NVIDIA GeForce GTX 670 Review Feat. EVGA: Bringing GK104 Down To $400
by Ryan Smith on May 10, 2012 9:00 AM ESTCompute
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.
Facebook
Twitter
RSS
414 Comments
View All Comments
Gastec - Tuesday, November 13, 2012 - link
Insanity!CeriseCogburn - Friday, May 11, 2012 - link
When amd has trouble with a game, the reviewer, completely unaware of any amd deficiency (as any good fanboy), has no explanation at the ready, nothing he has been watching for as an identified weakness - unlike the nVidia cards, where the reviewer is on constant watch for what he believes are nVidia card weaknesses." Skyrim is a game that for inexplicable reasons AMD just has some trouble with that NVIDIA doesn’t, possibly driver overhead."
Yes, we know, if amd doesn't do well, it's inexplicable. Something just has to be wrong. Somehow reality has warped.
Then after noting the 670 win fairly, we get this:
" At 1920 we’re clearly CPU limited even with all of Skyrim’s graphical features turned up. "
Instead of saying the 570 beats every amd card at that resolution, or even noting every amd card is stacked at the bottom, period, it's a "cpu problem" - amd didn't fail, the cpu did...
That's not all - after seeing the 570 spank every amd card at the 1920 resolution the reviewer goes on attack, since he mentioned of course earlier in the article nVidia brought up the 570, meaning it would be part of their line up he said, so attacking it is a must, as any good fanboy knows. Never mention is has spanked the 7870 and 7850 and even the 7970 once (civ5) and never mention it spanks the whole amd tier in this game with texture packs and it's "puny memory" at 1920, but go after it...
" GTX 670 is greatly improving on the GTX 570 due to the latter’s lack of memory. 1.25GB is cutting it close here with the high resolution texture pack "
There you have it. Once again. The GTX570 spanks every amd card at 1920 but all it gets is a big fat cut down for 2560.
Oh yes, the whole way it's done is consistently against nVidia.
CeriseCogburn - Friday, May 11, 2012 - link
Here again, an nVidia win is inexplicable. Instead we hear about the amd "strength".When nVidia wins, remember it's unclear why because it should not be happening, amd has a superior strength !
" At this point it’s not entirely clear why the GTX 600 series does so well here (both AMD and NV use SGSSAA), especially given the fact that the Radeons have a memory bandwidth advantage."
Now we can listen to the "on paper" amd fanboys, and the endless fantasy that future drivers and future games mean amd is the "future winner" for the newest cards tested and compared.
Never will we hear the amd core "technology" is weaker and has some severe caveats when it comes to game engines and implementation.
Instead, that with much superior "paper horsepower" on "notional multiplications" based upon less than fully accurate simplified calculations omitting severe bottlenecks in amd architecture and adding in guesses for the "tricks" amd has to "overcome" them, we will always be told the above, instead of the truth.
AMD core design is inferior for most of the game engines.
snakefist - Friday, May 11, 2012 - link
" At this point it’s not entirely clear why the GTX 600 series does so well here (both AMD and NV use SGSSAA), especially given the fact that the Radeons have a memory bandwidth advantage."oh, dear god - as much as i try to ignore you...
do you, or do you not remember NVIDIA cards with 192/384/512 memory bandwidth and how much good this "advantage" brought to them? compared to AMD 128/256 bandwidth of the same generation?
when commenting your post, one is necessarily becoming an AMD fan, since basically everything you say is so biased and (mainly) incorrect or misinterpretation of actual facts
i suspect you secretly work for AMD :)
CeriseCogburn - Friday, May 11, 2012 - link
Are you aware the post you copied is the reviewers words, smarty pants ?Do you realize I was criticizing the reviewers words, his "bandwidth" notional advantage for amd ?
Do you realize you just called the reviewer and idiot and agreed with me ?
YOUR words > " do you, or do you not remember NVIDIA cards with 192/384/512 memory bandwidth and how much good this "advantage" brought to them? compared to AMD 128/256 bandwidth of the same generation?"
