phreak0ut
The Thread Killer >:)
At SIGGRAPH 2007 conference Microsoft announced the details of the new DirectX updates version 10.1. They pointed out that to ensure full support of DirectX 10.1 you need not only to install Service Pack 1 for Windows Vista, but you may also need to replace a graphics card. Contemporary graphics accelerators from Nvidia GeForce 8800 and AMD/ATI Radeon 2900 may not support all the new features added to Direct3D 10.1.
Next-Gen web-site made certain things clear thanks to their phone interview with Microsoft's Sam Glassenberg, who said: "DirectX 10.1 fully supports DirectX 10 hardware. No hardware support is being removed. It's strictly a superset. It's basically an update to DirectX 10 that extends the hardware functionality slightly."
According to Sam Glassenberg, DirectX 10.1 will be fully compatible with all graphics cards supporting DirectX 10. He told that the current updates are very similar to those performed for DirectX 9 back in the days. All the company wants to do now is to increase the API life cycle. This statement was addressed to majority of worried gamers who got the impression that Microsoft announced GeForce 8800 and Radeon 2900 based graphics cards may become useless after the new updates have been pushed. However, Sam confirmed that existing graphics cards may still not be able to use all the new features of DirectX 10.1. At the same time he stress that applications designed specifically for DirectX 10.1 are very unlikely to appear, because overall, the updates aren’t that critical.
So, although DirectX 10.1 will support current DirectX 10 graphics hardware, today's DirectX 10 hardware will not be able to support all of the features of DirectX 10.1, which includes incremental improvements to 3D rendering quality.
However, the gamers who have already acquired contemporary DirectX 10 graphics accelerators shouldn’t be too upset. Upon developers’ request, version 10.1 sets whatever was available in the previous releases as a standard. As for the innovations, among them are 32-bit floating-point operations (instead of 16-bit ones, used today by default) and obligatory support of 4x FSAA.
Source