Over a year and a half – that’s how long the GeForce 8800 GTX remained in the position of what could be called Nvidia’s high-end GPU. Oh of course, six months after its release and – just a coincidence – just before the arrival of the R600, we did get an 8800 Ultra with slightly higher clock speeds, but it was nothing revolutionary. Then two and a half months ago, the arrival of the 9800GTX awakened hopes of substantial performance increases, but in the end the card offered only limited gains over the good old GTX, and was behind the Ultra version. For owners of these cards to be really happy with their investment, it was high time for Nvidia to offer more than a few extra megahertz or to rely on pairing two GPUs on the same card.
Finally Nvidia has heard our pleas: The GTX 280 is the first real reworking of the G8x architecture. And, yes, breaking with tradition, "GTX" is a prefix for this new architecture. Now we know the company’s modus operandi: Introduce a new architecture on a proven engraving process. Due to the often very high number of transistors, the chip is expensive to produce and the cards that use it remain expensive, but it stakes out the territory. Then during the ensuing years, Nvidia develops its architecture on all segments of the scale, using finer engraving, but it is less optimized for high clock speeds. Finally, when the new process is under control, Nvidia moves it into the high end, which then becomes much more affordable. We saw it with the G70/G71 and the G80/G92, and now history repeats itself with the GT200 – a true killer with 1.4 billion transistors engraved at 0.65 µm.
See our video of the GTX 280 (running) and the GTX 260 (shown).