Choice of thermal compound

[mubs]

Which thermal paste is best? I'm looking for good performance, and more important, long-term stability - I shouldn't have to clean up the goop and re paste it every year or two. Then again, I might be moving in a few months and will most certainly need to take the planned heatsink (Thermalright XP-120) off for the move.

Arctic have 3 varieties on their web site, with no comparison or guide: AS5, Ceramique and Alumina. Slightly differing characteristics, with AS5 having best performance by their own admission. An email to them has gone unanswered.

Please don't tell me to use the goop from the shop around the corner; I won't. I want this to be a do-it-once and forget it thing.

Incidentally, anybody know the maximum recommended weight for a HSF on socket-939? Searching AMD's website has been futile, and Google's hits indicate 450g as the magic number.

Thanks.

 

[Handruin]

I'm guessing you'll get some posts that say to use thermal pads for long-term stability. I can't say if they are any good or worse for heat transfer, but I tend to feel they are worse on an informal level (so take it with a grain of salt). I personally use the AS5 but I haven't had it long enough to know if it needs to be reapplied every year or two. Some sources say yes (for any grease), but I'm not sure I believe them. I went well over two years on my Athlon 2200+ before switching the HSF and I had no problem with the run-of-the-mill coumpond that came with the Spire HSF.

As for the weight of a heatsink, right now I'm sporting the thermalright XP-90C and their website tells me it weighs a heafty 690g "(heat sink only)". Hopefully that gives you some comfort. Also keep in mind I had to buy the special backplate for my gigabyte board to support this monster. That was under $10.

 

[Gilbo]

Dan did the definitive thermal goop analysis. I believe vegimite was the ultimate champion. It may be hard to source outside of Australia though. So you may have to go with some variety of real goop (vegimite also dries out anyway (high performance in the short term, but messy in the longterm ).

If you don't want to bother with that excellent article, everyday thermal goop should be just fine .

 

[Mercutio]

My understanding from several articles on the subject is that pretty much anything that's advertised as thermal goop is going to work about as well as anything else.
I prefer - and tell my students - that thermal pads are the better deal if you can get them. It's very easy to misapply thermal paste. The ideal is a nearly see-through coating, which to a lot of people is counterintutitive. Much more than that and it decreases contact between the HSF and the CPU...
Anyway, silver compound does seem to work better in my experience than white stuff, even if it's just one or two degrees F, so that's about the level to which I am selective.

 

[Sol]

If you really only want to do it once some thermal epoxy would work... The performance isn't as good but it will never dry out... Or it will, but it's supposed to...

Of course there isn't really any going back when you've glued the heatsink on so it's not such a brilliant solution...

I tend to use arctic silver but I'll guess that the goop that comes on the bottom of a lot of heat sinks would last longer. Probably the difference between a silver based goop and the stuff that'll come with your heat sink will be very small though, maybe a few months of wet time...

 

[time]

Dow Corning thermal transfer compound (white goop) has long been proven to last many years in general electronics applications. The pads are even better, partly because of their tolerance of movement - an issue with oversized coolers.

I'm surprised that no-one has pointed out that with the advent of heat-spreaders, thermal conductivity of goop is no longer such a big issue anyway.

Although I've used Artic Silver a lot, in hindsight I'm unconvinced that it's achieved anything. I'm also not confident about its long term stability.

 

[Bozo]

I use Dow Corning 340 silicone heat sink compound.
At work this stuff last for years in our electronic equipment. I'm talking 10-15 years or more on SCR drives that take more abuse and heat than a CPU will ever get.

Bozo :mrgrn:

 

[Tannin]

The white goop dries out. No question of that, I've seen it too many times, especially on stuff from the Athlon Thunderbird and Pentium III generation. Did they get hotter? Has the goop improved? Or, as Time so rightly points out, did it just stop mattering now that we have vastly better HSFs and built-in heat spreaders?

If you want it to last forever, use a pad. Otherwise, any compound will do, but try to avoid the white goop, it dries and cracks. I don't care what people say about its use in other contexts, I've seen it dried out many, many times with CPUs.

 

[Fushigi]

I believe vegimite was the ultimate champion. It may be hard to source outside of Australia though. So you may have to go with some variety of real goop (vegimite also dries out anyway (high performance in the short term, but messy in the longterm ).

For those who want a partially edible motherboard, Cost Plus World Market normally carries vegimite in their B&M stores.

 

[mubs]

Thanks for the opinions, folks, I really appreciate it. Buck hasn't chimed in yet, though.

The white goop dries out. ... Otherwise, any compound will do, but try to avoid the white goop, it dries and cracks.

Having built 0.01% of the number of systems you've built, I'm not sure I understand, Tannin. I thought all of them were white. When you say to avoid white goop, are you referring to a brand / type? What other colors is it available in??

It looks like Dow Corning stuff is the way to go. I certainly don't see industrial users buying large vats of Arctic Silver. And a company that cannot differentiate between its products and ignores requests for information makes me suspicious.

