Strangely, Matrix Multiplications on GPUs Run Faster When Given "Predictable" Data! [short]

Great minds discuss flops per watt.

Read in full here:

https://www.thonking.ai/p/strangely-matrix-multiplications