On: Amdahl's Law

[monocasa]

> BTW how do GPUs solve the concurrent bus access you mention, with many cores are writing pixels?

Batching and buffering with spatial locality so that you can stream out relatively large bursts of pixels.