Comment by ShabbosGoy

Wouldn’t the memtransistor be faster though? Even if you use CUDA, you still have to compile down to assembly and probably use some kernel level module to communicate with the GPU.