NVIDIA GeForce GTX 770 Review: The $400 Fight
by Ryan Smith on May 30, 2013 9:00 AM ESTCompute
Jumping into compute, we aren’t expecting too much here. Outside of DirectCompute GK104 is generally a poor compute GPU, and other than the clockspeed boost GTX 770 doesn’t have much going for it.
As always we'll start with our DirectCompute game example, 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. While DirectCompute is used in many games, this is one of the only games with a benchmark that can isolate the use of DirectCompute and its resulting performance.
Civilization V at least shows that NVIDIA’s DirectCompute performance is up to snuff in this case. Though as is the case with GTX 780, we’re reaching the limits of what this benchmark can do, due to just how fast modern cards have become.
Our next benchmark is LuxMark2.0, the official benchmark of SmallLuxGPU 2.0. SmallLuxGPU is an OpenCL accelerated ray tracer that is part of the larger LuxRender suite. Ray tracing has become a stronghold for GPUs in recent years as ray tracing maps well to GPU pipelines, allowing artists to render scenes much more quickly than with CPUs alone.
Moving on to a more general compute task, we get a reminder of how poor GK104 is here. GTX 770 can beat the slower GK104 products, and that’s it. Even GTX 570 is faster, never mind the massive lead that 7970GE holds.
Our 3rd benchmark set comes from CLBenchmark 1.1. CLBenchmark contains a number of subtests; we’re focusing on the most practical of them, the computer vision test and the fluid simulation test. The former being a useful proxy for computer imaging tasks where systems are required to parse images and identify features (e.g. humans), while fluid simulations are common in professional graphics work and games alike.
CLBenchmark paints GTX 770 in a better light than LuxMark, but not by a great deal. The gains over the GTX 680 are miniscule since these benchmarks aren’t memory bandwidth limited, and the gap between it and the 7970GE is nothing short of enormous.
Moving on, our 4th compute benchmark is FAHBench, the official Folding @ Home benchmark. Folding @ Home is the popular Stanford-backed research and distributed computing initiative that has work distributed to millions of volunteer computers over the internet, each of which is responsible for a tiny slice of a protein folding simulation. FAHBench can test both single precision and double precision floating point performance, with single precision being the most useful metric for most consumer cards due to their low double precision performance. Each precision has two modes, explicit and implicit, the difference being whether water atoms are included in the simulation, which adds quite a bit of work and overhead. This is another OpenCL test, as Folding @ Home has moved exclusively to OpenCL this year with FAHCore 17.
Recent core improvements in Folding @ Home continue to pay off for NVIDIA. In single precision the GTX 770 is just fast enough to hang with the 7970 vanilla, though the 7970GE is still over 10% faster. Double precision on the other hand is entirely in AMD’s favor thanks to GK104’s very poor FP64 performance.
Wrapping things up, our final compute benchmark is an in-house project developed by our very own Dr. Ian Cutress. SystemCompute is our first C++ AMP benchmark, utilizing Microsoft’s simple C++ extensions to allow the easy use of GPU computing in C++ programs. SystemCompute in turn is a collection of benchmarks for several different fundamental compute algorithms, as described in this previous article, with the final score represented in points. DirectCompute is the compute backend for C++ AMP on Windows, so this forms our other DirectCompute test.
Unlike our other compute benchmarks, System Compute is at least a little bit memory bandwidth sensitive, so GTX 770 pulls ahead of GTX 680 by 11%. Otherwise like every other compute benchmark, AMD’s cards fare far better here.
Facebook
Twitter
RSS
117 Comments
View All Comments
khanov - Friday, May 31, 2013 - link
*sigh*You failed again.
khanov - Friday, May 31, 2013 - link
Sorry dude, that wasn't aimed at you. Anand your comments system has a mind of its own.If I reply to xyz I sort of expect my reply to be below xyz's comment and not inserted randomly in to the comments list.
chizow - Thursday, May 30, 2013 - link
Once again, a year late, but still a nice card. The updated cooler and higher memory clocks are impressive, but the max Boost clock was achievable on "FTW" type binned GTX 680s in the past.I guess this is Nvidia's "Gigahertz Edition", basically an overclocked SKU to bring parity in the performance midrange market.
