Long-term result of CLU (liquid metal TIM)

[Homeles]

Isnt this evidence that you used too much paste?

Too much > too little.

 

[BonzaiDuck]

My temperatures are still significantly reduced. I'm able to run at 4.8GHz / 1.37v and stay under 100c IBT loaded. It'd be a cold day in hell when you can do that without delidding.

Somewhere within the last two days -- maybe at Tom's Hardware but not sure (or maybe it was here and not sure) -- I found a link to a paper someone had written -- someone like IDontCare but likely someone else. Maybe -- an IB overclocking guide at OverClock or some similar site.

It explains how you can measure the current leakage on an IB at different temperatures, using LN2. The current leakage drops exponentially with each increment of temperature decrease. So temperature is an essential factor, maybe more than voltage. The article offers advice about OC limits depending on the choice of cooling. Of course, the central observation is that 22nm chips now require a "quantum physics" understanding, since the technology is approaching the atomic or molecular level.

And of course, if temperature is key, IBT or LinX generate heat that otherwise wouldn't arise with ordinary gaming, daily use -- the occasional "turbo-mode" increase to the 4.5+Ghz setting.

If we (or I) want to play this game, though, it's obvious that the old Noctua D14 may be satisfactory as a cooling strategy, but not optimal. Water (or better) is the way to go.