Crossfire vs SLI comparison on t-break using HD3870 and 8800GT cards.

[warhammer]

Hey all Interesting article regarding our Favourite subject enjoy the read.

http://www.tbreak.com/reviews/article.php?cat=grfx&id=580&pagenumber=1

Comparison

3DMarks: Two HD 3870 cards are definitely more effective than two 8800GT cards in all Futuremark tests.This extra efficiency also makes the Crossfire post higher scores than two 8800GT cards in SLI making AMD the clear winner.

CoH - Opposing Fronts: Moving on to our first game, the SLI setup has a clear advantage over the Crossfire setup posting higher efficiency and thus higher scores as the 8800GT is faster than the HD3870 in single card mode.

Crysis: Crysis prefers nVidia cards in single and SLI modes. The efficiency of two 8800GTs in SLI is no match for Crossfire in Crysis making nVidia the card of choice for this game.

ET- Quake Wars: The Quake Engine scales very well under SLI as well as Crossfire tilting slightly in favor of Crossfire. However, the slight advantage for AMD is not enough to match the raw power of the 8800GT which posts much higher scores.

F.E.A.R. Like Quake Wards, F.E.A.R. also scales well under both SLI and Crossfire setups, however, it tilts slightly in SLI's favor. That, combined with the higher scores of 8800GT make it a faster on nVidia's GPUs.

Half Life 2- Episode 2: No real surprises here as AMD has been heavily promoting Half Life 2 under their camp and although the differences are higher at lower resoluiotn/settings, the efficiency between SLI and Crossfire is pretty close at 1920x1200 with 4X AA. Because of this, the Crossfire setup loses at this particular resolution and detail but manages to keep up or takeover the nVidia in other tests thus making it the winner.

Lost Planet - Extreme Condition: Again, Crossfire proves to me no match for SLI which posts much higher efficiencies across all resolutions/details. Adding to that the raw horsepower provided by single cards, Lost Planet is certainly a game you'd want to play on the 8800GT.

S.T.A.L.K.E.R.: Things take a turn in Crossfire's favor with S.T.A.L.K.E.R. where it posts higher efficiencies, especially towards the lower resolution and details. Unfortunately, this still doesn't allow the Crossfire setup to post faster scores than the SLI setup.

Unreal Tournament 3: Again, we see Crossfire posting higher efficiency at lower resolution/details and this time around, the difference is enough to make it faster than an SLI setup at lower resolution/details but as we crank these two up, SLI manages to take over.

Word in Conflict: Ending up our games, Crossfire takes the lead in efficiency which also makes it post higher scores than an SLI setup even at high resolution/details.

 

[Xazax]

If you ask me, it really depends on the engine and programming. You can see the cards trade blows in certian games, take CoH thats a heavily promoted and well optimized for Nvidia's card where as HL 2/Eps. 1/2, are much in the favor of CF, but take in consideration the graphics engine where HL 2 takes very much advantage of the 320 SPs, where as Nvidia only has 112(though higher clocked).

 

[wolf]

six of one and half a dozen of the other, alot of people i hear almost bragging about how ATi has "much better" crossfire efficiency, i dont believe it for a second.

cards perform differently on different games in single GPU mode let alone CF/SLi, and the results vary across the board.

as usual my opinion states that a choice in cards comes down ONLY to personal preference, as the price/performance ratio on 3850/3870 and 8800 cards is pretty much spot on, especially since generally in Dual Gpu modes, the difference is so insignificant you couldn't perceive the difference unless fraps told you anyway.

 

[vega22]

nice even test that, with the ati cards being feed by ddr3 but the nvidia ones on ddr2, not much difference in the mem bendwidth then no?

 

[wolf]

the word around the water cooler is DDR3 is pretty superfluous so far.

 

[vega22]

between 3 and 7% increase in mem bandwidth (and im sure this will widen with penryn on the higher fsb) is still better than none, its not that its not better performing just the gains aren't worth the excesive cost atm.

its not as if this is really news any way, xfire has scaled better than sli for quite some time but the margin is quite slim in most cases. the difference isnt that much in most games and depending on which games you play you should go with what ever performs best in those.

i love the way they go on about the source engine, its a very cpu bound engine and that, for one, is one of the tests were the ddr3 will shine over the ddr2 setup.

personally i dont like reviews where they dont give the things on test a even playing feild, it show a bias from the onset, its not as if you cant get a x38 with ddr2 slots to even the test is it?

 

[niko084]

If you ask me, it really depends on the engine and programming. You can see the cards trade blows in certian games, take CoH thats a heavily promoted and well optimized for Nvidia's card where as HL 2/Eps. 1/2, are much in the favor of CF, but take in consideration the graphics engine where HL 2 takes very much advantage of the 320 SPs, where as Nvidia only has 112(though higher clocked).

You hit it right on the nose...

There was a team that tore apart a game, I don't remember what it was, but it has been proven that when a game is directly built for a video card they can show as much as a 300% better frame rate over a card that is not well supported by the same game... Now this varies from game to game obviously and different cards vary differently.

But yes the direct programming of a lot of games specifically for Nvidia cards really hurts ATI in the performance scores.... Look at how bad cheap ATI cards beat Nvidia's top cards in early Half life and CS:S..... Thats a good show for what direct card favoring can do...