Comment by bwat48

I think the limitation here is more with the process nodes then with the GPU architecture