Sign in with
Sign up | Sign in

GPGPU Benchmarks

Meet The 2012 Graphics Charts: How We're Testing This Year
By

Introduction to the GPGPU Benchmarks

Now that OpenCL is established as a multi-platform GPU-computing API, it's relatively simple to benchmark general-purpose GPU application performance, as long as the applications are truly platform agnostic. Previously, when the GPGPU universe was divided into CUDA (Nvidia) and Stream (AMD), we faced the problem that most applications supported only one of the two environments, and could thus not be directly compared to each other.

Bitcoin Mining

Apart from synthetic benchmarks, Bitcoin mining is one of the few practically useful, albeit somewhat narrowly focused, examples for GPU computing. A problem of using Bitcoin mining as a benchmark is that the existing OpenCL miner will typically not support a new graphics card generation upon release. Thus, we have to keep updating the OpenCL miner and benchmark results.

LuxMark

LuxMark is the second practical GPGPU application in our benchmark suite. It's based on the free render program LuxRender. A second version of this benchmark is now available, but we still use the well-known scene LuxBall HDR; we want to test a broad range of graphics cards, and this scene is the least-complex one. The other two test scenes are too difficult for entry-level cards. Since we want you to be able to compare the GPGPU scores of graphics cards with the GPGPU scores achieved by APUs and IGPs, we don’t plan to replace this popular scene with another one until after 2012 at the earliest.

GPU Caps Viewer

GPU Caps Viewer is a synthetic benchmark with an interesting mix of computation done via OpenCL and post-processing and graphical output with or without anti-aliasing. The PostFX test is a variant of the Nvidia demo for the API call oclPostprocessGL in the company's GPU Computing SDK, where a blur effect is added in a post-processing step after the image has been rendered.

The particle test is traditionally one of Nvidia's strengths. However, AMD’s new GCN architecture is catching up.

N Queens

The so-called N Queens Problem has nothing to do with Freddy Mercury; it's a complex mathematical chess problem. Eight queens need to be positioned on a chess board in such a way that none of them could take another one according to chess rules (but disregarding black and white). The goal is to obtain the full solution set in the minimum time.

Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 85 comments.
This thread is closed for comments
Top Comments
  • 21 Hide
    quixoticism , April 3, 2012 6:46 AM
    I would have preferred to see all the benchmarks at 1920x1080 done with maxed out graphics settings.
    I want to see how the base performance is in games with full detail, and save the heavy AA and resolutions for extreme.
  • 19 Hide
    randomkid , April 3, 2012 5:51 AM
    Where's the 5760x1080? In the area where I come from, 3x 1920x1080p 22" monitor cost around the same or even less than a single 2560x1440/1600 27" monitor so this is a more likely configuration among gamers.

    The 5760x1080 resolution will also push the GPU's harder than a 2560x1440/1600 could so why limit the resolution there?
  • 19 Hide
    johnny_utah , April 3, 2012 4:47 AM
    While I love the new techniques, using BITCOIN to bench GPUGPU performance instead of Folding @ Home? Um, okay.
Other Comments
  • 19 Hide
    johnny_utah , April 3, 2012 4:47 AM
    While I love the new techniques, using BITCOIN to bench GPUGPU performance instead of Folding @ Home? Um, okay.
  • 15 Hide
    Anonymous , April 3, 2012 4:49 AM
    Still with the bar charts? Would *love* to see scatter plots with price/score on the axes... So much more useful in picking out a card.
  • 3 Hide
    AznCracker , April 3, 2012 5:10 AM
    Man the charts are dying to be updated. Too bad it isn't done more often since it takes a lot of work.
  • 5 Hide
    Anonymous , April 3, 2012 5:12 AM
    You havent added how many cheese wheels it can run in skyrim as a benchmark... wth?
  • 9 Hide
    DjEaZy , April 3, 2012 5:14 AM
    ... i like the pile of card's @ the end of the article.... a beautiful pile...
  • 14 Hide
    pharoahhalfdead , April 3, 2012 5:32 AM
    johnny_utahWhile I love the new techniques, using BITCOIN to bench GPUGPU performance instead of Folding @ Home? Um, okay.


    I agree. I know Tom's spends a lot of time benchmarking, but Folding@home is something that is a bit more common. I would love to see F@H in some articles.

    BTW, I appreciate all the work you guys do.
  • 19 Hide
    randomkid , April 3, 2012 5:51 AM
    Where's the 5760x1080? In the area where I come from, 3x 1920x1080p 22" monitor cost around the same or even less than a single 2560x1440/1600 27" monitor so this is a more likely configuration among gamers.

    The 5760x1080 resolution will also push the GPU's harder than a 2560x1440/1600 could so why limit the resolution there?
  • 9 Hide
    Anonymous , April 3, 2012 5:56 AM
    Quote:
    We'll add up to 20 new boards each month until the lower end of the performance range is filled out, too.

    How far back in GPU generations are you going to test, if at all? I saw the power consumption charts and could only see GTX 500, 600 and Radeon 6000, 7000 series. I have an EVGA GTX 480 SC for two years and do like to know how it compares to the newer series of GPUs. Much appreciated.
  • 8 Hide
    Yargnit , April 3, 2012 5:58 AM
    MMO FanYup no surprise here typical Nvidia benchmark suite fuck sakes.


