As I write this, it’s a beautiful Southern California Sunday afternoon. We’re deep into April and it’s already 80 degrees outside. I’m holding off on firing up the air conditioning for as long as possible, because that’s when power bills shoot up from $100 to $400.
The absolute last thing I need right now is a pair of GeForce GTX 480s (or any other high-end graphics cards, for that matter) pumping heat into the lab. Nevertheless, here I am, setting up a tower system with a Core i7-930 processor overclocked to 3.33 GHz, 6GB of Crucial memory, a 160GB Intel SSD, and…yeah, two GeForce GTX 480s (actually, I’m dropping three 480s into the machine, but for reasons I’ll explain in more depth, you probably don't want to try this at home quite yet).
Why are revisiting these cards a month after they first debuted?
When I first got my hands on the GeForce GTX 480 and 470, the boards were about six days from being officially unveiled by Nvidia. Getting the requisite results for my launch coverage meant setting up a sweat shop of benchmarking madness, owned and operated by yours truly. I used open-air racks and kept the numbers pouring in. I measured power and heat, determining that, damn, these things suck down quite a bit of juice and will readily fry your finger if you brush up against exposed metal. Moreover, I used Gigabyte’s X58A-UD5 motherboard, which doesn’t space its x16 PCIe links properly, forcing anyone using SLI to choose back-to-back x16 slots or a x16/x8 combo with room to breathe.
That’s not the way Nvidia recommends testing, though. The company suggests using one of a few different cases for the best acoustic experience, and a number of motherboards for the best balance between circulation and performance in SLI. So, with the help of Cooler Master and MSI, I built a machine that looks a lot like what I’d personally recommend to someone sinking serious cash into a gaming system.
At the same time, I thought it'd be prudent to revisit availability and 3D Vision Surround, the former of which is still dismally spotty, and the latter of which is still conspicuously missing in action.
Finally, I figured it’d be a great time to expound on my performance results. At launch, I was limited to just a couple charts worth of SLI testing on a motherboard that was only giving me one x16 link and one x8 connection. Without question, I wanted x16/x16 numbers, a larger sample of games, and even some 3-way numbers, if possible. MSI was kind enough to step up with a third GeForce GTX 480, yielding one-, two-, and three-card configurations, to which we’re able to compare scores from one and two Radeon HD 5870s. At the end of this menagerie, you’ll know whether SLIed GeForce GTX 480s or CrossFired Radeon HD 5870s get your more performance for your dollar.
I don’t mean to be too titillating of a tease here, but this one doesn’t end the way you think it does.