- Joined
- Feb 22, 2009
- Messages
- 757 (0.14/day)
System Name | Lenovo 17IMH05H |
---|---|
Processor | Core i7 10750H |
Video Card(s) | GTX 1660 Ti |
Audio Device(s) | SSL2 |
Software | Windows 10 Pro 22H2 |
Benchmark Scores | i've got a shitload of them in 15 years of TPU membership |
Question whether to invest into NVIDIA SLI setup bothers us every year, because every year new games are released, yet developers very often release the games too early and forget about SLI/CF or general performance.
Many gamers perhaps only now have the chance to upgrade their 2007 - 2010 gaming video cards into something modern and probably could not even play many awesome games on max settings, released since 2007. So now they want to know whether to buy one high-end card or two mid-range gaming cards for the same price - this is exactly what this review is about! Right now two GTX560 1 Gb cards are priced at half the price of GTX570 or HD6970 2 Gb and though it is clear that theoretically 2X GTX560 have more processing power than single high-end counterparts, the insufficient scaling of SLI might make them worse. We will test 40 games and find out should anyone invest into NVIDIA SLI.
Test setup
AMD Phenom II X6 1075T, 3000 MHz core
HyperX 4X2 Gb DDR3, 1333 MHz, CL7
Gigabyte GA-990XA-UD3
CobaNitrox IT-7750SG, 750 W, 80+
Corsair Force GT 60 Gb
SLI GeForce GTX560 1 Gb, 800 MHz core
Radeon HD6970 2 Gb, 890 MHz core
Windows 7 Pro SP1 64-bit/Forceware 296.10/Catalyst 12.3/
A mid-range PC for mid-range video cards sounds fair. Actually i have bought two GTX560 cards and my friend has bought a HD6970 - we both payed the same price. No overclock is applied on any of the hardware. Obviously HD6970 is no longer the crop of the cream card, but it is still a very competitive card. If you have not checked the many reviews about PCI-Express 8X/8X VS 16X/16X scaling difference, then i am telling you now - the difference is unnoticeable.
1) All of the tested games are tested on maximum possible settings, and all effects are turned on, except anti-aliasing is set to 4X, resolution set to 1600X1200.
2) Only games supporting over 60 FPS are tested.
3) Games not scaling in SLI will not be tested.
4) Game engines will not duplicate - every tested game has it's unique engine; for example only one Unreal 3 engine game will be tested.
5) Every game screenshot shown will represent the actual place in the game tested.
6) 15 seconds fraps benchmark is used, except in games those have their own bench. 1 fraps run is skipped, second or third is selected.
Now, before we start the test, immediately there 10 games, which do not scale at all or even scale negative, so do not use SLI on these games:
Blur 1.024
Alan wake 1.02
F1 2010 1.01
Neverwinter Nights 2 1.23
Singularity 1.1
Section 8 Prejudice
Starcraft 2 1.43
Dead Space 2 1.0
Assasin's Creed 1.02
Let's start the test! I will remind the anxious that the resolution in all games is 1600X1200 + 4X AA
Crysis Warhead DX10, patched 1.1.1711, CryEngine 2 engine
Guess it's no surprise which game starts the test - the father of all benchmark games Crysis! Even though SLI benefit is not great, it makes a difference between playable and unplayable.
Crysis 2 DX11 + high resolution texture pack, patched 1.9, CryEngine 3 engine
This is completely turned upside down from original Crysis - even a single GTX560 beats HD6970. SLI is a must, as it almost doubles the frame rates. Crushing NVIDIA victory.
Stalker Clear Sky DX10, patched 1.5.10, x-ray 1.5 engine
WTH is this? There is a game that lags more than Crysis and Metro? Yes! It's Stalker Clear Sky! But with SLI you can make unplayable into playable!
Battlefield Bad Company 2 DX11, unpatched, Frostbite 1.5 engine
Even unpatched BF BC 2 is SLI whoop-ass ready and takes over HD6970! Nice!
Metro 2033 DX11, patched, 4A engine
I did not use the in-game build benchmark, since it crashes often for no reason. SLI is an absolute must - it beats HD6970 and turns unplayable into playable.
