What is the manufacturing process(32nm or 28nm) and what does a smaller manufacturing process do for the CPU or GPU
In chip manufacture the manufacturing process refers to the distance between transistors on an integrated circuit. As the manufacturing process gets smaller you can fit more transistors in a given space. Increasing density.
Dropping the manufacturing process, making it smaller, means the chip will usually perform better in terms of speed and will usually use less power. However as you go smaller and smaller heat output increases.
Right now Intel is on the 22nm process and is looking to go smaller to 14nm in 2014. However when you go this small the features are so tiny that the silicon it is printed on beings to behave strangely and often times falls apart and at 10nm, the next jump, quantum tunneling becomes a serious issue. The shrinking process has allowed for Moore's Law but now it seems we may have hit a wall at least for now.
Heatout put actually decreases. because the volatge need to jump that smaller nm gap has shrunk. this is why haswell is low voltage. More heat is seemingly produced (and it is) because there are more transistors in a given area.
I'm no phycisist But i believe silicon won't work past 6 or 5nm because of the heisenberg uncertenty principle.
Ahh thank you for that clarification. I knew that the smaller process used less voltage so I was wondering why there was more heat. Didn't even think of the density issue.
At 10nm tunneling becomes an issue and at 5nm I believe it becomes too great, like you said, to overcome.
I think you'd find this thread interesting: https://teksyndicate.com/users/rsilverblood/blog/2014/02/06/gpu-wars-enter-maxwell-nvidias-successor-kepler
I made a comment talking about why the 290 and 290x have heat issues compared to other 28nm chip
Quite an interesting thread. Soaking in the knowledge right now. Thanks :)
I thought lithography was the distance between the gate and drain of the transistors. After a certain point quantum phenomena will disrupt the electrical side of things by allowing electrons to be present at more than one place at the same time.
So as the manufacturing process becomes smaller there is less voltage required but more heat output because of increased density. That's why ivy bridge needs less power than sandy bridge, but why does Haswell need less power than ivy since it's based on same lithography as ivy bridge?
It is a different architecture with different power management features and throttling. It is 22nm like IvyBridge but not the same.
Intel has the Tick-Tock principal. Every microarchitecture change is followed by a die shrink. Every tick is a shrink and every tock is a new architecture. So SandyBridge (32nm tock) is the same architecture as IvyBridge (22nm tick). Haswell (22nm tock) to Broadwell (14nm shrink of Haswell tick). So on to Skylake and Canonlake. If we get there that is because of quantum interference getting in the way.
