Before the data comes from the setup engine to be processed in the pixel Shader engine, it will meet an "Ultra" threading dispatch processor. It’s a key part of the chip that will tell R520 data where to go, and the data will hear and obey. There is a choice of four quad pixel Shader core or texture address units. Texture addressing Unit will later send its processed data's to texture units and later if not stored in texture cache those data's will go to General Purpose Register Arrays and further. If it sounds complicated that’s because it is.The Radeon X1600 has three pixel shader cores each, processing four pixels which equals twelve pixels per clock. And the Radeon X1300 has one pixel shader core which processes four pixels.
Each of ATI's next-generation cards features the Ultra threading dispatch processor that will process data prior to sending them to the pipes.
Source: The Inquirer