Yeah, so a large part of my optimization is to move from memory to compute bound. But things like matrix inversion, for example, thanks to the hard work of libr

Yeah, so a large part of my optimization is to move from memory to compute bound. But things like matrix inversion, for example, thanks to the hard work of library authors can scale pretty well. This happens on the CPU because the data typically exceeds GPU memory. So, you can get out the box scaling for much of your code. The 3D rendering is taken care by the GPU.