Maker Pro
Maker Pro

unbiased transistor leakage analysis

F

Fred Chen

Jan 1, 1970
0
Does anyone know of where one can get unbiased (i.e., objective)
transistor leakage power figures for the various CPU processors? I
have checked various benchmarking web sites, but sometimes it is hard
to get the feel that it is not subjective.

Also is there any detailed analysis available of the electric field
associated with the leakage current?

Much appreciated,
Fred
 
R

Robert Baer

Jan 1, 1970
0
Fred said:
Does anyone know of where one can get unbiased (i.e., objective)
transistor leakage power figures for the various CPU processors? I
have checked various benchmarking web sites, but sometimes it is hard
to get the feel that it is not subjective.

Also is there any detailed analysis available of the electric field
associated with the leakage current?

Much appreciated,
Fred

Transistor leakage is measured in microamps, not in watts.
CPUs have millions of transistors, in various states depending on
operations (if any) being done.
CPU power requirements can vary widely, depending on the program or
instruction mix.
One "benchmark" can be "tuned" tomake a CPU look better than a
(horrors!) competitors.
And benchmarks are rarely representative of actual useage by "the
average user", what ever "average" means...
 
F

Fred Chen

Jan 1, 1970
0
Robert Baer said:
Transistor leakage is measured in microamps, not in watts.
CPUs have millions of transistors, in various states depending on
operations (if any) being done.
CPU power requirements can vary widely, depending on the program or
instruction mix.
One "benchmark" can be "tuned" tomake a CPU look better than a
(horrors!) competitors.
And benchmarks are rarely representative of actual useage by "the
average user", what ever "average" means...

Yeah, I was afraid of that.

I have read that tunneling thru the gate dielectric is now the
dominant factor in the leakage current. So I wanted to find out more
about the electric field associated with this current.
 
R

Robert Baer

Jan 1, 1970
0
Fred said:
Yeah, I was afraid of that.

I have read that tunneling thru the gate dielectric is now the
dominant factor in the leakage current. So I wanted to find out more
about the electric field associated with this current.

AFAIK, that tunneling is most dominate in the (latest) dinky
geometries, and we may not see CPUs based on the latest and "greatest"
dinky geometries for a few years.
Maybe we will be lucky in that the manufacturers will learn that the
resulting excessive power drain cannot be tolerated in light of the
green movement (and laws) by various government organizations,
especially being compounded by their propensity to increase the number
of transistors by such large amounts...
....furthermore, the equipment costs seem to go up by the cube (or more?)
of the inverse of linewidth (in the billions now).
 
F

Fred Chen

Jan 1, 1970
0
Robert Baer said:
AFAIK, that tunneling is most dominate in the (latest) dinky
geometries, and we may not see CPUs based on the latest and "greatest"
dinky geometries for a few years.
Maybe we will be lucky in that the manufacturers will learn that the
resulting excessive power drain cannot be tolerated in light of the
green movement (and laws) by various government organizations,
especially being compounded by their propensity to increase the number
of transistors by such large amounts...
...furthermore, the equipment costs seem to go up by the cube (or more?)
of the inverse of linewidth (in the billions now).

With oxide thickness ~ 1 nm, tunneling has to be an issue. The
wave-function of the electron or hole should be important at this
scale.

I don't have any "real" numbers on the actual current magnitude, but
if leakage is something like 100 nA/transistor, and there are a 100
million transistors, operating at 1V, that is 10 W of power
dissipation right there.
 
Top