Introduction

In part 1, I described how I?d fitted a cheap heatsink with a 120mm fan, and was raving about my 35C full load & 22.5C (2.4C delta) idle temps.

At the end of the piece, I noted that the "idle" and "load" temps were creeping up.
This time, I'm going to attempt to describe the problems in more detail, diagnose them, and explain what I'm going to do about them; part 3 will be whether or not they worked.


Testing

Graph 1: Idle temps




First up is idle temps: up to 5 hours they're stable, i.e. in sync with the ambient temperature, although not as good as the 2.4C delta I was getting immediately after installation. After 5 hours running though, they begin to creep up.


Graph 2: Full load temps





Next up, the full load (folding@home) temperatures. From 30 minutes (warmed up) to five hours running, the temps show a gradual increase.
From five hours (300 minutes) running an additional problem appears: the CPU temperature is going up _faster_ than the ambient temp.





Diagnosis

So, what's the cause of these problems? I won't know until I fix them, but it seems there are a few likely candidates.

The generic thermal grease I've been using is the prime suspect: the fact that the temperatures get worse the longer the PC is running, and are getting worse from day to day suggests that not only is the grease turning to mush while in use, it's also degrading rapidly.

The budget heatsink is also a likely cause: with a small(er) number of chunky fins, you've got to wonder about its heat dissipating properties.

Heat soak from the power regulating MOSFETs close to the CPU socket could also be an issue. These produce a fair amount of heat, with radiation into the case and conduction into the motherboard being the only ways of heat dissipation. After several hours running, it's quite possible than heat is "soaking" from the MOSFETS far enough into the motherboard to result in false readings from the thermistor under the CPU.







Plans

Moving on to what I'm planning to do about it, first up is a Coolermaster Aero 7+, which I'll be fitting sans blower & rheostat (I'll be using the 120mm fan & ducting I already have).




My reasons for choosing this particular cooler were:
* Unscientifically, intuition tells me copper is a better material for heatsinks;
* Twice as many fins as the existing Q-Tec sink means four times the surface area, which has got to help;
* The blower fan is rated at 10.9 to 20 cfm, so I'm thinking that 78cfm 120 will give it quite a bit more pep;
* It's an 80mm fan fitting, so I can use all the gubbins I bought for part one;
* The three knib mounting clip has got to be more secure than the Q-Tecs one;
* It wasn't too expensive (£20).

In hindsight, I should have searched around until I found a heatsink with all the properties I wanted, but that used motherboard mounting instead. It recently occurred to me that _all_ heatsinks that use a spring clip are flawed, as immediately over the area on the CPU that produces the heat, there's a 1/4" of copper (or aluminium), a steel clip, and _no_ fins


Better thermal grease is next up on the shopping list: I don't think you can go far wrong with Arctic Silver 5, unless you spread it all over the shop & cause a short circuit.





Something to aid MOSFET cooling is next up, so I've picked out some Coolermaster Copper RAM Chip Coolers: cut in half they should be ideal.

I estimate 1/3rd of the spent air from the CPU sink will blowing over them, which will be sucked out by the 92mm fan in the PSU immediately above & the 120mm fan on the back of the case immediately behind.

Although that air will be prewarmed, I'm hoping that with an efficient CPU cooler with a lot of air blowing through it, the air won't be more than a couple of degrees C above ambient temperature. Well, here's hoping: there's no room for another fan in there.





Last on the list is some Arctic Alumina 2-Part Thermal Adhesive to stick them on with. Not that I'll be using much of it, I plan to use the ZENNZZO method of using tiny dabs of adhesive, with AS5 doing the bulk of the interfacing.








It would have been so much easier if ...

... Instead of attempting to justify why I needed to fork out on a new heatsink, MOSFET sinks & quality thermal transfer materials, I'd just said "I'm getting a new heatsink, MOSFET sinks & quality thermal transfer materials! Woof!"

It would have made for a very short article though.




Thanks to:
Im_gumby (although his step by step method didn't work for me) & Monsignor Funkibut for help with the Excel graphs;
ZENNZZO for input on the possible causes of the heat problems

JE I have seen on several sites that the Coolermaster Aero 7 Lite cools better then the Coolermaster Aero 7+ and its cheaper :+\

Quote: Originally Posted by TYCOON JE I have seen on several sites that the Coolermaster Aero 7 Lite cools better then the Coolermaster Aero 7+ and its cheaper :+\

The 7+ has been in my case for a few weeks now, but thanks - I'll go warm up the time machine

Intresting

One of the first things I look at when getting a motherboard is too see if it has the 4 holes for bolting on a HSF..if they aren't there...Time to look elsewhere, because it probably isn't going to be a seriuos overclocking board with only "Clips" for the HSF.

So it all boiled down to cheap thermal paste and a HSF that was moving around?

EDIT: BTW..I need some BH-5 since your making a trip...

My guess would be the mating area between the heat sink and cpu core is changing as the hsf heats up. Any metal will expand when heated. The absolute best surface for a heat sink would be slightly concave, so once it heats up it would be flat. Easier said than done.

Add the mounting pressure on the corners of the hs where the fan bolts on. And the less than optimum pressure a tab mount provides. The aluminum could easily warp and bow until only a fraction of the sink is touching the core. A thicker TIM like ceramique might help, but you'll be better off with the copper cooler.

Quote: Originally Posted by $SOLID$ Necro ... So it all boiled down to cheap thermal paste and a HSF that was moving around? ...

Well, that was the theory ... in practice, the new HS & AS5 gave results so contrary to expectations, part 3 could be a long time in coming