Compute Performance

For our look at compute performance this is going to be a brief look. Our OpenGL AES and DirectCompute Fluid Simulation benchmarks simply don’t scale with multiple GPUs, so we’ll skip though (though the data is still available in Bench).

Our first compute benchmark comes from Civilization V, which uses DirectCompute to decompress textures on the fly. Civ V includes a sub-benchmark that exclusively tests the speed of their texture decompression algorithm by repeatedly decompressing the textures required for one of the game’s leader scenes. Note that this is a DX11 DirectCompute benchmark.

Given the nature of the benchmark, it’s not surprising that we see a performance regression here with some setups. The nature of this benchmark is that it doesn’t split across multiple GPUs well, though that doesn’t stop AMD and NVIDIA from tying. This doesn’t impact real game performance as we’ve seen, but it’s a good reminder of the potential pitfalls of multi-GPU configurations. Though AMD does deserve some credit here for gaining on their single GPU performance, pushing their lead even higher.

Our other compute benchmark is SmallLuxGPU, the GPU ray tracing branch of the open source LuxRender renderer. We’re now using a development build from the version 2.0 branch, and we’ve moved on to a more complex scene that hopefully will provide a greater challenge to our GPUs.

Unlike the Civ V compute benchmark, SLG scales very well with multiple GPUs, nearly doubling in performance. Unfortunately for NVIDIA GK104 shows its colors here as a compute-weak GPU, and even with two of them we’re nowhere close to one 7970, let alone the monster that is two. If you’re looking at doing serious GPGPU compute work, you should be looking at Fermi, Tahiti, or the future Big Kepler.

Civilization V Power, Temperature, & Noise
Comments Locked

200 Comments

View All Comments

  • CeriseCogburn - Friday, May 4, 2012 - link

    I disagree
  • chadwilson - Thursday, May 3, 2012 - link

    I have some issues with this article, the first of course being availability. Checking the past week, I have yet to see any availability of the 680 besides $200+ over retail premium cards on ebay. How can you justify covering yet another paper launch card without blaring bold print caveats, that for all intents and purposes, nVidia can't make for a very long time? There is a difference between ultra rare and non-existant.

    Is a card or chip really the fastest if it doesn't exist to be sold?

    Second, the issue of RAM, that's a problem in that most games are 32 bit, and as such, they can only address 3.5GB of RAM total between system and GPU RAM. This means you can have 12GB of RAM on your video card and the best you will ever get is 3GB worth of usage.

    Until games start getting written with 64 bit binaries (which won't happen until Xbox 720 since almost all PC games are console ports), anything more than 2-3GB GPU RAM is wasteful. We're still looking at 2014 until games even START using 64 bit binaries.

    Want it to change? Lobby your favorite gaming company. They're all dragging their feet, they're all complicit.
  • Ryan Smith - Thursday, May 3, 2012 - link

    Hi Chad;

    While I'm afraid we're not at liberty to discuss how many 680 and 690 cards NVIDIA has shipped, we do have our ears to the ground and as a result we have a decent idea as to how many have shipped. Suffice it to say, NVIDIA is shipping a fair number of cards; this is not a paper launch otherwise we would be calling NVIDIA out on it. NVIDIA absolutely needs to improve the stock situation, but at this point this is something that's out of their hands until either demand dies down or TSMC production picks up.

    -Thanks
    Ryan Smith
  • silverblue - Thursday, May 3, 2012 - link

    The 690 is a stunning product... but I'm left wanting to see the more mainstream offerings. That's really where NVIDIA will make its money, but we're just left wondering about supply issues and the fact that AMD isn't suffering to the same degree.
  • CeriseCogburn - Sunday, May 6, 2012 - link

    A single EVGA GTX680 sku at newegg has outsold the entire line up of 7870 and 7850 cards combined with verified owners reviews.
    So if availability is such a big deal, you had better ask yourselves why the 7870 and 7850 combined cannot keep pace with a single EVGA 680 card selling at Newegg.
    Go count them up - have at it - you shall see.
    108 sales for the single EVGA 680, more than the entire combined lot of all sku's in stock and out of the 7870 and 7850 combined total sales.
    So when you people complain, I check out facts - and I find you incorrect and failing almost 100% of the time.
    That's what happens when one repeats talking points like a sad PR politician, instead of checking available data.
  • ltcommanderdata - Thursday, May 3, 2012 - link

    Have you considered using WinZip 16.5 with it's OpenCL accelerated file compression/decompression as a compute benchmark? File compression/decompression is a common use case for all computer users, so could be the broadest application of GPGPU relevant to consumers if there is an actual benefit. The OpenCL acceleration in WinZip 16.5 is developed/promoted in association with AMD so it'll be interesting to see if it is hobbled on nVidia GPUs, as well as how well if scales with GPU power, whether it scales with SLI/dual GPU cards, and whether there are advantages with close IGP-CPU integration as with Llano and Ivy Bridge.
  • ViRGE - Thursday, May 3, 2012 - link

    Doesn't WinZip's OpenCL mode only work with AMD cards? If so, what use would that be in an NVIDIA review?
  • ltcommanderdata - Thursday, May 3, 2012 - link

    I actually don't know if it's AMD only. I know AMD worked on it together with WinZip. I just assumed that since it's OpenCL, it would be vendor/platform agnostic. Given AMD's complaints about use of vendor-specific CUDA in programs, if they developed an AMD-only OpenCL application, I would find that very disappointing.
  • ViRGE - Thursday, May 3, 2012 - link

    Going by their website it's only for AMD cards.

    "WinZip has been working closely with Advanced Micro Devices (AMD) to bring you a major leap in file compression technology. WinZip 16.5 uses OpenCL acceleration to leverage the significant power of AMD Fusion processors and AMD Radeon graphics hardware graphics processors (GPUs). The result? Dramatically faster compression abilities for users who have these AMD products installed! "
  • CeriseCogburn - Friday, May 4, 2012 - link

    Oh, amd the evil company up to it's no good breaking of openCL misdeeds again.
    Wow that's evil- the way it's meant to be unzipped.

Log in

Don't have an account? Sign up now