Shaders/HYPERZ/AA

SMARTSHADER 2.1

Besides the optimizations made to the RADEON 9800 PRO core, ATI has also improved its pixel shading engine with a new F-buffer. The F-buffer works like a form of cache memory, storing pixels that require multiple passes rather than writing them out to the frame buffer each time. This feature in particular was meant to address the shortcomings of RADEON 9700’s instruction length. With RADEON 9700 limited to 64 instructions, some complex shader effects required the pixel shading engine to make multiple passes. While this produced lifelike images, performance is crippled in the process. The F-buffer eliminates some of the redundancy from the graphics pipeline, saving time and reducing memory bandwidth requirements.

ATI also likes to point out that the F-buffer allows them to support fragment shader programs of unlimited length.

HYPERZ III+

HYPERZ III+ builds on the HYPERZ III occlusion culling technology first introduced in the RADEON 9700 family last year. The key improvement is the enhanced Z-cache that has been optimized to work better with stencil buffers. This is important for the next generation games that like Doom III that will be using real-time shadow volumes extensively. Take a look at some of our E3 2002 Doom III screenshots for a few eye candy examples.

We ran a few tests with PowerVR’s FableMark to get an indication of the improvements:






In the tests we underclocked our RADEON 9800 PRO to RADEON 9700 PRO levels, as you can see we saw a performance improvement of up to 2%. Not a whole lot, but as scene complexity is increased so should performance.

SMOOTHVISION 2.1

Besides the aforementioned improvements, ATI also states that it has incorporated optimizations in the memory controller that improve efficiency. The end result is greater performance when anti-aliasing is enabled, especially in the higher AA modes. We ran a few quick benchmarks to test this as well:






Other than 1600x1200, where we noted a 2% performance improvement, we really don’t see any gains. You’ll need to go up to 6xAA to really get any benefits from the improvements implemented in SMOOTHVISION 2.1 and by then the added bandwidth demands may be too much for the RADEON 9800 PRO core to handle.