenjahh]

Ah, I see. And I hear 4-way SLI isn't supported at all with 580s and probably never will... although it supposedly works with 480s. What a mess :lol:

Pretty weird though, I do get excellent scaling in Dirt 2 :? (and Just Cause 2 to some extent, but there's sth fishy with my 2-way SLI results in this one)

[SEI]

maybe just cause 2 in 2 way sli has problems? because i tried that with my 580 and on max its unplayable.. but everything else (mafia 2, dirt 2, wow, batman AA works fine at max)

[Paradigm Shifter]

Unless it's a Just Cause 2/GTX580 specific issue, JC2 has no problems at all with 2-way SLI.

[jethc9]

Hi everyone. I am new to the forums here, but I can tell you that 3 GTX 580's in 3 way SLI works great. I have had zero problems and I get near perfect scaling across all three cards. I am using 3 HP LP3065's in Surround, bezel corrected to 8100x1600. If anyone wants me to run a particular benchmark and post the results I would be happy to do it.

BTW.. I have had no problems with JC2 in 3 way SLI.

[scavvenjahh]

Here are my results... Take it with a grain of salt, this is really amateur work - I tested only 7 games and ran ONE pass of each benchmark at max settings. I just wanted a rough but documented idea of GPU scaling in some of my currently installed and easy-to-bench games, so here goes.

My system:

i975 @ 4.08 GHz
6 GB DDR3 1700 MHz CAS9
EVGA X58 SLI
GTX 580 driver v263.09

Multi-mon display: 3x 1680x1050. Remember, that's 40% less pixels than 3x 1920, so your mileage - GPU load and SLI scaling - may vary.
PCI-E link speed: I found no difference between SLI 16x/16x and 16x/8x, so let's say it has no impact and leave it at that. 3-way SLI on my system works at 16x/8x/8x. PCI-E bandwidth couldn't explain the varied results below, anyway.

R.U.S.E.
Built-in benchmark.
Insane preset, 'Disable SLI-unfriendly settings' On (removes motion blur, apparently).

1) GPU Rating @ 2560x1600

Single GPU: 47.4
2-way SLI: 87.3
3-way SLI: 88.3

Typical peak VRAM usage: 1130 MB.

2) GPU rating @ 5220x1050

2-way SLI: 67.7
3-way SLI: 63.9

Typical peak VRAM usage: 1450 MB.

3) Quick conclusion

The additional GPU power of 3-way SLI is unused. Higher multi-mon resolutions may yield better scaling, but wouldn't VRAM usage become a problem then ?

DiRT 2
Built-in benchmark
DX11, Ultra High preset, 4x AA.

1) Avg framerate @ 2560x1600

Single GPU: 71.6
2-way SLI: 127.5
3-way SLI: 146.6

2) Avg framerate @ 5220x1050

2-way SLI: 95.7
3-way SLI: 123.2

Typical peak VRAM usage: 1400 MB.

3) Quick conclusion

Satisfactory scaling. Each additional GPU boosts both min & max framerates.

Mafia 2
Built-in benchmark
Max settings, APEX PhysX Medium.

1) Avg framerate @ 2560x1600

Single GPU + PPU: 37.7
2-way SLI: 45.4
2-way SLI + PPU: 64.0
3-way SLI: 50.4

Typical peak VRAM usage: 650 MB (500 when PPU is present).

2) Avg framerate @ 5220x1050

2-way SLI: 39.5
3-way SLI: 51.5

Typical peak VRAM usage: 820 MB.

3) Quick conclusion

Satisfactory scaling in multi-mon, but a dedicated PhysX card is even more useful performance-wise. This could be because of my "low" resolution and/or the level of PhysX content in the benchmark.

Apache Air Assault
Beginning of the same Free Flight mission
High preset

1) Avg framerate @ 2560x1600

Single GPU: 68
2-way SLI: 65
3-way SLI: 44

Typical peak VRAM usage: 450 MB.

2) Avg framerate @ 5220x1050

2-way SLI: 35
3-way SLI: 23

Typical peak VRAM usage: 600 MB.

3) Quick conclusion

Not only does this game have no SLI support yet, but multi-GPU actually degrades performance dramatically. Hopefully a driver update will fix this.

GTA 4
Built-in benchmark
Max settings

1) Avg framerate @ 2560x1600

Single GPU: 62
2-way SLI: 65
3-way SLI: 64

Typical peak VRAM usage: 1000 MB.

2) Avg framerate @ 5220x1050

2-way SLI: 64
3-way SLI: 61

Typical peak VRAM usage: 1200 MB.

