Compute

Shifting gears, as always our final set of benchmarks is a look at compute performance. As we have seen with GTX 680, GK104 appears to be significantly less balanced between rendering and compute performance than GF110 or GF114 were, and as a result compute performance suffers.  Cache and register file pressure in particular seem to give GK104 grief, which means that GK104 can still do well in certain scenarios, but falls well short in others.

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.

It’s quite shocking to see the GTX 670 do so well here. For sure it’s struggling relative to the Radeon HD 7900 series and the GTX 500 series, but compared to the GTX 680 it’s only trailing by 4%. This is a test that should cause the gap between the two cards to open up due to the lack of shader performance, but clearly that this not the case. Perhaps we’ve been underestimating the memory bandwidth needs of this test? If that’s the case, given AMD’s significant memory bandwidth advantage it certainly helps to cement the 7970’s lead.

Our next 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.

SmallLuxGPU on the other hand finally shows us that larger gap we’ve been expecting between the GTX 670 and GTX 680. The GTX 680’s larger number of SMXes and higher clockspeed cause the GTX 670 to fall behind by 10%, performing worse than the GTX 570 or even the GTX 470. More so than any other test, this is the test that drives home the point that GK104 isn’t a strong compute GPU while AMD offers nothing short of incredible compute performance.

For our next benchmark we’re looking at AESEncryptDecrypt, an OpenCL AES encryption routine that AES encrypts/decrypts an 8K x 8K pixel square image file. The results of this benchmark are the average time to encrypt the image over a number of iterations of the AES cypher.

Once again the GTX 670 has a weak showing here, although not as bad as with SmallLuxGPU. Still, it’s enough to fall behind the GTX 570; but at least it’s enough to beat the 7950. Clockspeeds help as showcased by the EVGA GTX 670SC but nothing really makes up for the missing SMX.

Our foruth benchmark is once again looking at compute shader performance, this time through the Fluid simulation sample in the DirectX SDK. This program simulates the motion and interactions of a 16k particle fluid using a compute shader, with a choice of several different algorithms. In this case we’re using an (O)n^2 nearest neighbor method that is optimized by using shared memory to cache data.

For reasons we’ve yet to determine, this benchmark strongly dislikes GTX 670 in particular. There doesn’t seem to be a performance regression in NVIDIA’s drivers, and there’s not an incredible gap due to TDP, it just struggles on the GTX 670. As a result performance of the GTC 670 only hits 42% of the GTX 680, which is well below what the GTX 670 should theoretically be getting. Barring some kind of esoteric reaction between this program and the unbalanced GPC a driver issue is still the most likely culprit, but it looks to only affect the GTX 670.

Finally, we’re adding one last benchmark to our compute run. NVIDIA  and the Folding@Home group have sent over a benchmarkable version of the client with preliminary optimizations for GK104. Folding@Home and similar initiatives are still one of the most popular consumer compute workloads, so it’s something NVIDIA wants their GPUs to do well at.

Whenever NVIDIA sends over a benchmark you can expect they have good reason to, and this is certainly the case for Folding@Home. GK104 is still a slouch given its resources compared to GF110, but at least it can surpass the GTX 580. At 970 nanoseconds per day the GTX 670 can tie the GTX 580, while the GTX 680 can pull ahead by 6%. Interestingly this benchmark appears to be far more constrained by clockspeed than the number of shaders, as the EVGA GTX 670SC outperforms the GTX 680 thanks to its 1188MHz boost clock, which it manages to stick to the entire time.

Civilization V Synthetics
Comments Locked

414 Comments

View All Comments

  • Galidou - Saturday, May 12, 2012 - link

    Massive smackdown, lol it made AMD lower it'S prices by 50-60$, that's not massive.

    You know what's massive, my two radeon 6870 that I paid 130$ each and still compete with those 400$ cards... That's a massive smackdown because I bought them almost A FREAKING YEAR AGO. Take that abuser of the word ''massive'' :P
  • SlyNine - Sunday, May 13, 2012 - link

    Not sure that comments was meant for me. I agree with you.
  • CeriseCogburn - Sunday, May 13, 2012 - link

    I very recently ( a couple months ago) recommended a 6870 to a very close friend, I think the price was about $159 and it had maybe some rebate at the egg.
    He bought it and has been running it for his several sims, it was the best bang for the buck at the time in a level above his then current 3 or 4 running vid cards ( 4000+5000 series).

