GeForce GTX 680, Part 2: SLI, 5760x1080, And Overclocking

Noise And Temperature

One of the biggest worries to address when it comes to adding a second graphics card is noise, which is of course an artifact of greater power consumption leading to increased heat dissipation.

Fortunately, at idle, two GeForce GTX 680s are still quieter than a single GeForce GTX 580. And although the inside card gets a little bit warmer, a single degree is hardly worth sweating.

Strangely, the inside Radeon HD 7970 heats up quite a bit more (though it’s worth noting that the outside card switches off completely and cools down to about 33 degrees).

We wouldn’t have expected trouble at idle; under load is where you see all of the action.

In this case, the inside GeForce GTX 680 gets up to 83 degrees pretty easily—warmer than a GeForce GTX 580 under load, but not as bad as prior-generation dual-slot boards.

Fortunately, the GeForce GTX 680’s fans don’t have to blow particularly hard to maintain those temperatures. A pair of GK104-based cards gets louder than a GeForce GTX 590, but they’re still less noisy than a single Radeon HD 7970. Two 7970s are pretty distracting.

Chris Angelini
Chris Angelini is an Editor Emeritus at Tom's Hardware US. He edits hardware reviews and covers high-profile CPU and GPU launches.
  • tacoslave
    amds driver team needs to get off its ass i mean look at those crossfire results thats downright pitiful .
    Reply
  • bystander
    tacoslaveamds driver team needs to get off its ass i mean look at those crossfire results thats downright pitiful .What is wrong with their crossfire performance?

    You do have to look at the high resolution benchmarks to see actual crossfire results, as 1080p benchmarks are being bottlenecked by the CPU.
    Reply
  • I am a PC
    It is quite clear that the Radeons are more powerful but that once again the Nvidia favoring benchmark suite once again favors Nvidia.
    Reply
  • Would like to see CUDA compute instead of openCL stuff. Something like Blender/Cycles benchmark.
    Reply
  • cangelini
    I am a PCIt is quite clear that the Radeons are more powerful but that once again the Nvidia favoring benchmark suite once again favors Nvidia.If you were to hand-pick a suite to favor AMD, what would it include? :) You did notice the Radeons doing really well in Battlefield and Metro, right?
    Reply
  • cangelini
    AgonothetaWould like to see CUDA compute instead of openCL stuff. Something like Blender/Cycles benchmark.Our Blender test is being working on right now--currently we're only utilizing the Tiles/Cycles engines for CPU reviews.
    Reply
  • My MSI 7870 came with Catalyst 12.3, so why use 12.2 if 12.3 is out there?
    Reply
  • weatherdude
    You know, I'm starting to wonder what the Tom's Hardware labs are like. I hope it has something to keep the staff sane as they run tests over and over again.

    Anyways, looks like the competition in the GPU world is going strong right now. To me both the GTX 680 and Radeon 7970 are fine pieces of work. The general compute performance of Tahiti is really really good though so will AMD really reduce the prices significantly below the GTX 680?

    Great review as usual. All your hard work is appreciated.
    Reply
  • cangelini
    sniper13xMy MSI 7870 came with Catalyst 12.3, so why use 12.2 if 12.3 is out there?Because it's not?
    Reply
  • cangelini
    weatherdudeYou know, I'm starting to wonder what the Tom's Hardware labs are like. I hope it has something to keep the staff sane as they run tests over and over again.Anyways, looks like the competition in the GPU world is going strong right now. To me both the GTX 680 and Radeon 7970 are fine pieces of work. The general compute performance of Tahiti is really really good though so will AMD really reduce the prices significantly below the GTX 680?Great review as usual. All your hard work is appreciated.If ever you're in Bakersfield, CA, you're welcome to drop by and check the lab out. It's like a gamer candy store, literally stacked with graphics cards higher than I can reach!
    Reply