3) Quick conclusion

Not a SLI-friendly game. GTA4 loves beefy CPUs and large amounts of RAM/VRAM above all, so the additional CPU power required to handle 3 cards has an obvious (albeit negligible) impact here.

Lost Planet 2
Benchmark Test B
DX11, max settings

1) Avg framerate @ 2560x1600

Single GPU: 35.9
2-way SLI: 59.8
3-way SLI: 63.3

Typical peak VRAM usage: 1050 MB.

2) Avg framerate @ 5040x1050 (bezel comp. not supported)

2-way SLI: 52.6
3-way SLI: 55.7

Typical peak VRAM usage: 1200 MB.

3) Quick conclusion

A third GPU brings remarkably little additional performance. SLI scaling is good though, so a more aggressive CPU overclock or higher resolutions should help.

Metro 2033
Benchmark tool (metro2033benchmark.exe)
DX11 High, AAA + Tesselation + PhysX.

1) Avg framerate @ 2560x1600

Single GPU + PPU: 39.72
2-way SLI: 65.29
2-way SLI + PPU: 66.24
3-way SLI: 68.67

Typical peak VRAM usage: 830 MB (750 when PPU is present)

2) Avg framerate @ 5220x1050

2-way SLI: 55.04
3-way SLI: 59.88

Typical peak VRAM usage: 990 MB

3) Quick conclusion

Again, SLI scaling is good up to two GPUs. A dedicated PhysX card has little impact at these settings.

Just Cause 2
Concrete Jungle benchmark
DX10, max settings incl. CUDA water

1) Avg framerate @ 2560x1600

Single GPU: 27.33
2-way SLI: 41.47
3-way SLI: 43.75

Typical peak VRAM usage: 750 MB

2) Avg framerate @ 5220x1050

2-way SLI: 32.87
3-way SLI: 32.97

Typical peak VRAM usage: 950 MB

3) Quick conclusion

Negligible gain of 3-way SLI. Again, is the CPU holding me back, would higher resolutions help ? Or are some of the advanced settings "3-way SLI-unfriendly" (shadows) ?

Summary of benchmark results vs. 3-way SLI
:!: Remember that this is for MY resolutions, with MY settings, on MY system, after only one run of each benchmark, in only seven games. Actual gameplay has often little to do with benchmark scenes, too so this is NOT a general SLI scaling evaluation of games, but a summary of GTX 580 3-way SLI results.

Good scaling:
Dirt 2
Mafia 2

Zero gain: (CPU-limited results @5M pixels)
R.U.S.E.
GTA 4
Lost Planet 2
Metro 2033

Negative gain: (SLI issues)
Apache Air Assault

[SEI]

thanks for that! i dont think its worth it for the extra fps gains.. think we would benifit from more vram then anything.. some games seem to be using near max

[0utf0xZer0]

Hi everyone. I am new to the forums here, but I can tell you that 3 GTX 580's in 3 way SLI works great. I have had zero problems and I get near perfect scaling across all three cards. I am using 3 HP LP3065's in Surround, bezel corrected to 8100x1600. If anyone wants me to run a particular benchmark and post the results I would be happy to do it.

BTW.. I have had no problems with JC2 in 3 way SLI.

I'd pretty much need to win the lottery to be able to afford the setup you have... but I'm actually really curious as to what games are playable at that res. I'm not sure I have any particular benchmark requests though.

[Uldred101]

I plan to have a new computer built in a couple months or so (end of Feb, early Mar). I'm going with three monitors and currently planning to ultimately end up with 3 SLI. I say ultimately as I will only purchase 2 580's (or better) at the start and later on will get the third. I will however, purchase everything else beforehand to be ready for the 3rd SLI (motherboard/psu, etc), assuming there are no complications.

As I continue to see various posts and reviews, I'm seeing a mixed bag of results making me steer away from going with 3 SLI.

Any other thoughts from those that are currently using 3 SLI and 3 monitors? For now, I'm just compiling information, but if I stay away from 3 SLI, then there's no sense in getting a motherboard (etc.) to support it.

Time will tell.

Uldred

[Naieve]

Just found this review after looking around for awhile. Tri-SLI 480's vs 480 and 460 SLI multi monitor. Seems to show decent scaling on the games chosen.

http://www.tomshardware.com/reviews/quad-sli-nvidia-surround-geforce-gtx-480,2745-10.html

[csaris]

Just found this review after looking around for awhile. Tri-SLI 480's vs 480 and 460 SLI multi monitor. Seems to show decent scaling on the games chosen.

http://www.tomshardware.com/reviews/quad-sli-nvidia-surround-geforce-gtx-480,2745-10.html

That's because they test in canned benchmarks which have nothing to do with real gaming.

I've tried 3x580s in surround and the ingame gain was zero.