Cryostasis DX10, patched 1.1, Atmosphere 4 engine
This game is most known not for it's extensive NVIDIA Physx use, but as the most unoptimized and worst laging game for ATI cards of all time - it used to produce some 10 FPS on lowest settings working even on HD5970! Now things have changed and the newest AMD cards can keep up the pace with NVIDIA cards, but the game is still unplayable with AMD cards, as there is a strange micro-stuttering when playing with AMD cards. Noway in hell do you play this game on AMD cards - never! Lucky for NVIDIA cards SLI works really nice!
Red Faction Guerrilla DX10, patched 1.02, Geo-Mod 2.0 engine
SLI does not scale over 100 %? Wrong. This is an amazing SLI own-age example.
Deus Ex Human Revolution DX11, patched 1.2.633, Crystal engine
I really had to find a "suitable" place in this game for frames to drop below 100 per second to stress it most... SLI GTX560 easy beats HD6970.
Aliens vs. Predator DX11, patched, Asura engine
When i first reviewed this game a year ago on CF HD4870, it was unpatched, and scaling was twice negative from single card! Now it is twice positive even though it is SLI and not CF - but that is still not enough for SLI GTX560 with almost 100 % scaling to beat a single HD6970!
FEAR 3 DX11, unpatched, Despair engine
Before Crysis became the father of all benchmarks, there was FEAR. All FEAR games scaled very well on both ATI and NVIDIA. The latest installment tends to keep the tradition as SLI doubles the frames and leaves HD6970 behind.
Far Cry 2 DX10, patched 1.02, Dunia engine
Actually SLI is not worth the deal - a single GTX560 will produce all the minimal frames you need. In-game benchmark was used.
Red Orchestra 2 Heroes Of Stalingrad DX9, patched, Unreal 3 engine
I have tested many Unreal 3 games and non of them scaled at all, but this one not only scales - it does SLI very good!
Call Of Juarez The Cartel DX11, patched 1.1, Chrome 5 engine
Single GTX560 beats HD6970, but SLI scaling is poor - same can be said for all the other Chrome 5 games. The options menu of this game is the worst i have ever seen and it looks more shitty than original Call Of Juarez - WTH???...
Syndicate DX9, unpatched, Starbreeze engine
It looks like there is no patch for this game yet, but SLI still works moderately. SLI is not recommended, however, as single GTX560 beats HD6970.
Jericho DX9, patched 1.01, Mercury engine
Older game does not mean it's bad for testing 2011 hardware - as it was a system hog back in 2007, it is perfect for testing modern video cards and boy... Sick SLI scaling over 100 % crushes HD6970 to dust! Remember that the price of SLI GTX560 is equal to HD6970...
Timeshift DX9, patched 1.02, Saber3D engine
Another older game, yet such a good SLI example - it is only that you will not be needing any SLI when one GTX560 can handle all.
Necrovision DX10, patched 1.2, Pain engine
The good old Painkiller Battle Out Of Hell would run smooth on these settings on a 6800GT at 120 FPS, yet Necrovision laughs at this fact and demands brutal hardware for an old, but modified 2004 engine. Both SLI and CF always scaled very good in this game and here SLI scaling is almost 100 %, beating HD6970.
Hard Reset DX9, patched, Road Hog engine
SLI scaling is weak and not enough to beat HD6970.
Serious Sam 3 DX10, unpatched, Serious 3.5 engine
SLI scaling is nothing special and thus can not touch HD6970. For a cartoon style game the graphics look just awesome, unlike in Serious Sam 2..
Duke Nukem Forever DX11, patched 1.01, Unreal 613 engine
Again i strove to find a place in this game where it would lag below 100 FPS - and i could not find one! No need for SLI really, but if it will work - it will beat HD6970.
Just Cause 2 DX10, patched, Avalanche 2.0 engine
Good SLI scaling still can not touch the mighty HD6970. SLI is still recommended as it scales over 50 %, but generally this is a really AMD biased game.
Dead Space DX9, unpatched 1.0.0.222, Visceral engine
Since Dead Space 2 was SLI non existing, i had to use the original game and even though there is no need for SLI to enjoy the game, it is amazing how well optimized this game is having in mind that there was never released a patch!
