cooling experiment

[hollo]

this was an idea related to dealing with the heat that excapes from the CPU die into the package and CPU socket rather than into the heatsink, which i've found from past experiments is quite a lot

it didn't do anything spectacular, took off about a degree or two of heat, but i thought i'd post it for interest
my original install was arctic silver 5 between the CPU and a loose Integrated Heat Spreader, and the same between the IHS and heatsink (a thermaltake silentpipe tower)

first i took apart the CPU, and set it up with a little blob of arctic silver under the IHS and on the CPU-die-
http://img263.imageshack.us/img263/5959/198.th.jpg (http://img263.imageshack.us/my.php?image=198.jpg)

then dunked a big blob of cheap silicon heat transfer compound on the IHS. it's a bit runnier than the AS5, so i'm hoping the two blobs of AS5 made contact when it was reassembled. i used the silicon stuff for the big blob because of concerns about AS5 shorting something out, and because it's really cheap ($5 NZ, about $3 USD)-
http://img263.imageshack.us/img263/4803/199.th.jpg (http://img263.imageshack.us/my.php?image=199.jpg)

put the two parts back together hoping to see silicon heat transfer compound squeezed out the edges after it was compressed-
http://img263.imageshack.us/img263/6132/200.th.jpg (http://img263.imageshack.us/my.php?image=200.jpg)

IHS pushed down with silicon compound squeezed out the sides-
http://img263.imageshack.us/img263/4601/202.th.jpg (http://img263.imageshack.us/my.php?image=202.jpg)

i tidied up the excess silicon compound, and reassembled it with AS5 for the heatsink as per usual (the heatsink and top of the IHS were cleaned with isopropyl alcohol, so the install conditions were the same, or similar, between this and my previous install) for stress testing and temperature monitoring-
http://img263.imageshack.us/img263/5049/201.th.jpg (http://img263.imageshack.us/my.php?image=201.jpg)

results? earlier this morning i ran orthos for about an hour at stock speed and voltage, and temperature was mostly 44, sometimes 45.
right now it's been running for about an hour and temperature is mostly 42, sometimes 43
measured with coretemp
my room's definitely a little warmer than it was a few hours ago when i did the first run, maybe a degree or two, i had a heater on briefly, but unfortunately i can't measure that. the arctic silver might cure over the next few days, improving the current setup a little further compared to the previous install that was a week old.

interesting :)
i was originally planning to do this with a foam shape (to hold the silicon heat-transfer stuff in) and a bare CPU-die, so the silicon compound would sit right against the heatsink, but i realised it'd be a lot easier to do this for the sake of experiment by just filling in the gap under the IHS (because cooling isn't critical on this stock-running internet browsing machine).

the power output of this single core 1.8ghz socket 939 CPU is very low at stock settings, the temperature is about 32-33 degrees idle, mid 40s load. i bet that if this was done by holding the silicon (or similar heat transfer compound) directly between the CPU-package and heatsink with a foam shape on a hot overclocked dual-core IHS-free setup this could be pretty worthwhile (this computer is retired as a performance machine, it normally runs undervolted at stock speed, which is why the thermaltake silentpipe tower has no fans mounted normally to blow across the fins, just the blue fan giving a little breeze aimed towards the power mosftets (it's designed to run passive, so it works fine)).

[cegras]

Interesting experiment, it's always nice to see new literature coming in for peer review =P

The one thing I'd like to point out that if you use too much goop to fill in the space, at some critical thickness it'll start acting more as an insulator than a conductor of heat.

Another thing that's weird is that you put the huge silicon blob right over where the die would contact the IHS. Wouldn't that form a core -> as5 -> silicon grease -> IHS layer?

[ben333]

why not just shorten the legs on the socket mounting gear and run it with out the lid and a small ammount of paste? I did that and it worked quite well.

[hollo]

cegras: i was planning to simply use excess and let it be squeezed out, mostly straight away but also gradually over time because the stuff is pretty viscous, which i think is the best way of getting full coverage. that's my normal TIM application method - a dollop in the centre that is spread over the contact area by the pressure of the heatsink, with the plan that there'll be a little excess but that the TIM material should end up very thin given time and a nice tight heat-sink mounting method.

and for the big silicon blob, putting a huge blob in the centre and squeezing it out seemed to be the best way to get the entire under-IHS area filled with thermal stuff without air-bubbles. i thought about how i could do that and still get AS5 between the CPU-die and IHS but couldn't think of anything better than what i did - the hoped for result was that the runnier silicon-based stuff would squeeze out first, and the two AS5 blobs would end up forming most of whatever thermal transfer stuff remained between the CPU and IHS after the pressure was applied

ben33: i used to run my CPU IHS-less, on another motherboard, and liked it a lot. but i couldn't be bothered doing the necessary modification to the s939 CPU-retaining mechanism needed for the heat-sink to sit flat on the CPU-die afterwards for this current setup, because this computer is a browsing machine that doesn't need to be overclocked. it was something of a theoretical exercise. i'd definitely use both methods (this and running without the IHS) on a future performance (overclocking) setup

[ben333]

btw a good method of modding the 939 socket HS mount is to set it on a flat piece of metal and heat it up and let it melt even.

[hollo]

hey nice tip, it took me ages to scrape the plastic down to size last time :D