It was the first system of a modern design in a lot of ways.

Six hardware threads, so the thread per subsystem model didn't really work. Instead you got the best results from a DAG of tasks submitted to a job stealing scheduler running on most of the CPU resources. But beside that, no little DSPs that you had to separate work out onto other than the GPU.

And it had the first consumer GPU to incorporate unified shaders (so vertex and pixel shaders ran on the same cores, dynamically scheduled to improve utilization).