G80 chip will intro TurboCache era

THERE HAS BEEN a lot of weird expressions on faces of key players when first pictures of G80 or GeForce 8800GT/GTX, showing 12 memory chips instead of regular, computer-friendly numbers of eight or 16. However, Nvidia's latest chip is far from being a patchwork, since the company has greater plans in scale.

Basically, if we take a look at Nvidia's G71-neé-RSX chip, the physical memory interface is 128-bit wide with 128-bit TurboCache addressing of system XDR memory and this is the first case of Nvidia using TurboCache technique on a higher end part.

Graphzilla decided to go for a 512-bit memory interface including a physical 384 one. So, you can expect that the company will deliver a marketing speech about 768MB of system memory and several gigabytes of additional system memory. It only remains to be seen if the TurboCache will be switched on for the Quadros only, or will the GeForce 8800 get the same treatment.

However, the real goal for Nvidia with G80 chip is the Quadro series. This means no Folding@Home speedups, since the company wants to make money by selling GPGPU 4K-res special FX rendering beasts. Expect TurboCached Quadros with 768 of video memory - and a 1.5 GB is also planned, as we managed to learn in Frisco Bay- with addressing of all available system memory.

The company already rules the HD broadcast market and with this chip, it's pretty much poised to become the de facto standard in the special effects arena as well. Nvidia's Gelato is pretty much what ATI's Stream Computing is, only being focused on movie production alone (rendering).

ATI is already designing its boards for integration into 2U servers, and it is only a question of time when mainstream hardware vendors start to offer processing boxes featuring several nV/DAAMIT GPUs instead of CPUs. µ