Dragon Age 2 DX11 + high resolution texture pack, patched 1.04, Lycium engine
Once again SLI makes a difference from unplayable to playable. Sadly Neverwinter Nights 2 did not scale at all, because in places it is unplayable with a single GTX560 just like at places Dragon Age 2 is.
Assassin's Creed Brotherhood DX10, patched 1.02, Anvil engine
The original game uses a different engine, but does not scale, this one, however, at least gives some try to scale and that is enough to beat HD6970.
Lost Planet Extreme Condition DX10, unpatched, MT Framework 1.2 engine
Could not test Lost Planet 2, because it simply kept crashing... Well whatever - the first game is also a demanding one and needs SLI badly! Even unpatched the game buffs the frames from SLI so good, that it beats HD6970.
PT Boats Knights Of The Sea DX10, patched 1.3, Storm 4.0 engine
Another example of "SLI over 100 %" - don't troll on me, i am not imagining things, just writing what i see on my monitor! Even though an older game, but with all the explosions, smoke and water splashes the frames can drop dramatically so SLI is an absolute must!
Battlestations Pacific DX9, patched 1.2, engine - eee what engine?
Never i have seen a so NVIDIA biased game... Suffice to say that an 8800GT trashes HD4870 to dust in this game, so there should not be any surprise that GTX560 does the same to HD6970.
HAWX 2 DX11, patched 1.01, again wtf is the engine of this?
How nice - a modern simulator and over 100 FPS! An in-game benchmark was used. Single GTX560 beats HD6970 and SLI is of the charts!
Dirt 2 DX11, patched 1.1, EGO engine
Since F1 2010, based in EGO 1.5 engine did not scale at all, i had to use an older example of the EGO engine and to my surprise the scaling is ballistic!
World In Conflict DX10, patched 1.009, Mass Tech engine
Do not trust the minimal frame rates of this game! I removed them since they show BS! For example on lower resolutions the minimal FPS would be lower than on 1600X1200 and in general the minimal frame rates are always the same within the range of 20 +/- 3 no mater what card you toss... SLI GTX560 is equal to HD6970 here.
-------------------------------------------------------------------------------------------------------------------------------------------------------
Conclusions:
SLI GTX560 wins in 24 out 30 games, 1 tie
HD6970 wins in 5 out of 30 games, 1 tie
Many gamers perhaps only now have the chance to upgrade their 2007 - 2010 gaming video cards into something modern and probably could not even play many awesome games on max settings, released since 2007. So now they want to know whether to buy one high-end card or two mid-range gaming cards for the same price - this is exactly what this review is about! Right now two GTX560 1 Gb cards are priced at half the price of GTX570 or HD6970 2 Gb and though it is clear that theoretically 2X GTX560 have more processing power than single high-end counterparts, the insufficient scaling of SLI might make them worse. We will test 40 games and find out should anyone invest into NVIDIA SLI.
Test setup
AMD Phenom II X6 1075T, 3000 MHz core
HyperX 4X2 Gb DDR3, 1333 MHz, CL7
Gigabyte GA-990XA-UD3
CobaNitrox IT-7750SG, 750 W, 80+
Corsair Force GT 60 Gb
SLI GeForce GTX560 1 Gb, 800 MHz core
Radeon HD6970 2 Gb, 890 MHz core
Windows 7 Pro SP1 64-bit/Forceware 296.10/Catalyst 12.3/
A mid-range PC for mid-range video cards sounds fair. Actually i have bought two GTX560 cards and my friend has bought a HD6970 - we both payed the same price. No overclock is applied on any of the hardware. Obviously HD6970 is no longer the crop of the cream card, but it is still a very competitive card. If you have not checked the many reviews about PCI-Express 8X/8X VS 16X/16X scaling difference, then i am telling you now - the difference is unnoticeable.
1) All of the tested games are tested on maximum possible settings, and all effects are turned on, except anti-aliasing is set to 4X, resolution set to 1600X1200.
2) Only games supporting over 60 FPS are tested.
3) Games not scaling in SLI will not be tested.
4) Game engines will not duplicate - every tested game has it's unique engine; for example only one Unreal 3 engine game will be tested.
