1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Two 760's SLI or one 780? Not like the other threads.

Discussion in 'NVIDIA' started by pabloottawa, Oct 3, 2013.

  1. Aithos

    Joined:
    Sep 30, 2013
    Messages:
    175 (0.17/day)
    Thanks Received:
    35
    Which benchmarks were you looking at? There were a couple where the SLI setup won, but if you look at the 1080p numbers the 780 won in quite a few games and the author was running stock speeds so the overclocking still means more headroom for the 780. The SLI setup I imagine would have won in the higher resolution numbers because SLI scales better at higher resolutions, but that's not relevant because you're running 1080p and not a 3 monitor "surround" setup. I'll take another look this afternoon, but were there specific numbers you had questions on?

    The author is also selecting the single card because stuttering and scaling is an issue with SLI setups regardless of what you're running. The performance will vary wildly in games each time a driver is released, sometimes the newest driver will improve your games and other times they will get worse as a change improves the newest titles but hurts older ones. Give me a bit and I'll edit this...

    Edit: Ok, so looking at it again there are a couple sets of numbers for each resolution so I can see how it could be confusing. On most of the 1080p numbers (medium and high/ultra) it looks like the 760 SLI and 780 card are virtually even, the 760 setup wins by a little on some and the 780 wins on a couple. The takeaway at the end of the day is that even though the 760 wins by a little it's small single digit gains and real world performance is virtually identical. That's why the author recommends the single card, because it's virtually the same but is simpler.

    You could argue the overclocking point because you'll still get a lot more out of a classified 780 than a SLI overclock setup (same is true of my setup vs a 780ti classified, I can't OC as much). Add that to the potential of adding a second 780 later and you get why the author recommends that and the same reason I was advocating the same advice. My original point was that even if the 760 wins by a little bit it's not worth messing with SLI for virtually no real world gains. Thus far I haven't run into any SLI issues myself, but my cards are so overkill right now it wouldn't matter...later in their lifespan I may regret going with this monitor and forcing myself into SLI.
     
    Last edited: Dec 13, 2013
  2. pabloottawa

    Joined:
    Jan 27, 2011
    Messages:
    266 (0.13/day)
    Thanks Received:
    15
    3DMark Vantage : Two 760's in SLI beat one 780 1/1

    3DMark11 Performance and Extreme: Two 760's in SLI beat one 780 2/2

    3DMark Fire Strike Normal and Extrreme: Two 760's in SLI beat one 780 2/2

    All the Battlefield 3 tests ( in 5 out of 8 tests) : Two 760's in SLI beat one 780 5/8

    All the BioShock Infinite tests (in 3 out of 8 tests): Two 760's in SLI beat one 780 3/8

    All the Crysis 2 tests (in 8 out of 8 tests): Two 760's in SLI beat one 780 8/8

    All the Crysis 3 tests (in 2 out of 7 tests): Two 760's in SLI beat one 780 2/7

    All the Dirt: Showdown tests (in 4 out of 4 tests): Two 760's in SLI beat one 780 4/4

    All the Far Cry 3 tests (in 6 out of 8 tests): Two 760's in SLI beat one 780 6/8

    All the Hitman: Absolution tests (in 8 out of 8 tests): Two 760's in SLI beat one 780 8/8

    All the Max Payne 3 tests (in 4 out of 8 tests): Two 760's in SLI beat one 780 4/8

    All the Metro 2033 tests (in 3 out of 4 tests): Two 760's in SLI beat one 780 3/4

    All the Metro: Last Light tests (in 2 out of 4 tests): Two 760's in SLI beat one 780 2/4

    All the The Elder Scrolls V Skyrim tests (in 6 out of 8 tests): Two 760's in SLI beat one 780 6/8

    All the Tomb Raider tests (in 5 out of 8 tests): Two 760's in SLI beat one 780 5/8

    All the Unreal 3 Engine tests (in 4 out of 4 tests): Two 760's in SLI beat one 780 4/4


    Conclusion:

    Out of all the FPS and response time test combined, 92 total: Two 760's in SLI beat one 780 65 times. In my opinion and my interpretation I'd say the 760's in SLI win and admittedly, one of the first things the author mentions before testing is: "We have to admit that we're not the biggest fans of SLI and CrossFireX"

    So it makes me wonder just how unbiased this article really is. With regards to raw power and FPS, two GTX 760's will beat a 780. This is what I was looking at when I made my decision. I really didn't care about scaling or syncronization..... these things to me are pretty minuscule when considering the power of a dual card setup.
     
