EVGA's GeForce GTX 560 Ti 2Win: The Raw Power Of Two GPUs

Final Thoughts

If you’ve read our GPU reviews for any length of time then you’re probably familiar with our editorial stance on multi-GPU configurations: multi-GPU is nice to have, but only after you’ve exhausted single-GPU performance. Normally this is an easily defensible position, as dual-GPU cards are priced well above single-GPU cards and multi-GPU otherwise involves the hassle of dealing with multiple video cards.

EVGA’s GeForce GTX 560 Ti 2Win makes that a much harder position to defend. Technically speaking it’s pricier than a GTX 580, but not significantly so; it’s a GTX 580 competitor instead of something that comes after the GTX 580. Furthermore as we already know from the regular GTX 560 Ti SLI, a pair of GTX 560 Tis can beat a single GTX 580 by 30% if not more under the right circumstances. On a pure performance per dollar basis the 2Win is considerably faster than the GTX 580, and that’s a fact that’s very hard to argue with.

Ultimately the existence of the 2Win is a major vote of confidence in SLI by EVGA. If you believe as they do – that NVIDIA will continue to quickly add SLI support to games, that SLI scaling will always be strong, and that multi-GPU timing issues are easily overcome – then the 2Win makes the GeForce GTX 580 redundant at current pricing. It’s that simple.

On the other hand if you don’t share EVGA’s confidence in SLI, then very little has changed. If you believe that new games will have teething issues with SLI, that microstutter will continue to exist, and that not every game will scale well with SLI, then the 2Win is a poor choice in light of the more consistent performance of the GTX 580. Certainly the performance of the 2Win is phenomenal when SLI is working, but if SLI falters that means the performance of the 2Win is reduced to that of a $220 GTX 560 Ti. It’s not necessarily a deal breaker, but it’s a real concern that must be evaluated when buying any dual-GPU card, including the 2Win. We’re going to continue to be conservative and recommend the consistency of a single-GPU card over the performance of a dual-GPU card, but for the individual buyer the 2Win’s performance makes a very good argument to throw caution into the wind.

Moving beyond the scope of SLI it’s clear that the 2Win is a solid product. EVGA’s use of an open air cooler is definitely an interesting choice. It’s not the only card using this style of a cooler – a number of overclocking focused vendor custom cards do so – but it’s more surprising to see it on a multi-GPU card. The end result is that given a suitable case this cooler allows the 2Win to dissipate as much heat as it does for relatively little noise. It’s subjectively noisier than a GTX 580, but just barely.

Wrapping things up, the only aspect I feel that EVGA has left underdeveloped on an otherwise very strong card is VRAM. As a result of SLI 2Win is a $520 card with 1GB of effective VRAM. We’ve already seen 1GB of VRAM pose limitations in a couple of our tests, and going forward it’s only going to get worse.  Case in point: Battlefield 3, which we’re currently looking at. In a technical presentation DICE has stated that the combined memory consumption at 1920x1080 for the gbuffer, Z-buffer, and MSAA resolve data is 158MB; and this is before other buffers let alone textures. As a $200 card meant for 1920 and lower resolution, 1GB of VRAM makes sense for the GTX 560 Ti. But as a $500 dual-GPU card meant for higher performance, higher quality, and higher resolutions, 1GB of effective VRAM is the biggest bottleneck going forward for the 2Win. Realistically EVGA is in a hard place since using higher density GDDR5 would drive up the price of the card and make it even more expensive than the GTX 580, but at the end of the day I think the 2Win needs 2GB of effective VRAM to spread its wings through 2012.