5) Every game screenshot shown will represent the actual place in the game tested.
6) 15 seconds fraps benchmark is used, except in games those have their own bench. 1 fraps run is skipped, second or third is selected.
Now, before we start the test, immediately there 10 games, which do not scale at all or even scale negative, so do not use SLI on these games:
Blur 1.024
Alan wake 1.02
F1 2010 1.01
Neverwinter Nights 2 1.23
Singularity 1.1
Section 8 Prejudice
Starcraft 2 1.43
Dead Space 2 1.0
Assasin's Creed 1.02
Let's start the test! I will remind the anxious that the resolution in all games is 1600X1200 + 4X AA
Crysis Warhead DX10, patched 1.1.1711, CryEngine 2 engine
Guess it's no surprise which game starts the test - the father of all benchmark games Crysis! Even though SLI benefit is not great, it makes a difference between playable and unplayable.
Crysis 2 DX11 + high resolution texture pack, patched 1.9, CryEngine 3 engine
This is completely turned upside down from original Crysis - even a single GTX560 beats HD6970. SLI is a must, as it almost doubles the frame rates. Crushing NVIDIA victory.
Stalker Clear Sky DX10, patched 1.5.10, x-ray 1.5 engine
WTH is this? There is a game that lags more than Crysis and Metro? Yes! It's Stalker Clear Sky! But with SLI you can make unplayable into playable!
Battlefield Bad Company 2 DX11, unpatched, Frostbite 1.5 engine
Even unpatched BF BC 2 is SLI whoop-ass ready and takes over HD6970! Nice!
Metro 2033 DX11, patched, 4A engine
I did not use the in-game build benchmark, since it crashes often for no reason. SLI is an absolute must - it beats HD6970 and turns unplayable into playable.
Cryostasis DX10, patched 1.1, Atmosphere 4 engine
This game is most known not for it's extensive NVIDIA Physx use, but as the most unoptimized and worst laging game for ATI cards of all time - it used to produce some 10 FPS on lowest settings working even on HD5970! Now things have changed and the newest AMD cards can keep up the pace with NVIDIA cards, but the game is still unplayable with AMD cards, as there is a strange micro-stuttering when playing with AMD cards. Noway in hell do you play this game on AMD cards - never! Lucky for NVIDIA cards SLI works really nice!
Red Faction Guerrilla DX10, patched 1.02, Geo-Mod 2.0 engine
SLI does not scale over 100 %? Wrong. This is an amazing SLI own-age example.
Deus Ex Human Revolution DX11, patched 1.2.633, Crystal engine
I really had to find a "suitable" place in this game for frames to drop below 100 per second to stress it most... SLI GTX560 easy beats HD6970.
Aliens vs. Predator DX11, patched, Asura engine
When i first reviewed this game a year ago on CF HD4870, it was unpatched, and scaling was twice negative from single card! Now it is twice positive even though it is SLI and not CF - but that is still not enough for SLI GTX560 with almost 100 % scaling to beat a single HD6970!
FEAR 3 DX11, unpatched, Despair engine
Before Crysis became the father of all benchmarks, there was FEAR. All FEAR games scaled very well on both ATI and NVIDIA. The latest installment tends to keep the tradition as SLI doubles the frames and leaves HD6970 behind.
Far Cry 2 DX10, patched 1.02, Dunia engine
Actually SLI is not worth the deal - a single GTX560 will produce all the minimal frames you need. In-game benchmark was used.
Red Orchestra 2 Heroes Of Stalingrad DX9, patched, Unreal 3 engine
I have tested many Unreal 3 games and non of them scaled at all, but this one not only scales - it does SLI very good!
Call Of Juarez The Cartel DX11, patched 1.1, Chrome 5 engine
Single GTX560 beats HD6970, but SLI scaling is poor - same can be said for all the other Chrome 5 games. The options menu of this game is the worst i have ever seen and it looks more shitty than original Call Of Juarez - WTH???...
Syndicate DX9, unpatched, Starbreeze engine
It looks like there is no patch for this game yet, but SLI still works moderately. SLI is not recommended, however, as single GTX560 beats HD6970.