  3. Aithos

    Joined:
    Sep 30, 2013
    Messages:
    175 (0.17/day)
    Thanks Received:
    35
    The only tests that matter are the 1080p game ones. The synthetic benchmarks are pretty meaningless because they aren't real world, and the higher resolution (triple monitor) ones are meaningless because you're not running a triple monitor setup and we should all realize that SLI scales better in higher resolution settings. If you had a 2560x1440 monitor instead of a 1080p I might change my recommendation to, but you don't. So there are two things that matter:

    1) 1080p medium/high settings benchmarks
    2) Real world difference

    Here are the 1080p benchmarks with a couple notes:

    BF3:
    medium - 760sli (165) 780 (140)
    ultra - 760sli (71) 780 (95) - they mislabeld this one, they show the 3 way as 71 and the 2 way as 113..

    Bioshock Infinite:
    medium - 760sli (203) 780 (176)
    ultra - 760sli (104) 780 (88)

    Crysis 2:
    very high - 760sli (97) 780 (97)
    ultra - 760sli (96) 780 (91)

    Crysis 3:
    medium - 760sli (84) 780 (105)
    very high - 760sli (51) 780 (45)

    Dirt Showdown:
    medium - 760sli (139) 780 (138)
    ultra - 760sli (106) 780 (103)

    Farcry 3:
    medium - 760sli (102) 780 (104)
    ultra - 760sli (66) 780 (55) - another one they mislabeled 3 vs 2 way sli

    Hitman Absolution:
    medium - 760sli (74) 780 (74)
    high - 760sli (63) 780 (63)

    Max Payne 3:
    normal - 760sli (197) 780 (190)
    highest - 760sli (124) 780 (113)

    Metro 2033:
    medium - 760sli (98) 780 (80)
    very high - 760sli (56) 780 (46)

    Metro Last Light:
    medium - 760sli (81) 780 (94)
    very high - 760sli (62) 780 (61)

    Skyrim:
    normal - 760sli (257) 780 (255) - noted as random because cpu limited in article
    high - 760sli (171) 780 (146)

    Tomb Raider:
    normal - 760sli (295) 780 (250)
    ultra - 760sli (64) 780 (56)

    So if you look at these results, you can immediately throw out everything in medium settings. Half of them are over 120-144fps which means it doesn't matter, and the rest are virtually identical. Only the Metro games seem to have any difference at medium and it flip flops which card wins, so I'll attribute that to bad drivers. Looking at the max settings (high/ultra) there are some differences (BF3, Bioshock, Farcry and Tomb Raider) but most of them are within a couple frames. Even the ones I listed aren't huge gains, and because of driver issues with SLI are bound to change. Every time NVIDIA releases new drivers SLI users will get difference performance in various games and it isn't always positive, that's why people bring up the micro stuttering and frame pacing. Framerate isn't a vaccuum where higher is always better. If it's higher but you're getting more tearing and stuttering then the experience is actually worse. You won't have that issue compounded by a single card.

    Then there is the overclocking thing, as I've previously mentioned...the 780 (which I'd get a classified and I'm sure they used reference for these benchmarks) will go considerably higher over stock than any SLI setup could hope to. There are some 780 classified cards that as a single card can rival a 780ti in pure power even with the lower CUDA cores. Again, the point isn't that the 780 flat out beats the 760s, it's that it's close enough that in real world terms it doesn't matter. That's why the article people recommend it, they (and most people) prefer single card solutions when the performance is close enough that you'd never see it when playing the games...
     

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)