Building a computer – another digression for thermal paste

I was doing the research for what I think the next ‘thing’ in the project is – yeah, I’m actually doing that, not just spouting – when I got sidetracked by thermal compounds.

Now, one of the Big Deals for most electronics is heat dissipation. Too hot, and what used to be Really Good is now a paperweight. (And probably not a good paperweight at that.) I started in on cooling devices (yet another post), but noticed a lot of assumptions. I’d also noticed the same when reviewing Xbox 360 repair and mod discussions. (digression – I’ve now tried a couple of the warranty-breaking xbox 360 RRod fixes. Where the error is appropriate they’re great fixes. They won’t fix all the problems. Offhand, I’m guessing between half and 3/4 of 3-light RRods can be fixed that way.)

Here’s the deal. In a perfect world the metal of the component mates perfectly with the metal of the heat-sink, regardless of how that sink then rids itself of the resulting thermal load. A perfect world, however, doesn’t happen. There are uneven spots – divots, stripes, warps, etc – which prevent a perfect match. Adding nothing whatsoever results in those gaps being filled with air. Air is a lousy thermal conductor – better than vacuum, but not by a lot. Anything is better than that, but we can beat “anything” by putting something designed for the purpose into place. Thus, thermal compound (aka thermal paste aka thermal interface material aka TIM).

The really interesting thing about TIM is that the industry as a whole has improved markedly over the past decade. In 1999, tests comparing ‘good’ stuff to ‘standard’ stuff would show a processor temperature difference of over 10 degrees celcius – that is, the ‘low’ would be around 35C and the high around 45C. Today that difference is halved. If you’re not pushing the extreme, then the quality isn’t enough to justify the difference in price. Except… that last sentence assumes you pay more for quality. Turns out that’s not necessarily true. In fact, some of the best are as cheap as the ‘standard’ stuff.

For those pushing the extremes, the ‘best’ – actually the best two – come from Shin Etsu. That’s the G751 and the X23 7783D. Both (as well as the middle of the pack X23 7762) list at around US$9.50 per gram. The third best – which definitely pushes the two leaders – is the modder forum old favorite, Arctic Silver 5, which lists around US$8.00 for 3.5 grams. If you’re only going to do one heat sink, once, there’s no real difference in price. If you’re going to do several the quantities start to matter. By the way, MG Chemical’s Non-Silicone Heat Transfer Compound lists at about US$9 for 60 grams. Yes, that’s close to 20 times as much paste for the price as AS 5. How does it do? In Hardware Enthusiast’s test (most recent of many I ex it was the definite trailer, letting the CPU get almost three degrees (Celsius) than G751 at heavy load. 3C may or may be a lot, so let’s use a better measure. 5% hotter than the best. Remember what I said a bit earlier – we’re getting good results from ‘standard’. (Digression – list is not what you’ll find retail. I’ve seen G751 as low as $2.50 .)

If my business takes off I will probably need industrial lots of compound, and having 60g is worth the purchase. Now when I put this thing together I’ll possibly be adding more than one heat sink – in addition to the CPU I may be adding heat sinks to graphic processors. (depends on card and whether it has one or not). Still, that’s only two to four applications. G751 will go for at least two and possibly four if applied correctly.

Applied correctly? Yeah. Remember, all we’re trying to do here is fill the air pockets with something that works better than air. TIM is better than air – but it’s still worse than metal to metal. We don’t want to apply so much as to insulate the latter.

A couple of comparison sites I found independently discovered that ‘the best’ application method depended on the material. Hmm, that needs more. OK, there are two groups of recommended application process. Group one says put a small amount on as a drop and let the pressure of the plates against each other do the spreading. Group two says spread it thinly and completely first. GENERAL GUIDE – the thicker the paste, the more likely group two is right. G751 is a group two, while AS5 is a group one. (MG is a group two.)

Bottom line – for what I’m building, picking based upon price is sufficient. Barring a (strong) possibility, I need to avoid false economy (ie, paying for a lot due to the great price/quantity, only to throw most of it away after not using it). If and only if prices are equal do I need to pay attention to quality – and even then there’s no clear ‘best’. (Yes, the two Shin Etsu are ‘best’ in the aforementioned test. They beat the ‘pack’ by about a degree C. The “pack” is then within a fraction of a degree C and contains several choices, to include AS5, AS Ceramique, and AS MX-2.

A note on applying the ‘thick’ stuff. Overclocker forum discovered the ‘best’ way to apply this was to put it on the heat sink instead of the cpu, then press the two together. To avoid ‘excess smear’, they recommend using a marker (toothpaste is one example) on the cpu, press together, then separate. Clean the CPU. Mark the ‘edges’ of the mark on the heatsink, clean the toothpaste (thoroughly) and then create a mask from masking tape. Now the credit card application process could be used without worry of breaking the motherboard under the pressure of spreading the paste. (Which also, by the way, allows the film to be even thinner in application.)


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s