Introduction

When we first reviewed the new GeForce GTX 480 and 470 graphics cards back in March one of the few things we left short in our discussion was the possibility of running a pair of the cards in SLI.  We touched on it briefly, but only in passing and we really wanted to spend some quality time with the dual-GPU configurations to get a better feel for how the system would perform and how much actual benefit you would get going by doubling your investment cost.  After all, that is what it comes right down to: is spending twice as much money really going to increase your gaming performance and experience?

SLI Considerations

After a bit of delay with other reviews and a two week journey to Japan and Taiwan, we sat down with a pair of GTX 480s, GTX 470s (and yes even the new GTX 465s that were recently launched) to do our duty and game.  Sometimes it's tough, you know?

With any multiple graphics card configuration you have quite a few questions to ask yourself.  Do you have the right hardware to run SLI or CrossFire?  Do you have a chassis that is capable of containing all of that power?  Does your power supply live up to the demands of these high power GPUs?  Most modern motherboards for the AMD or Intel platform are going to support SLI and that should be something easily gauged on the motherboard box or on the motherboard manufacturer's website.  As for the power supply, that is a tougher question but I would generalize by saying a 850+ watt power supply from a quality vendor should be enough for someone running a pair of GTX 480s.  I say "generalize" because every system configuration is different and if you are running a set of 10 2TB hard drives in your system, you'll need to bump that power supply up a notch.

The chassis presents another issue worth noting: heat and noise.  NVIDIA has taken a lot of heat (only a slight pun intended) for the high temperatures and the noise levels of the fans on the GTX 400-series of cards under load.  Much of this criticism is warranted as even plugging in a second monitor is enough to significantly raise the idle temperatures on your graphics cards.  If you put a PAIR of them side by side (if your motherboard or hardware configuration doesn't permit spacing between them) that is only going to exacerbate the problem. 

To NVIDIA's credit, they are pushing the media to test their SLI configurations inside of actual cases rather than in an open-air environment.  Usually hardware reviewers are using open-air testing conditions for ease of access and to better and more reliably test the heat and noise variables.  (What better way to see if a card is hot than just reaching out and touching it to see if you get branded?)  We did much of our benchmarking in the same open air environment that we have traditionally used but we also spent some time with the GTX 480s shelled up inside a Silverstone Raven 2 case.

The Raven 2 is a unique design that actually mounts the system rotated 90 degrees clockwise - this has the added benefit of getting the heat of the GPUs (as well as everything else) rising out of the case.  In theory this helps aid in cooling as the fans are going WITH the laws of nature rather than against it.  And in truth putting the GTX 480s inside the Raven 2 did indeed negate many of the nagging sound issues arising from the fan on the GTX 400-series cards.  Temperatures benefited from the standardized airflow that is not really present on an open air setup and the noise was not only lessened because of that but quite frankly just because it was encased inside something.

NVIDIA wanted to prove to us that their cards are not TOO noisy and that
the complaints of the reviewers are unwarranted in relation to the
difference between them and the ATI Radeon 5000-series competition. 
They sort of got that point
across to this reviewer but I think it still has to be said that
NVIDIA's GTX 400 cards are louder and are hotter - but how much a
difference that makes to enthusiasts might be overblown.  These are not
the dust busters of GeForce FX days.

Of course what we were most interested
in seeing was the performance of the SLI configurations.  Let's take a
look at the testing setup and get to the results!