Nvidia GeForce GTX 680
The simple fact right up front here is that Kepler improves upon Fermi in every regard, from performance and power consumption, to features and software. On the surface it may not seem significantly different; after all, the GeForce GTX 680 doesn’t look too dissimilar from its predecessor, and the horsepower isn’t double, contrary to many unreasonable expectations we’ve seen posted in the forums. Is Kepler evolutionary or revolutionary? In truth, it’s a bit of both.
Kepler certainly improves the horsepower from Fermi, with the GTX 680 sitting atop the performance charts in every instance, soundly beating the Radeon 7970. Depending on the game title, some differences were nominal, about 5%, but there were a few instances where the GTX 680 utterly thrashed the 7970 by up to 30%. The average was about 15-20% ahead of the flagship Radeon card, which is consistent with what we’ve seen in previous product launches between the two titan graphics companies. So in that sense, Kepler is evolutionary.
The power consumption is a distinct improvement, and the 28nm process is a further development in Nvidia’s roadmap. This is expectedly evolutionary, but the performance per watt is a significant advancement, and in that sense, it is far less conventional. Providing better framerates is one thing, but bringing more horsepower while sucking less juice is very impressive. It appears that power consumption is a hallmark of Kepler, and it would stand to reason this should translate across the upcoming models, including the mid-range, budget, and even mobile lineup for Nvidia.
The Adaptive V-Sync is certainly innovative and we’re a bit surprised this hasn’t been developed sooner. Meanwhile, FXAA is an excellent move forward with image quality improvements, which notably take less of a performance hit, and we’re fans of bang for the performance buck, so to speak. And FINALLY seeing multi-screen support on a single card is very welcomed, though long overdue. But where Kepler really creates a stir and is ground-breaking is with GPU Boost.
GPU Boost, in our opinion, is nothing short of revolutionary. The idea isn’t entirely new though, as Intel’s Sandy Bridge Turbo was the real innovator, but Nvidia appears to be standing on the shoulders of giants, as it were. Regardless, GPU Boost is entirely a new way of thinking as it continually monitors and dynamically adjusts the card’s processes to provide optimal performance. And not simply gaming performance; power consumption, temperatures, and noise levels are all monitored and adjusted steadily for peak efficiency.
Seeing the new Base Clock and Boost Clock is certainly like deja vu for overclocking enthusiasts, and the "free" turbo boost can really be seen as a value-added performance feature that is provided out of the box. In that sense, "overclocked" cards that no doubt will be carried by board partners will simply be a product of shifting the target clocks to offer greater overhead potential. And it still depends on the particular card’s ability to hit certain speeds. That said, the days of manual tweaking to find the best combination of stable clock speeds and voltages are largely now redundant with the advent of GPU Boost. Since it’s hardware-embedded, you can’t turn it off, like it or not. We suspect after the intial shock wears off and dust settles, it will be very much a "liked" thing.
So then there’s the little matter of price. What does all this horsepower and innovation cost? $499. And that is less than what we expected. Just icing on the cake for consumers, really, and it certainly puts the Radeon 7970 in a very difficult position. Expect AMD to drop prices in the near future in response to the Kepler launch. Not only does the GeForce GTX 680 run faster and more efficient, it’s pleasantly well-priced for a premium flagship graphics card. Gotta love that.
The Nvidia GTX 680 brings both evolutionary and revolutionary ideas and features to the market, providing an outstanding package to gamers. Fermi may have paved the way, but Kepler brings it home. And it’s shaping up to be a juggernaut.