Jericho DX9, patched 1.01, Mercury engine
Older game does not mean it's bad for testing 2011 hardware - as it was a system hog back in 2007, it is perfect for testing modern video cards and boy... Sick SLI scaling over 100 % crushes HD6970 to dust! Remember that the price of SLI GTX560 is equal to HD6970...
Timeshift DX9, patched 1.02, Saber3D engine
Another older game, yet such a good SLI example - it is only that you will not be needing any SLI when one GTX560 can handle all.
Necrovision DX10, patched 1.2, Pain engine
The good old Painkiller Battle Out Of Hell would run smooth on these settings on a 6800GT at 120 FPS, yet Necrovision laughs at this fact and demands brutal hardware for an old, but modified 2004 engine. Both SLI and CF always scaled very good in this game and here SLI scaling is almost 100 %, beating HD6970.
Hard Reset DX9, patched, Road Hog engine
SLI scaling is weak and not enough to beat HD6970.
Serious Sam 3 DX10, unpatched, Serious 3.5 engine
SLI scaling is nothing special and thus can not touch HD6970. For a cartoon style game the graphics look just awesome, unlike in Serious Sam 2..
Duke Nukem Forever DX11, patched 1.01, Unreal 613 engine
Again i strove to find a place in this game where it would lag below 100 FPS - and i could not find one! No need for SLI really, but if it will work - it will beat HD6970.
Just Cause 2 DX10, patched, Avalanche 2.0 engine
Good SLI scaling still can not touch the mighty HD6970. SLI is still recommended as it scales over 50 %, but generally this is a really AMD biased game.
Dead Space DX9, unpatched 1.0.0.222, Visceral engine
Since Dead Space 2 was SLI non existing, i had to use the original game and even though there is no need for SLI to enjoy the game, it is amazing how well optimized this game is having in mind that there was never released a patch!
Dragon Age 2 DX11 + high resolution texture pack, patched 1.04, Lycium engine
Once again SLI makes a difference from unplayable to playable. Sadly Neverwinter Nights 2 did not scale at all, because in places it is unplayable with a single GTX560 just like at places Dragon Age 2 is.
Assassin's Creed Brotherhood DX10, patched 1.02, Anvil engine
The original game uses a different engine, but does not scale, this one, however, at least gives some try to scale and that is enough to beat HD6970.
Lost Planet Extreme Condition DX10, unpatched, MT Framework 1.2 engine
Could not test Lost Planet 2, because it simply kept crashing... Well whatever - the first game is also a demanding one and needs SLI badly! Even unpatched the game buffs the frames from SLI so good, that it beats HD6970.
PT Boats Knights Of The Sea DX10, patched 1.3, Storm 4.0 engine
Another example of "SLI over 100 %" - don't troll on me, i am not imagining things, just writing what i see on my monitor! Even though an older game, but with all the explosions, smoke and water splashes the frames can drop dramatically so SLI is an absolute must!
Battlestations Pacific DX9, patched 1.2, engine - eee what engine?
Never i have seen a so NVIDIA biased game... Suffice to say that an 8800GT trashes HD4870 to dust in this game, so there should not be any surprise that GTX560 does the same to HD6970.
HAWX 2 DX11, patched 1.01, again wtf is the engine of this?
How nice - a modern simulator and over 100 FPS! An in-game benchmark was used. Single GTX560 beats HD6970 and SLI is of the charts!
Dirt 2 DX11, patched 1.1, EGO engine
Since F1 2010, based in EGO 1.5 engine did not scale at all, i had to use an older example of the EGO engine and to my surprise the scaling is ballistic!
World In Conflict DX10, patched 1.009, Mass Tech engine
Do not trust the minimal frame rates of this game! I removed them since they show BS! For example on lower resolutions the minimal FPS would be lower than on 1600X1200 and in general the minimal frame rates are always the same within the range of 20 +/- 3 no mater what card you toss... SLI GTX560 is equal to HD6970 here.
-------------------------------------------------------------------------------------------------------------------------------------------------------
Conclusions:
SLI GTX560 wins in 24 out 30 games, 1 tie
HD6970 wins in 5 out of 30 games, 1 tie
Last edited: