Low Power Caches

  • Thread starter Thread starter Patrick
  • Start date Start date
P

Patrick

Hi

I am just doing some experiments with low power caches, called drowsy
caches. I am using
Hotleakage to simulate the behaviour of the cache. I am simulating a
datacache with 1024 lines each holding
64 bytes. The data array of the cache has a power consumption of 0.025
W if all lines are active. If some lines are in
drowsy mode the power consumption is less, each line in drowsy mode
saving around 10e-08 W. Obviously I can detect this difference in my
simulation but I wonder if I could also detect such a change in
practice with real measurement devices? Are there devices that can
detect a difference of lets say 10 lines are in drowsy mode to all
lines are active or will there be too much noise?

Many thanks for usefull comments,
P
 
Patrick said:
I am just doing some experiments with low power caches, called drowsy
caches. I am using
Hotleakage to simulate the behaviour of the cache. I am simulating a
datacache with 1024 lines each holding
64 bytes. The data array of the cache has a power consumption of 0.025
W if all lines are active. If some lines are in
drowsy mode the power consumption is less, each line in drowsy mode
saving around 10e-08 W. Obviously I can detect this difference in my
simulation but I wonder if I could also detect such a change in
practice with real measurement devices? Are there devices that can
detect a difference of lets say 10 lines are in drowsy mode to all
lines are active or will there be too much noise?
Click to expand...

I think you want comp.arch
 
Back
Top