Well, folks, the mysterious "it" is finally here. Today at the GeForce LAN and Nvidia Gaming Festival event in Shanghai, China, Nvidia pulled the curtains back on what should be the fastest graphics card the world has seen to date. The GeForce GTX 690 will feature dual GK104 chips, just like those used in the GeForce GTX 680. Two of them together on a single card will team up to deliver roughly twice the theoretical peak performance of a single GTX 680.
Most of the GTX 690’s dossier reads like the spec sheet for a pair of GTX 680s in SLI. The card has dual GK104 chips, based on the Kepler architecture, with all units enabled for 3072 total shader ALUs. Onboard is 4GB of memory, or 2GB dedicated to each GPU, running at 6 GT/s, the same speed as on the 680.
GPU clock frequencies are only slightly lower than a discrete GTX 680’s. Nvidia it expects the GTX 690 to perform very similarly to two GTX 680s in SLI, and the firm estimates the GTX 690 will be between 25% and 65% faster than the prior-generation dually, the GeForce GTX 590. Since the GK104 is a more power-efficient chip, Nvidia anticipates the GTX 690 will be 1.5-2X more power efficient than its predecessor, as well.
Perhaps even more impressive than the specifications, which aren’t far from what one might have expected, is the card’s design. Nvidia CEO Jen-Hsun Huang reputedly directed the GTX 690 design team to build the best graphics card possible, and the product does indeed appear to set some new standards for materials and workmanship in the sometimes-staid world of PC graphics cards. The attention to detail is obvious in the GTX 690’s cooling shroud, where the usual glossy plastics have been banished. The lighter-colored outer portions of the shell are fashioned from chromium-plated aluminum, and the inner portion surrounding the cooling fan is made of magnesium. On either side of the fan are clear polycarbonate windows revealing the nickel-plated heatsink fins beneath. We’re quite pleased to see a premium-class product from a major GPU maker adopt premium-class materials, which are a rarity even among aftermarket cards.
Nvidia says much attention has been paid to the GTX 690’s acoustics, as well. Not only is the fan a top-quality spinner, but the PCB beneath the cooler has been flattened to reduce turbulence via the use of extremely low-profile capacitors. The result is a dual-GPU offering that is purportedly quieter than most of the firm’s single-GPU solutions.
Of course, products made with premium materials don’t tend to come cheap, and the GTX 690 will be no exception.The list price will be double the GeForce GTX 680’s, or roughly a grand, give or take a few pennies. You’ll also need a PSU capable of feeding its dual 8-pin power inputs, necessary to match the card’s 300W max power rating. We’d expect many GTX 690 owners to be sporting multiple monitors, as well, to be fed by its array of three dual-link DVI ports, along with a single mini-DisplayPort output.
Nvidia has done something notable by nearly matching the speeds of two GTX 680s in SLI. Given the power and thermal constraints involved, we’re dubious about whether AMD can build a Tahiti-based solution to match. Although the Radeon HD 7970 performs quite similarly to the GTX 680, the Tahiti GPU is larger, draws more power, and has 50% more memory channels to feed. Then again, Tahiti’s larger memory capacity and additional bandwidth could prove to be an asset at very high resolutions, so perhaps only time will tell.
The GeForce GTX 690 should be available in limited quantities on May 3, with wider availability planned for May 7.