Homeles - Thursday, May 30, 2013 - link
How in the world is this card a year late? Nvidia was still winning at this time, one year ago. Now they have not one, not two, but three single GPU cards that are on parity or are faster than the 7970 GE. Nvidia is in a far better position than they were with their GTX 500 series.chizow - Thursday, May 30, 2013 - link
Full GK104 should've been GTX 670 and below from the outset, as Nvidia initially planned. That's why it's a year late, at this price point anyways.Also, AMD reached parity with Nvidia's GTX 680 last year with the 7970GE launch in June/July, which then distanced itself by 5-10% with the Never Settle Drivers in Sept/Oct last year.
Now that the GTX 770 has launched and is ~10% faster than the 680, it again, reaches parity with the 7970GE.
JPForums - Thursday, May 30, 2013 - link
I thought the 104/114 series was historically reserved for the x60, while the 100/110 series was meant for the x70/x80 chips. Thus this new highend GK104 model should have been a 760Ti. GK110 should have maxed out at the 780 and the 770 should have been the paired down model. If they really had to have a Titan, it should have been a DPFP uncapped 780 (so they got that almost right).Of course the prices should have been the usual highend price points and not the massive price jumps they are currently pushing. Sure you can justify the price with the current performance relative to the previous generation, but if we always did that, the high end cards would get perpetually more expensive as the performance of each new generation of cards would justify a price hike over the previous generation. In reality, these prices are the unfortunate result of a lack of competition. Of course not all companies handle lack of competition the same way. nVidia has shown that, when uncontested, they will jack introductory prices into the stratosphere (8800 Ultra - $800-1000, $650 - GTX280, Titan/GTX780 - $1000/$650). Under normal competitive conditions, the top single GPU card from either nVidia or AMD/ATi of each generation comes in at $500. In similarly uncontested situations AMD/ATi has proven to be much less abusive to their customers (7970 - $550, 5870 - $400). Granted the relatively low price of the Dual GPU GTX295 probably kept the 5870s price in check until the GTX400 series launched, but at that point there was a significant difference in stability between single and dual GPU cards. Now I must mention, lest anyone gets the wrong idea, that AMD/ATi was probably only taking this route because marketshare/mindshare was more important to them than profit margins. Nonetheless, the facts remain.
chizow - Thursday, May 30, 2013 - link
I agree with virtually everything you said, although I never really had a problem with Nvidia jumping GK104 up a SKU to the x70 range. The performance was certainly there especially relative to last-gen performance and full GK104 also beat AMD's best offering at the time.The problem I had was Nvidia's decision to turn this 2nd tier ASIC into their flagship and subsequently, hold off on launching their true flagship ASIC a full year AND charge $1000 (and later, $650) for it.
All events predicated on the fact AMD launched 7970 at flagship prices when it really didn't deserve the asking price. Tahiti launch set the stage for Nvidia to not only undercut AMD pricing but to beat them in performance as well with only their 2nd tier chip.
JPForums - Thursday, May 30, 2013 - link
True, the 7970 could definitely be considered overpriced when it launched, but it was the undisputed performance champ until nVidia finally launched the GTX680 to bring back competition. Though, this begs the question, was the 7970 really this underperforming, or was the GK104 simply larger and faster (relatively speaking) than midrange chips in the past. Given that the GK104 die size is smaller than the GTS250, GTX460, GTX555 die sizes, I'd say larger is out. That said, they removed a lot of compute resources to get the gaming performance they were targeting, so faster might hold some weight.The 7000 series sudden proficiency in compute combined with the equally sudden removal of compute focus in the GTX600 series meant the 7970 would need to be far larger to maintain equivalent performance. Given the fact that Tahiti XT (352mm) was much closer to the size of GK104 (294mm) than GK110 (561mm), the 7970 should probably be considered a mid-weight. That is to say I can conclude that Tahiti XT was under performing (in games) AND GK104 was an overachiever. So the question becomes, is compute capabilities important enough to sacrifice gaming performance that a year ago likely would have clocked in closer to the GTX780 (GTX775 class?) for compute performance that in many cases exceeds Titan, but gaming performance roughly on par with a GTX680?
JlHADJOE - Friday, May 31, 2013 - link
IMO AMD's initial, higher price on the 7970 was justified. People forget that it was a much bigger chip than the 6970, with a 384-bit bus instead of 256. Any 384-bit part is effectively big, IMO. Same size as the 580, and now the Titan and 780.The fault here IMO goes right back to AMD's marketing division. If they hadn't stupidly went from 5870 to 6970, then people might have noticed that Tahiti was in fact a bigger part than its two immediate predecessors, and properly deserving of the 7900-series naming.
EJS1980 - Thursday, May 30, 2013 - link
Pretty much this /I\I
I