Fuad from The Inquirer just can't stop writing about ATI's upcoming graphics cards. One of his new posts about the Radeon X1800 gives us some more insight about how the Shader engine actually works.
The Radeon X1800 uses eight vertex shaders and each of them can process two shader instructions per clock. Theoretically this will give you 10 billion instructions per second..
This vertex Shader 3.0 is upgraded to support Shader model 3.0 as ATI could not handle more than Shader model 2.0b before. It uses Dynamic flow control, it can execute 1024 instruction but due its dynamic flow control is practically unlimited to the number of the instructions and ATI has more temporary registers than before.
A 128 bit vector ALU processes the vertex data and the core features parallel 32-bit scalar ALU and after processing those data it is transferred via flow control to further processing.
Backface Cull, Clip, Perspective divide and Viewport transform will also be executed by the vertex shader engine prior to sending the data to the setup engine.