Introduction
Back in June we regaled the HEXUS readership with a fantastic tale of pixel fragments, vertex units and MADD instruction issuing. A yarn was spun with whispers of single-board performance to shame then current SLI, power consumption less than the previous generation and, most shocking of all, real retail availibility on the day of launch. Hardware websites throughout the land told of their accounts with the mystical new graphics boards and the myth became legend. NVIDIA had built a graphics chip to lay waste to the NV40s and R480s of the world. Its name? G70.
Powering GeForce 7800 GTX, G70 sits pretty as the most powerful pixel pusher yet released to retail. However one SKU, especially one at such a lofty price, is never enough, so rumours have run riot recently about what NVIDIA would do with G70 to create something a bit cheaper with most of the performance intact. Today they show their hand with the resulting GeForce 7800 GT probably the worst kept secret in graphics.
If you think back to the GeForce 6800 GT and how it related to the Ultra, you'll know that all NVIDIA did was reduce the frequencies of the chip and memory, keeping everything else intact. The smart money therefore would have been for NVIDIA to do the same with the 7800 GT, reducing the GTX clocks to create the new SKU. That hasn't happened, though, NVIDIA configuring the G70 differently with 7800 GT as well as changing clocks.
Time to explain the differences, examine a pseudo-reference board and show you what performance is like relative to 7800 GTX and a couple of other current parts.