The author has no idea where yield improvement comes from. I think they like to believe in pie-,in-the-sky magic. Getting to high yield 10nm process Intel have no choice but to start at low yield and ship dice into processors to find out what works: they have to commit to a process flow right the way up to packaging. Otherwise it truly is a pipe dream. You can't run the fab line efficiency if you mix outputs, it wouldn't surprise me if the process chemistry and timing and temps are so different you can't run mix nm scales at all.
Intel have to make and sell! What does the author want them to do with a low yield 10nm process? Design chips without the gpu and package or, design the dies for the future 2+1 you state, burn the silicon which is waiting for yield to work and then use it for the you makes more sense.
Lenovo took a deal. The price and energy budget work for them. It's win:win.
Intel have to make and sell! What does the author want them to do with a low yield 10nm process? Design chips without the gpu and package or, design the dies for the future 2+1 you state, burn the silicon which is waiting for yield to work and then use it for the you makes more sense.
Lenovo took a deal. The price and energy budget work for them. It's win:win.