Intellisample Technology: NVIDIA Strikes Back

Virtually everything up to this point has been NVIDIA more or less following the DirectX 9 spec and delivering what we have already seen from ATI. Granted, the GeForce FX does offer some enhancements in both the pixel and vertex shader categories but neither of those are going to be tangible to end users anytime soon. What will make games like Doom3 run faster however are things like NVIDIA's Intellisample technology.

Intellisample is NVIDIA's all encompassing term for their new color compression engine, improved fast z-clear, dynamic gamma correction, adaptive trilinear and anisotropic filtering, and anti-aliasing.

The first and most interesting of all of the new features is NVIDIA's color compression engine; the color compression engine sits between the rendering pipeline and the memory controller and compresses everything that comes out of the rendering pipeline before it is sent to memory. The lossless compression algorithm can obtain up to a 4:1 compression ratio which explains where NVIDIA gets their 48GB/s of memory bandwidth from. Remember that the 500MHz DDR2 memory on the GeForce FX provides 16GB/s of bandwidth on the 128-bit memory bus, but multiply that by 4 (don't forget to subtract out the original 16GB/s of data) and you'll get the 48GB/s of memory bandwidth NVIDIA is claiming.

It turns out that the compression engine is most useful when enabling anti-aliasing, and the reasoning behind this is simple; with conventional rendering you're taking one sample per pixel, and the color of that sample is rarely perfectly identical to that of adjacent samples. However, when anti-aliasing is enabled the GPU is taking multiple samples per pixel of identical colors which obviously compresses perfectly. The only time you don't get perfect compression is when you are sampling around the edges of polygons (vs. the interior of polygons).

The end result of this compression engine is that anti-aliasing now becomes a very low cost operation, since very little memory bandwidth is wasted. Essentially the only memory bandwidth used is on the edges of polygons, which make up a much smaller percentage of a scene than everything else. This should sound quite familiar as the results are similar to what Matrox promised (and delivered) with their Fragment Anti-Aliasing technology - only anti-aliasing the edges of polygons - however the difference is that there are no compatibility problems with NVIDIA's approach as it is still conventional multisampled AA.

The compression engine is completely invisible to the rest of the architecture and the software running on the GeForce FX, which is key to its success. It is this technology that truly sets the GeForce FX apart from the Radeon 9700 Pro.

The compression engine and the high clock speed of the GeForce FX enabled NVIDIA to introduce to new anti-aliasing modes: 6XS under Direct3D, and 8X AA under both OpenGL and Direct3D. Because of the compression engine, performance with AA enabled should be excellent on the GeForce FX.

Switching gears a bit, the GeForce FX also has a new, more efficient algorithm for clearing the Z-buffer. Known as Fast Z-Clear to those familiar with ATI's HyperZ technology, the GeForce FX basically has an improved version of the Fast Z-Clear that was in the GeForce4. The GeForce FX can also perform a fast color clear of the frame buffer but that is much less useful than fast z-clear.

On the topic of texture filtering, NVIDIA finally offers a non-purists anisotropic (and trilinear) filtering algorithm. This driver selectable option will enable users to choose from an accurate filtering algorithm (similar to what the GeForce4 currently has) or a slightly less accurate but very high performance algorithm (similar to what the Radeon 9700 Pro uses). The end result is that the performance hit incurred when enabling anisotropic filtering or even trilinear filtering when in this "performance" mode is considerably less. NVIDIA claims that their anisotropic filtering algorithm is more precise than ATI's, so the GeForce FX's anisotropic filtering should look just as good if not better than the Radeon 9700 Pro's.

NVIDIA's got 8 pipes too FX Flow - Cooling the Beast


View All Comments

Log in

Don't have an account? Sign up now