This all went over my head, but does anyone know either how much faster this will make things (4x faster than AVX512 at 2048-bit??), and if unified memory plus a basic GPU will render this dead in the water?
It's impossible to say what sort of improvement (if anything) this will give.
My assumption is the determining factor in performance is going to ultimately be memory bandwidth.
> and if unified memory plus a basic GPU will render this dead in the water?
Yeah, generally I think that is the case. The main benefit this thing would have is power consumption will likely be lower than what it is to send this operation to the GPU.
My assumption is the determining factor in performance is going to ultimately be memory bandwidth.
> and if unified memory plus a basic GPU will render this dead in the water?
Yeah, generally I think that is the case. The main benefit this thing would have is power consumption will likely be lower than what it is to send this operation to the GPU.