TIM pumpout, thermal cycling, WC & HLT

[RonnieG]

The ff conditions seem to exacerbate the TIM pumpout problem:

-watercooling
-thermal cycling (alternating idle/full load usage pattern)
-HLT commands (bus disconnect enabled)
-overclocking/overvolting

Watercooling allows large and fast decreases in temperature when coupled with HLT instructions. A usage pattern that alternates between full idle and full load conditions will cause die temps to swing WIDE AND FAST, particularly with overclocked/overvolted CPUs.

This thermal cycling causes TIM pumpout, wherein the TIM gets ejected and displaced by air, which is of course not a good thing.

Greases also creep, again worsening the situation.

How can this be prevented?

[Top Hat Theater]

Fold, 24/7 to keep the system at full load :D

Sorry, I have nothing constructive to contribute...just my "witty" remark. However, I do have a question...how much ejection/displacement are we talking about? I don't think it's happened to me enough to cause a change in my temps.

~THT

[Since87]

Originally posted by RonnieG
Watercooling allows large and fast decreases in temperature when coupled with HLT instructions. A usage pattern that alternates between full idle and full load conditions will cause die temps to swing WIDE AND FAST, particularly with overclocked/overvolted CPUs.


Temperature variation between powered off, idle, and load are less with watercooling than with aircooling.

My motherboard reports a 2 to 3 C difference between idle and Prime95 load with my watercooled CPU.

[PCphreak]

This is really only an issue with an actively cooled system (below ambient), or with a poorly cooled sytem. In either senario, there is significant thermal delta between rest (off) and active states (on).

Per AMD's System's Integrators specs' (for instance), AMD specifies using a phase-change TIM on a air cooled heatsink- not a grease type of TIM. This is because grease has a greater potential of being pumped out, or shifting over time.

This is all I really have to add to the discussion. But RonnieG asked: "How can this be prevented?".

I believe you cannot totally prevent TIM creep, but you can control it by limiting the rate of thermal cycling.

-PC

[RonnieG]

Temperature variation between powered off, idle, and load are less with watercooling than with aircooling.

My motherboard reports a 2 to 3 C difference between idle and Prime95 load with my watercooled CPU.



Assuming a 100W t-bred with copper maze-type waterblock temps are going to swing a LOT more than that, probably around 20*C above ambient. Heat from the tiny t-bred die will be tough to spread.

Watercooling can get die temps back to ambient FASTER than a heatsink.

At full idle temps are practically at ambient, so we'll have temps that QUICKLY go up and down 20*C... not good...

TIM pumpout can be enough to be noticeable after a few months under these conditions.

I see some solutions:

1) Curable rubber TIM. AMD used some kind of silicone rubber TIM for bonding the K6 metal cap to the tiny die. Once cured there is no possibility of pumpout, but performance may not be as good compared to grease. I believe intel uses something similar for bonding their IHS.

2) Application a sufficiently generous amount of grease such that there s no possibility of air getting into the joint. The grease "seals" the area around the joint. TIM still gets pumped out but gets sucked back in when the assembly contracts.





...unless I'm very much mistaken