What to do when performance declines because of the temperature?

hi,all
when the performance declines because of the temperature, we should lower the voltage or lower the frequency? why?
Thank you!

Re: performance problem

When temperature increases, performance degrades.. because the carrier mobility degradation leads to slower slew rates... and hence the slower the speed.
So when temperature increases, to keep circuit working(i.e. avoiding timing violations) you can slow down the frequency to avaiod the timing violations.... or you can increase the voltage(if permited and should be within tolerance) so that delays will be less and it will compensate the delay increase by tempaerature and hence it will work at freq. you targated for.
BTW, while doing timing clouser, once must make sure that circuit works and meets timing for all the corners(i.e. best corner and worst corner).
Best corenter : High Vdd, Lowest temperature,Best process
Worst Cornter : Low Vdd, highest temperature, Worst process.

Hope this will help you....

Re: performance problem

Hi viju,

With 65 and below it is other way round ...

Means

Temp decreases, slower device and so worst case is lower temp one...

performance problem

this is something interesting you have any doc or any ref for this bellow 65nm for less temp more delays or for worst corner use lowest temp.

Re: performance problem

You can Google it for more details .. I do not have much idea about why it happens .. But we do consider Temp -40 as worst case rather than 125 ...

Re: performance problem

Hi Jaydip,
I am aware of this fact.
At 65 ns, Vdd is quite low (0.9 V !?) and hence its threshold will be very low(0.3 !?). Now as temerature is increases Vt goes further low and hence high current will flow and conduction starts quickly.
This is not seen till 130 nm, becasue at that time Vdd and Vt quite higher than at 65 or 45 ns.

All, Please let me know your views as well....