It seems hard to get a nice flat HSF base nowadays. I wanted to show you the difference between a flat and an untouched HSF.
The first time I did this I dropped 6C off my CPU temp but didn’t document it very well. This time around, I kept careful notes to demonstrate the general concept and benefits of doing this, not to try and get the lowest values I could.
The test bed:
Abit KT7A mobo
T-Bird 1.1ghz 200FSB w/retail HSF (I used thermal grease, not the pad)
Western Digital 20gb ATA 100 HD
ATX full tower 300 watt
This is a stock computer with no mods or additional fans.
I got my temperature readings by running protein folding for 10 minutes, taking the load temp, then cooling for 5 minutes and taking an idle temp.
Here are the temperature reading I got before flattening:
10 Minutes Loaded: 60C
5 Minutes Idle: 33C
Here’s a couple of shots of the HSF as it came out of the box:
Nothing special here.
The base did have quite a rough surface, which seems apparently intentional in comparison to the (somewhat) machined aluminum that can be seen along the left side and bottom. (below)
Even though I handle the network at my place of employment, I seem to find myself in the machine shop a lot. I like watching whatever projects are going on and machining has always interested me. So here we have our HSF mounted in the mill, getting ready to fly cut.
On the last HSF I cut, we started in .001″ increments and I had a perfectly flat base by .005″
This one was a lot worse off, though, so we went in .003″ increments. Below is how it looked after the first run.
Pretty scary isn’t it? Here you have a HSF to mount to a core no bigger than your fingernail and this is the kind of surface they are giving you to do it with.
We dropped it down another couple thousandths and ran it back the other way (below).
You can see we didn’t even let it go all the way across this time. It was so far off, I just backed it off and dropped it another .003″.
The dark spots in the freshly cut areas are wet from cutting fluid (if that’s hard to tell by the pictures)
I’m not going to bother you with pictures of every single cut that was done. Overall, this poorly shaped beast required .011″ to make true!
Below are two shots of the finished piece. It was still wet and had particles on it, but it sure looks nice doesn’t it?
Quite a difference from the first picture at the top. These pictures don’t do it justice.
I washed it off and blew it out real well to make sure I had no cutting residue or metal flakes on it.
Ok, it looks nice, but looks aren’t everything! What did it do?
I remounted it back into the machine with thermal grease (I didn’t have anything better) and re-did my tests exactly as before.
Here are the new numbers vs. the old ones.
10 Minutes Loaded: 58C vs. 60C
5 Minutes Idle: 30C vs. 33C
System: 34C vs. 35C
Unfortunately, it’s hard to say what gains one will get from this modification. I had a 6c difference on the last one I did. The next one I do might have a minimal difference, who knows?
Since all HSF aren’t created equal, you can’t really put an exact number on it. I think you can agree though, tis better to do, than not to do!
I’d like to thank Randy Moya. He’s the guy that runs the machine shop and for giving me some of his time to do these tasks. I had fun doing this little article. Hope you enjoyed it!