    So what would YOU like to see used then? If they were trying to push Nvidia wouldn't Hawx 2 be in the suite?
  • 16 Hide
    shinym , April 3, 2012 6:18 AM
    For Starcraft II you say "This game doesn't stress the CPU, and is thus well-suited for GPU benchmarking." Looks like you got CPU and GPU mixed up there.
  • 5 Hide
    cangelini , April 3, 2012 6:28 AM
    shinymFor Starcraft II you say "This game doesn't stress the CPU, and is thus well-suited for GPU benchmarking." Looks like you got CPU and GPU mixed up there.

    More than likely, it's the sequence the German team picked that is less CPU-bound than other StarCraft II tests we've used.
  • 10 Hide
    CaedenV , April 3, 2012 6:31 AM
    CommieIBankerStill with the bar charts? Would *love* to see scatter plots with price/score on the axes... So much more useful in picking out a card.

    I would love something like this as well, but as Tom's has a world wide market and the prices vary so greatly from place to place such charts become impractical.

    What I miss about the old charts is that you could easily compare old GPUs to new ones. When upgrading I think people like to have a reference of what they currently have compared to what they are looking at getting. To throw in a few old cards like the 8600GT, 8/9800GT in the mix may help people feel more secure in their purchases, and have a better feel for what they are getting because they can relate the new card to their old one. Not saying Tom's needs to put in every single card from every generation, but a card or 2 from each generation from the last 6 years or so would be nice.

    Another thing I would like is something like Anandtech's bench where when you select your cards you get bar graphs instead of raw numbers in a chart. It is just easier to visually see a 1/3rd difference in performance on a line graph rather than in raw numbers.
  • 21 Hide
    quixoticism , April 3, 2012 6:46 AM
    I would have preferred to see all the benchmarks at 1920x1080 done with maxed out graphics settings.
    I want to see how the base performance is in games with full detail, and save the heavy AA and resolutions for extreme.
  • 0 Hide
    CaedenV , April 3, 2012 6:47 AM
    while speaking of comparison chart complaints:
    -Some cards (though not all) show up as their name, others have a picture of the card with a sales link, and others give very little information at all. It would be nice to make this consistent so that at the top of each column we could see the pic (if available), the name of the card, and then a sales link (hey, you have to pay of the site somehow)

    -organize the comparison charts a little. On the comparison page it just throws all the charts together with no apparent rhyme or reason. It would be nice to have groupings such as physical considerations (temps, noise, and power usage), gpgpu benchmarks, and game benchmarks sorted either by game.

    -Could we add physical dimensions? It would be helpful to some to know how long a card is, and now many slots it takes.

    -lastly, under "02 - Unigine Heaven DirectX 11 Performance 1920x1080, 4xAA, AFx16 Shaders medium Tessellation normal" the "score in" has a blank spot to enter in text (pretty sure it should say 'FPS' here) followed by a broken "Go" link.

    None of these changes should be all that hard to make, and would make the charts much simpler to use when comparing specific cards.
  • 2 Hide
    Anonymous , April 3, 2012 6:53 AM
    Glad to see Bitcoin Mining being included now.
  • 14 Hide
    kyuuketsuki , April 3, 2012 8:50 AM
    quixoticismI would have preferred to see all the benchmarks at 1920x1080 done with maxed out graphics settings.I want to see how the base performance is in games with full detail, and save the heavy AA and resolutions for extreme.

    This. I, along with probably a large chunk of your readers, are probably more interested in performance at maxed out settings at 1080p than skipping from mid-range settings at 1080p to maxed out settings at higher resolutions.
  • 1 Hide
    Anonymous , April 3, 2012 9:10 AM
    Thanks for your work Guys,reading trough since long time and each rig has been built based upon your tests!
    Gabriel
  • 2 Hide
    devBunny , April 3, 2012 12:15 PM
    "What would you rather listen to, a low hum or a high-pitched whine at the same loudness?"

    Lol. Good question. My tinnitus gives me a couple of the lows and several highs (but, just for fun and to keep me noticing them, the pitch and loudness vary). To be honest, I'd rather not hear any of them! ;O)
  • -1 Hide
    devBunny , April 3, 2012 12:33 PM
    Dear Tom's

    Regarding charts, I think it's way past time that you provided dynamic, configurable graphs generated from the data rather than static images. Images are fine if that's all that the web savvy that a site has. Surely Tom's is better than that?! [tease, tease] ;o)

    I'd like graphs that have options, including choice of colours (I sometimes can't easily see the difference between the colours that you use) and scale (so that I don't have to opening images in another tab and manually enlarging them to read the finer details).

    Someone above mentioned scatter diagrams. And why not, if that would suit him? A page that downloads the data rather than an image has scope for creating whatever kind of graph can help represent the data meaningfully.

    Tom's, I invite you to kill me with kindness. ;-)

    ps. By "dynamic", I don't mean animated. That's pointless, even annoying, eye candy. It's only necessary to show the data more clearly, in a form that people prefer.
Display more comments