NVIDIA GeForce GTX 680 Review: Retaking The Performance Crown
by Ryan Smith on March 22, 2012 9:00 AM ESTFinal Words
Once again we have reached the end of another GPU launch article and once again we have a lot of data to digest, so let’s get to it.
For the last few generations AMD has always put up a good fight and always managed to spoil NVIDIA in some manner, be it by beating NVIDIA to market by months like we saw with the 5000 series, or significantly undercutting NVIDIA and forcing them into a bloody price war as we saw with the 4000 series. This time AMD once again spoiled NVIDIA by releasing the Radeon HD 7970 nearly 3 months early, but as always, at the end of the day it’s NVIDIA who once again takes the performance crown for the highest performing single-GPU card.
What makes this launch particularly interesting if not amusing though is how we’ve ended up here. Since Cypress and Fermi NVIDIA and AMD have effectively swapped positions. It’s now AMD who has produced a higher TDP video card that is strong in both compute and gaming, while NVIDIA has produced the lower TDP and weaker compute part that is similar to the Radeon HD 5870 right down to the display outputs. In some sense it’s a reaction by both companies to what they think the other did well in the last generation, but it’s also evidence of the fact that AMD and NVIDIA’s architectures are slowly becoming more similar.
In any case, this has ended up being a launch not quite like any other. With GTX 280, GTX 480, and GTX 580 we discussed how thanks to NVIDIA’s big die strategy they had superior performance, but also higher power consumption and a higher cost. To that extent this is a very different launch – the GTX 680 is faster, less power hungry, and quieter than the Radeon HD 7970. NVIDIA has landed the technical trifecta, and to top it off they’ve priced it comfortably below the competition.
Looking at the bigger picture, I think ultimately we still haven’t moved very far on the price/performance curve compared to where we’ve gone in past generations, and on that basis this is one of the smaller generational jumps we've seen for a GTX x80 product, or for that matter one of the smaller leads NVIDIA has had over AMD's top card. But even with NVIDIA’s conservative pricing we’re finally seeing 28nm translate into more performance for less, which of course is really what we're interested in. To that end, based on GK104’s die size I’m left wondering where GTX 680 is going to be sitting by the end of the year as 28nm production improves, as there’s clearly a lot of potential for price cuts in the future.
But in the meantime, in the here and now, this is by far the easiest recommendation we’ve been able to make for an NVIDIA flagship video card. NVIDIA’s drive for efficiency has paid off handsomely, and as a result they have once again captured the single-GPU performance crown.
Facebook
Twitter
RSS
404 Comments
View All Comments
Jamahl - Thursday, March 22, 2012 - link
And you were all too willing to do so without evening-up the initial crime. Don't insult our intelligence Ryan.SlyNine - Thursday, March 22, 2012 - link
You need to open your mind alittle bit. It's easy to see what would happen in Anand did something that actually limited out of the box performance.Why are you even suggesting they do such a thing. This is how the card ships, and thats how you will be getting it.
Maybe they should lower the memory clock on AMD cards to make it fair. Or wait, their are different number of shaders. Maybe Anand should somehow limit that.
It just doesn't make any sense.
MattM_Super - Friday, March 23, 2012 - link
Yeah you can't please all the people even some of the time when it comes to GPU reviews. This seems like a through enough review of the card as it comes out of the box. Overclocking is also important, but considering the hassle, increase in temps and noise, and possible voiding of the warranty, it seems unreasonable to demand that the OC scores be treated as more important than the stock scores.Scott314159 - Thursday, March 22, 2012 - link
Any chance of running FAH on the 680... it will only take a few minutes and would give us folders a view into its relative performance compared to the outgoing 580 (and the Radeons).I'm looking to buy a new card in the short term and FAH performance is a factor.
Thanks in advance!
Ryan Smith - Thursday, March 22, 2012 - link
Tried it. It wouldn't run.cudanator - Thursday, March 22, 2012 - link
C'mon guys, why isn't there a single CUDA-Test? And don't say "cause AMD doesn't support it" :P For me most interesting would be the CUDA-Speed compared to other nVidia-Models.Wreckage - Thursday, March 22, 2012 - link
Not to mention PhysX. Sadly there are a lot of features AMD does not support and so they don't get benchmarked often enough. h.264 encoding is another one.CeriseCogburn - Thursday, March 22, 2012 - link
On other sites they turn on full PhysX in Batman on the 680 and keep it off on the 7970, and the 680 still wins.LOL
If you watched the release video they show PhysX now has dynamic on the fly unique in game destruction that is not repeatable - things break apart with unique shatters and cracks. I say "it's a bout time!" to that.
My 7970 needs to go fast, in facts it's almost gone as we speak. No way am I taking the shortbus.
SlyNine - Thursday, March 22, 2012 - link
Umm, the GNC (7xxxHD) has a fixed function H.264 encoder. Afaik the 680GTX doesn't even have a fixed function h.264 encoder. So I'm pretty sure it would mop the floor with cuda H.264 encoding.Ryan Smith - Thursday, March 22, 2012 - link
We have the data, although it's not exactly a great test (a lot of CUDA applications have no idea what to do with Kepler right now). It will be up later today.