NVIDIA GeForce GTX 780 Review: The New High End
by Ryan Smith on May 23, 2013 9:00 AM ESTSynthetics
As always we’ll also take a quick look at synthetic performance, though as GTX 780 is just another GK110 card, there shouldn't be any surprises here. These tests are mostly for comparing cards from within a manufacturer, as opposed to directly comparing AMD and NVIDIA cards. We’ll start with 3DMark Vantage’s Pixel Fill test.
Pixel fill is traditionally bound by ROP and memory throughput, but with enough of both the bottleneck can shift back to the shader blocks. In this case that’s exactly what happens, with the GTX 780 trailing GTX Titan by about the theoretical difference between the two cards. On the other hand it’s very odd to see the GTX 680 get so close to the GTX 780 in this test, given the fact that the latter is more powerful in virtually every way possible.
Moving on, we have our 3DMark Vantage texture fillrate test, which does for texels and texture mapping units what the previous test does for ROPs.
Unlike pixel fill, texel fill is right where we expected it to come in compared to cards both above and below the GTX 690.
Finally we’ll take a quick look at tessellation performance with TessMark.
NVIDIA’s tessellation performance is strongly coupled to their SMX count, so the high number of SMXes (12) on the GTX 780 helps it keep well ahead of the pack. In fact we’re a bit surprised it didn’t fall behind GTX Titan by more than what we’re seeing. On the other hand the lead over the GTX 580 is right where we’d expect it to be, showcasing the roughly trebled geometry performance of GTX 780 over GTX 580.
Facebook
Twitter
RSS
155 Comments
View All Comments
mac2j - Thursday, May 23, 2013 - link
The problem with $650 vs $500 for this price point is this:I can get 2 x 7950s for <$600 - that's a setup that destroys a 780 for less money.
Even if you're single-GPU limited $250 is a lot of extra cash for a relative small amount of performance gain.
Ytterbium - Thursday, May 23, 2013 - link
I'm disappointed they decided to cut the compute to 1/24 vs 1/3 in Titan, AMD is much better value for compute tasks.BiffaZ - Friday, May 24, 2013 - link
Except much consumer (@home type) compute is SP not DP so it won't make much difference. SP performance is around equal or higher than AMD's in 780.Nighyal - Thursday, May 23, 2013 - link
I don't know if this is possible but it would be great to see a benchmark that showed power, noise and temperature at a standard work load. We can get an inferred idea of clock per watt performance but when you're measuring a whole system other factors come into play (you mentioned CPU loads scaling with increased GPU performance).My interest in this comes from living in a hot climate (Australia) where a computer can throw out a very noticeable amount of heat. The large majority of my usage is light gaming (LoL) but I occasionally play quite demanding single player titles which stretches the legs of my GPU. The amount of heat thrown out is directly proportional to power draw so to be able to clearly see how many less watts a system requires for a controlled work load would be a handy comparison for me.
TL:DR - Please also measure temperature, noise and power at a controlled workload to isolate clock per watt performance.
BiggieShady - Friday, May 24, 2013 - link
Kudos on the FCAT and the delta percentages metrics. So 32,2% for 7990 means that on average one frame is present 32,2% more time than the next. Still, it is only an average. Great extra info would be to show same metrics that averages only the deltas higher then the threshold delta, and display it on the graph with varying thresholds.flexy - Friday, May 24, 2013 - link
NV releases a card with a ridiculous price point of $1000. Then they castrate the exact same card and give it a new name, making it look like it's a "new card" and sell it cheaper than their way overpriced high end card. Which, of course, is a "big deal" (sarcasm) given the crazy price of Titan. So or so, I don't like what NV does, in the slightest.Many ages ago, people could buy *real* top of the line cards which always cost about $400-$500, today you pay $600 for "trash cards" which didn't make it into production for Titan due to sub-par chips. Nvidia:"Hey, let's just make-up a new card and sell those chips too, lols"
Please AMD, help us!!
bds71 - Friday, May 24, 2013 - link
for what it's worth, I would have like to have seen the 780 *truly* fill the gap between the 680 and titan by offering not only the gaming performance, but ALSO the compute performance - if they would have done a 1/6 or even 1/12!! to better fill the gap and round out the performance all around I would HAPPILY pay 650 for this card. as it is, I already have a 690, so I will simply get another for 4k gaming - but a comparison between 3x 780's and 2 690's (both very close to $2k) at 8Mpixels+ resolution would be extremely interesting. note: 3x 30" monitors could easily be configured for 4800x2560 resolution via NVidia surround or eyefinity - and I, for one, would love to see THAT review!!flexy - Friday, May 24, 2013 - link
Well compute performance is the other thing, along with their questionable GPU throttle aka "boost" (yeah right) technology. Paying premium for such a card and then weak compute performance in exchange compared to older gen cards or the AMD offerings... Seriously, there is a lot to not like about Kepler, at least from an enthusiast point of view. I hope that NV doesn't continue that route in the future with their cards becoming less attractive while prices go up.EJS1980 - Wednesday, May 29, 2013 - link
Cynical much?ChefJeff789 - Friday, May 24, 2013 - link
Glad to see the significant upgrade. I just hope that AMD forces the prices back down again soon. I hope the AMD release "at the end of the year" is closer to September than December. It'll be interesting to see how they stack up. BTW, I have shied away from AMD cards ever since I owned an X800 and had SERIOUS issues with the catalyst drivers (constant blue-screens, had to do a Windows clean-install to even get the card working for longer than a few minutes). I know this was a long time ago, and I've heard from numerous people that they're better now. Is this true?