Nvidia GeForce 6800 Tech Preview - PAGE 1

The GeForce 6800 Tech Brief(NV40)

On April 2nd, we were invited down to San Jose to Nvidia's Editor Day where the day started off with a keynote from Nvidia's Jen-Hsun Huang. Actually it was more of a rallying cry for Nvidia. Jen-Hsun defended Nvidia's position of choosing FP32 over FP24 but conceded the fact that implementation on the NV3x series might have been an issue. Another theme was features and having a lot of them- Nvidia's position with the NV40, or as we all know now, the GeForce 6800 Ultra, is that a lot of features is a good thing. Huang illustrated the classic chicken and egg problem: if the hardware support did not exist, then the development efforts will not be there. Therefore Nvidia is trying to encourage developers to use different features on the 6800 and beyond because in his vision of the gaming industry, game technology is going to diverge rather than converge. As with the NV3x series, Nvidia is going to offer a top to bottom solution that will encompass the value range all the way to the high-end enthusiast solution. The proliferation of new technology in this way allows all cards to run all games albeit at different speeds, resolutions and with different texture sizes. The advantage though is that all games should work on all cards as they are functionally equivalent. This is a semi serious problem at this time as the feature set from a GF4Mx varies drastically from a NV3x board and even more so than a NV4x. By June, Nvidia expects to have no less than 5 different variants of the NV4x family out. Nvidia's goals and future is all fine and dandy but let us get to the real meat of the issue- the GeForce 6800.

The GeForce 6800 Ultra

This time around the FX moniker has been dropped and we go back to the naming scheme of the old GeForce 2 series with the Ultra designation. Switching between architectures like from the GeForce 2 to GeForce 3 and the GeForce 4 to the GeForce Fx did not always result in speed increases. There were cases where the GF2 and GF4 were faster than their latter counterparts. This time around, Nvidia claims that this is the largest performance increase that they have had between architectures.

What does the 6800 Ultra bring to the table today?

- 0.13 Micron Architecture
- 16x1 / 32x0 Architecture
- 6 Vertex Units
- 222M Transistors
- Shader Model 3 Support
- DX9c Support
- Full fp32 throughout pipeline
- Non-Power of 2 Textures with Mip Mapping
- Displacement Mapping