GPU memory vs. compute capacity at each precision level

Linear axes. Dense (non-sparse) TFLOPS.