Now tell it to the reviewer dumb dumb, since you copied and pasted HIS WORDS from my post, his words, words for word from the review !
ROFLMAO - yes you maybe should become a paid operative for amd, you have the intelligence for it - get everything wrong then attack - you're perfect for them :)
snakefist - Friday, May 11, 2012 - link
nope. i do not realize that. i don't memorize each word of review to be able to write 2/3 comments of 22-pages thread...and i didn't call anyone "an idiot".
what i said is what i believe (furthermore, it could be backed up by numerous reviews from that time - all of them also written by reviewers, perhaps on your surprise). whether this reviewer agrees with me or not on this topic is a simple matter of opinions, and review itself doesn't burst with hate as all your posts do.
CeriseCogburn - Friday, May 11, 2012 - link
Who knows what you said, it isn't clear, because you didn't make it clear.Believe whatever it is you believe,as that is also unclear, and only your opinion, since no facts are present for you, according to you.
Of course you only see hate from me, since I point out the amd flaws, and amd is your fanboy fave, as your other posts clearly show, and no you don't sense hate from the reviewer as he coddles amd, as I pointed out, and that gives you a nice warm feeling of "good opinion" vibrations.
ROFL
It's all too clear.
CeriseCogburn - Friday, May 11, 2012 - link
PS- you called me a lot worse than an idiot, so don't cop out on your hate filled spewing. Of course who expects anything less from an amd fanboy.Let's go with this, so you don't forget, or try to claim white snowy innocence - I certainly hope a bunch of obvious amd fanboys ignore every fact and facet I've presented, and dive, "driven" as you noted, right (back) into their beloved failure, with someone like you doing that, I couldn't receive any greater and lovely reward for my efforts here.
snakefist - Friday, May 11, 2012 - link
i did question your sanity, which i found more than ever a reasonable question. i'm surely not only one who shares this doubt. as for anger-management - i don't find this insulting, you should rather stop living in denial that you have onei did not called reviewer "an idiot", which you implied.
i was not aware that you are making efforts of any kind, except to comment every sensible post of any author with your irrelevant raging about something else... now that you explained it so well, i'll probably stay off your noble crusade of enlightening people with wildly subjective and/or incorrect claims... at least as long i feel like it
facts behind my claims are clearly stated and usually well-known and proven. your last two posts doesn't have single fact, btw
now reap the benefits of your great reward, whatever you might think it is
CeriseCogburn - Saturday, May 12, 2012 - link
Oh stop being a liar, you got outed so now you can't stand yourself.Let's go with near page one where one of your fellow amd fans haphazardly claims the 365mm die of the amd card is so close to the 300mm die of the nVidia card that amd price dropping again is no problem.
Another amd snake applauded the post for the sanity it contained. ( merely because amd fanboyism was served well, of course)
I had to point out, with considerable effort and for the greater good ...
(let's face it amd fans should be ponying up profit dollars for the hurting amd, not demanding low prices be lowered again - it's called putting your money where your mouth is, and where your heart is in this case, something the amd zealouts apparently have, to only their personal and selfish advantage, trained themselves out of - a rude, and debilitating issue for amd - who IMO has greatly encouraged that suicidal behavior )
...that the 365mm die is over 43% larger than the 300mm die.
Now there's just a single example of what needs to happen much more often, so that we can be well informed persons instead of lying brainwashed monkeys.
I certainly do not mind a big 'ol fanboy, in fact that's great, but let's bring it up to a level where a fan can be a self respecting and respected addition, not in need of constant lies, and endless unethical misconduct, right ?
A fanboy should easily support his view with the truth and not be in need of anything other, and as I fairly point out, an honorable fanboy won't be hellbent on squeezing every last dollar from their favorite producer while at the same time preaching corporate profitability "lessons" or sideway gloatings of the same, in direct opposition to their stated personal savings conduct goal and oft lofted talking point rhetoric.