    So no doubt amd can have a deal worth purchasing, it's just not there at all in these 2 new generations. Not even a tiny bit.
  • Galidou - Sunday, May 13, 2012 - link

    And what's the reason for that, do you know? It's not because Nvidia IS SO FAR SUPERIOR ffs stop with that, it may be superior but it's not a civic vs a ferrari. AMD had the price/performance/die size superiority because they'Ve been doing(since radeon 4870) a shift in making the most heavy gpus in the world.

    Nvidia was focusing on a more computationnal approach and biggest gpu, strongest performance since... well since it's alive I think. Nvidia simply took AMD's way for this gen, forget about compute power a little and focus on smaller die for mazimum performance and low consumption while AMD made the move the other way, improved their compute power at the cost of die size and power consumption.

    That's what GCN is all about. it'S about the same thing that happened with GTX 2xx vs radeon 48xx cept that 4870 was more than TWICE smaller than GTX 2xx and almost as performant and sometimes even more. Not a mere 43% difference in die size, MORE THAN DOUBLE the size and still were so close in performance.

    And why am I doing this analysis and you didn't, because you were too much occupied at launching inflammatory disrespectful stuff, like lots of nvidia fanboys do. People don't deserve this in a ''computer part'' oriented discussion. You were at the same time saying people were stupid and disrespectful while you were doing the same, let's say even worse.

    Now get a neutral vision of things before commenting like you do, you wanna be a fanboy, at least do it in the light of respect. Did I call you any names now? Did I have to use the words ''stupid'', ''analphabet'', ''dumbfuck'', and so on... No because that was meaningless. Learn from the best of die like the rest.
  • CeriseCogburn - Sunday, May 13, 2012 - link

    Yes you're a smack talker name calling disrespectful jerk.

    Yes of course you did, and you're a red fanboy and angry about all my absolutely valid points.

    That wall of text you have there is a big zero of hate - some toward me, so playing some innocent jay card doesn't work either, and BTW you've now posted a curse word, something I have never done.

    So you're a garbage mouth, and very disrespectful.
  • Galidou - Monday, May 14, 2012 - link

    I never called you any names NEVER look at all my posts, I never called you even one name and you just did it again and again but I'm the one who lacks respect what's that:

    ''disrespectful jerk.''
    ''you're a garbage mouth''

    That word I posted was an example, it wasn't meant toward anyone it was an example.
  • CeriseCogburn - Thursday, May 31, 2012 - link

    Oh stop your lying, you attack and call names more than anyone, it's your whole text every time.
  • CeriseCogburn - Sunday, May 13, 2012 - link

    PS
    angry disser Galidou " And why am I doing this analysis and you didn't, "

    Why you are doing this analysis is because you are a pure amd troll and have to go back to the last generation to try to make some idiotic historical point because you have no current points and no current rebuttals.

    Furthermore you claim I am somehow unsound by not having made your crybaby last generation whining no real point but I have made the more pertinent point as nVidia beat amd in every single metric here.

    I guess for your "win" you need to squeal about respect as you do the same trash talk, then go back to a prior generation to try to whine about what exactly ?

    You having a neutral vision is one big fat joke.
    You're in with all the amd fanboys here telling lies and going back a generation and whining about that...

    It would be nice if your snide remarks could be directed at pertinent points I have made but so far you're incapable, and often just on a troll attack against me personally in at least two of your other posts.

    What were you spewing about up there anyway ? Power ? LOL
    Guess who loses that sonny.
  • Galidou - Monday, May 14, 2012 - link

    these are the words you currently used in about 4-5 posts of yours, I don'T want to make any point with you toward video cards related discussion because of that, these are the words you use, I just want you to realize what you do:

    ''pathetic amd fanboys
    massive ignorance
    disrespectful jerk
    smack talker
    angry disser
    you're so ignorant
    most ignorant and clueless fools
    Where exactly is the brain ?
    (roll eyes at the immense ignorance, again)''

    and then at the same time use a sentence like:

    ''Thank me when you grow up enough to realize rebutting lies and fibs by others is an adult and responsible''

    The attitude words you used up there are so responsible... but no, you could just make your point and end it with a ''.'' but you ahve to add those words attacking everyone in your way, but sorry I forgot, I'm the one who attacks you, you're the perfect one... I'm maybe a troll but I'm not mad like you are.

    But hey, your arguments are so right that you gotta use all this stuff in your argumentation and call everyone a mad hater. Who's the hater? I already like Nvidia I'm no AMD fanboy.
  • CeriseCogburn - Thursday, May 31, 2012 - link

    You're a total amd fanboy and a liar, and attacker, and name caller, and cursing rude jerk.
    I'm sure you enjoy playing an angel in between all of that.

Log in

Don't have an account? Sign up now