Sign in with
Sign up | Sign in
Your question

A simple question

Last response: in CPUs
Share
June 16, 2007 4:24:48 AM

Does manufacturing a processor at a smaller process make it inherently faster at the same MHz rating? Sorry for the stupid question, but I'm not very technologically inclined and was curious about this. Someone told me it does and I thought I'd ask people that actually know what they are talking about. :?

More about : simple question

June 16, 2007 4:30:47 AM

No it doesn't. From what I have learned it mainly just lowers the power requirements and lowers the heat output from the processor.
June 16, 2007 4:42:43 AM

MAYBE

if the chip is smaller then each path is smaller - an electron takes a long time to move as compared to light. shortening the path decrease the time and speeds up the process of a calculation

in fact, they bounce off atoms as they travel - shortening the distance by shrinking the die increase the speed.

it also decrease heat as the electrons bounce off less atoms, the heat is a by product of collisions - like a pin ball machine. i took my 400 level physics classes in the early 80's so i could be wrong!

if you decrease heat the electrons flow even faster - i think?

important part is you can then send more electrons or increase the fsb

hope that helps!
Related resources
June 16, 2007 4:45:45 AM

No, manufacturing process has no effect on performance
June 16, 2007 4:46:43 AM

nlrs basically has it.

Beyond that, on the manufacturing side, besides R&D and cost to refab, they can put more chips on a single wafer (what they build the processor on) meaning each wafer become inherently worth more.

But on our end, it means it's easier to cool the chip, and it's easier on the electric bill, especially when overclocking.
June 16, 2007 4:51:16 AM

MAYBE

its far more complicated then simplu no - the answer is yes
June 16, 2007 5:59:40 AM

The only numbers I've seen say no, that's AMD's fault.

The shrinking of the die causes changes to be made to the chip in order for it to even function on the smaller scale. These changes seem to ineviteably lead to the processing power to remain the same.
June 16, 2007 6:05:30 AM

Quote:
MAYBE

if the chip is smaller then each path is smaller - an electron takes a long time to move as compared to light. shortening the path decrease the time and speeds up the process of a calculation

in fact, they bounce off atoms as they travel - shortening the distance by shrinking the die increase the speed.

it also decrease heat as the electrons bounce off less atoms, the heat is a by product of collisions - like a pin ball machine. i took my 400 level physics classes in the early 80's so i could be wrong!

if you decrease heat the electrons flow even faster - i think?

important part is you can then send more electrons or increase the fsb

hope that helps!


Electrons actually travel very very fast. You won't notice any more speed difference in the CPU die due to smaller distances than you will notice time compression while driving your car.
June 16, 2007 7:17:36 AM

No, the die shrink shouldn't have a direct effect on processing speed. Everything moves at the same frequency, it just does it more efficiently. Now, when a die shrink happens there could be an improvement in processing speed, but that would come from a redesign of the core, not the smaller transistors. When AMD did their last die shrink I believe they did a little tinkering with the core's design, but I think that ended up hurting the performance, didn't it? I can't remember clearly.

The end result is generally cooler, more efficient processors that can usually overclock better. The smaller cores also result in more cores fitting onto a wafer, meaning that they can make more processors for the same amount of money, resulting in a drop of prices as well. Everyone wins!
June 16, 2007 7:27:58 AM

Quote:
When AMD did their last die shrink I believe they did a little tinkering with the core's design, but I think that ended up hurting the performance, didn't it? I can't remember clearly.


I believe JumpingJack was saying that latencies to the memory increased ever so slightly, and this is what caused the performance hit.
June 16, 2007 7:54:23 AM

Yes, the cache latencies on Brisbanes were higher than those on Windsors, making them slower. On top of that, the Brisbanes didn't really overclock better than the Windsors either.
!