 

[mubs]

According to Dow Corning, AMD uses Dow's TC-5022 in it's factory heatsinks for A64 Proc-in-box.

Dow claims "TC-5022 improves thermal performance 10-15 percent compared to existing materials".

This page at Dow lets you compare the compounds, including TC-5022 and 340. Not that I understand all of it.

 

[mubs]

Bozo, where do you buy the Dow 340? I can't seem to find it anywhere online (pricewatch, pricegrabber, google).

 

[iGary]

Now that we are on the subject of cooling, an idea that has been kicking around in me noggin' for years is the fact that when you turn computers off, the hot toasty parts no longer get cooled, even though there is still plenty of heat emanating from those components. If some of these components were operating at borderline-critical thermal levels with cooling fans running, their temperatures would likely rise into critical territory for a short duration once the system was shutdown.

Component life would be significantly increased if you had a BIOS that could delay the cooling fan shutdown a few minutes after the system was shutdown (this would be pre-programmed BIOS settings), or have the BIOS to independently monitor temperatures of the microprocessor(s), hard drives, and/or GPU, shutting those cooling fans off to those components once the temperature was low enough to ensure that no thermal damage could occur from latent heat.

Maybe this cooldown cycle capability is already a reality on some computer systems. If it isn't, it really should.

PS: In the modern era, all I've ever used is a *thin* even coat of Arctic Silver 5.

The idea with thermal compound is to couple the heatsink to the component as efficently as possible. This means filling in all the tiny imperfections on the surfaces of the chip package and the heatsink and compensating for the likelihood of minutely non-parallel or uneven surfaces. The thermal compound also has to stay pliable due to the thermal expansion and contractions of the heatsink throughout the service life of the system.

 

[Mercutio]

I noticed on one of my X2-based systems that the PSU fan and CPU fan stayed on for ~5 seconds after the system powered off. I don't know how much good that does, but I don't remember PCs doing it before, and I thought it was odd.

 

[iGary]

Bozo, where do you buy the Dow 340? I can't seem to find it anywhere online (pricewatch, pricegrabber, google).

You can buy the stuff at any *industrial* electronic parts store (i.e. -- not a Radio Shack).

By the way, unless something recently changed, it's actually Dow/Corning thermal compound. It's been around for decades.

 

[iGary]

I noticed on one of my X2-based systems that the PSU fan and CPU fan stayed on for ~5 seconds after the system powered off. I don't know how much good that does, but I don't remember PCs doing it before, and I thought it was odd.

Since the modern desktop computer and operating systems issue commands to the CPU to shutdown (as opposed to being powered off hard via a power switch supplying the power supply), in your case it sounds more like the 5-volt fan power connections on the mobo are designed to simply bleed-off the reservoir capacitors in the power supply once the rest of the system has been shutdown.

 

[Will Rickards]

Maybe this cooldown cycle capability is already a reality on some computer systems. If it isn't, it really should.

It does seem to be implemented in cars though. It really freaked me out when I shut off my 94 honda accord and some time later the fan kicked in.

 

[Bozo]

I've used the Dow/Corning cream for over 20 years on heatsink/SCR assemblys.
I don't know where we get the stuff though.


Bozo :mrgrn:

 

[Sol]

I'm pretty sure that unlike car engines a CPU can not get any hotter after it stops running. With no electricity running through it it will not produce any more heat and since the heatsink will generally be slightly cooler than the CPU the laws of thermodynamics (as I understand them) would tend to preclude the actual core getting any hotter.

Actually I don't think a car engine can get any hotter overall either just in that case some parts of it which are kept cool when running can heat up which could be bad. In the case of the CPU the only part that can get any hotter is the heat sink so no one really cares.

Several products have been produced which will keep the fans running after the system shuts down, but I suspect that they are more based on the fact that people intuitively feel that this would help and hence buy them rather than on any scientific evidence.

 

[Sol]

Oh and I'm pretty sure I've seen Dow Corning thermal paste in Dick Smith or Jaycar so int the US I guess I'd try Radioshack or somewhere.

 

[Tannin]

I am 100% with Sol on this one. With a car, the heat-critical components of the system (typically the lubricants in the turbo bearings) are not the bit that generates the heat. They have to stay cooler than the heat source (exhaust gases). In a computer, the heat critical component is the heat source.

Ergo, after-shutdown cooling is 100% pure fart gas.

 

[mubs]

Enermax's Noisetaker line of PSUs are designed to exhibit "diseling" characteristics - fans run on for a few minutes even after computer is shut down. See 4th item in product features here, labeled "Smart Fan". IIRC, Merc's power supply couldn't power an X2, so he bought a 600W Enermax that solved the problem. Methinks he bought this one.

Oh, and Tannin, Sir, you did not answer the question in my earlier post.

 

[Tannin]

Just generic white goop. Mubs. Maybe some of the white stuff is perfectly OK, but who knows which sort is which? I use white goop for stuff that doesn't matter, and either Arctic Silver or a gold-coloured one that Geil gave us a dozen tubes of a few years ago. Both seem fine.

 

[mubs]

Thanks.