Mounting a Heat Sink

[Will Rickards]

So I've done 1 AMD processor heatsink install about 2 years ago.
For my new computer I ordered the artic silver 5 and plan to use the stock AMD heatsink. The processor is an AMD x2 3800+. I think I'm irrationally fearing installing this heatsink. Any picture or video guides out there to allay my fears?

 

[Tannin]

Unless you are overclocking or have bad case temps or something, you don't even need the Arctic Silver. The hardest part about installing the HSF is getting the cardboard box open. Serously!

Plug in CPU, place HSF on top (either way around), slip handleless clip over clipping point, rotate the other one (with the little handle) around so that it can reach the other clipping point and slip it on, rotate it (it's an over-centre device) so that it's firm.

Really, really easy.

(Actually, I liked the ones you used a screwdriver for best, but most people will think that these are the easiest HSF units to fit since .... um... there was a very similar one used by Intel maybe a couple of years back. The latest Intel ones seem to require a degree in spherical geometry to figure out, at least until you get used to them.)

 

[Buck]

I concur with the captain.

 

[Bartender]

I've seen Buck clamp heatsinks on a few times, it seems really easy -- even a monkey could do it.

 

[sechs]

Also, if you're getting a retail CPU, it will come with the interface material -- usually a thermal pad already on the heatsink. You will have to use that to maintain your warranty.

 

[Handruin]

When I built a Sempron a few weeks back, the retail heatsink had thermal grease, not a pad. I bought the Arctic silver's ArctiClean product and it worked very well for removing thermal grease from the stock AMD heatsink. Apply a few drops and let it sit for 30-60 seconds. Wipe off with a cloth...it was that easy. Two thumbs up for anyone else considering it.

 

[time]

The Athlon 64 contact area is huge compared to Socket A, so there's no point in using Artic Silver unless the thermal pad becomes damaged (after removal, for example).

 

[Handruin]

Even in the situation if/when AMD ships with thermal grease rather than a pad, would you suggest leaving the original grease?

The Sempron I spoke of has the large heat spreader (socket 754). I felt it was worth applying ArcticSilver5 vs the OEM gunk. I might add they apply a large amount to the stock HSF. Much thicker than I prefer to use.

 

[tazwegion]

I think the question is self serving... rather would you use K-mart/payless (recycled) OR Penzoil/GTX2/Valvoline (you know what I mean ) oils in your Ferrari?

Both are lubricants, though only one is the result of an intensive QA program, hence the same with thermal compounds... if you've got AS5 why not use it instead of the OEM 'vanilla' pastes, Silicone Vs. Silver thermal (POLYSYNTHETIC) properties?

 

[Tannin]

Why not use Arctic Silver?

• Because Arctic Silver is expensive and, in the case of (e.g.) a Sempron 3100, provides no added benefit whatever. In this type of CPU, thermal issues are a complete non-issue. Zip, nada, nil, not part of the agenda. You could use stawberry jam and it wouldn't get anywhere near its thermal limits.
• Because, for all its virtues, heat sink paste dries out over time and gradually becomes less efficient. Thermal pads don't. So you might get (say) an immediate 3 degree drop by using paste instead of a pad, but in two year's time, the paste-based installation will be running 5 degrees hotter than the one with the pad. If you use paste, you have to make sure you lift off the heat sink every year or two, inspect, and renew the compound. (Learning about this stuff is one of the things you have to do to become a qualified (i.e., certified) AMD reseller. Doubtless Intel have a similar scheme that teaches the same things, though I wouldn't know about that as I don't sell enough Intel CPUs to worry about Intel certification.)
• Because with several other types of CPU, cooling is critical and you need every degree you can get. With (e.g.) an Athlon XP 2000 (old style), a Thunderbird 1400, or some of the older P4s, you need Arctic Silver. It's only common sense to save the expensive paste in the tiny tube for the jobs that actually need it.

BTW, quite a few of the recent AMD parts come with jut a smear of el-cheapo paste instead of a pad. Obviously, AMD regard cooling as a non-issue for them, as otherwise they would use something better. They always do where cooling actually matters.

 

[Handruin]

I'll agree that it wouldn't make sense to buy an expensive tube, just for a Sempron install. However, since I had some, and it takes very little to make an effective application, I decided to use it. My reasoning was to aid the cheap AMD HSF as much as possible with transferring heat.

I've yet to be convinced that thermal paste needs reapplying every year or two. Have you had situations where it's made a significant difference? In my current machine (XP 2200+), I've used the existing thermal past for 2+ years. I only recently changed to AS5 because I switched my HSF. The existing grease looked fine, it wasn't baked on.

 

[LiamC]

Dan Rutter over at DansData, who had the best cooler comparison anywhere for a couple of year--and had (as much as could be expected) a sound methodology--found absolutely zip difference between greases--including AS 3/AS 5/AS Ceramique, other super dooper wonder greases (KY? ) and plain old white Dick Smith/Tandy/Radio Shack/Yum Cha grease. He did find a repeatable, measurable, difference between thermal pads and grease (1 or 2 C), but nothing like some sponsored links/reviews would have you believe. He also showed that pads and/or grease is better than no interface material at all. What Dan did find was that if the grease was applied too thickly, or contaminated (carpet fibre, human/pet hair etc), then its performance dropped off markedly.

Note: heatsinks from a few years ago have measurably rougher contact areas than now. The retail heatsinks that AMD/Intel ship today are very much better today in terms of bulk (ability to absorb heat), design (ability to radiate heat) and contact surface (they now approach mirror like).

AMD identified a couple of years ago the problem of "pump-out", where grease is forced out of the junction between CPU and heatsink--but this would take at least a couple of years (depending on # of power cycles--it is the heating/cooling phase that does this). Also, this was the time of the spreader-less K7. With the heat spreader on the K8--and consequant
larger surface area--this phenomena would take much longer.

AMD's use of thermal pads was based upon a little less efficiency for greater longevity--they test and base the lifespan of their processors on ten year intervals.

Until someone can point me in the direction of a definitive study on this (other, and contradictory), then I'll go with Dan's analysis.

 

[Tea]

Quite so, Doug. Our situations are a little different. You will most likely never use up the tube of AS before it gets lost, turns into putty, mistaken for toothpaste and swallowed, stood on by an elephant, whatever.

For me, on the other paw, evrery time I use a bit of Arctic Silver, it's one day closer to the day I have to tell Tannin to order a fresh tube. And then I have to listen to him grumble about how much money I waste. (Shhesh! Me waste money! You should see Tannin when he gets hold of a credit card!)

Have we seen situations where the paste has dried up and become useless? Yes indeed, lots of times. Quite a few times anyway. It seems to happen most frequently where the machine tends to run hot anyway (which sort of stands to reason) and where the paste in question is that cheap cruddy white stuff. (Possibly only certain brands, but who can tell after it's been on for a year or two.)

We used to see it quite often back around Thunderbird 1300 through Athlon XP 1800 days, much less often since then. Have the paste manufacturers improved their quality? Or is it that processors tend not to run so hot anymore? Or is it just that we switched to using Arctic Silver on any machine that seemed to warrant it several years ago and AS is (probably) better in this regard than cheap brands? Or perhaps it's because for the last year or two on new builds we mostly use the stock heatsink with the stock thermal pad and only reach for the paste infrequently now? Maybe all of the above.

 

[Tea]

Great post, illi. One point I disagree on: we have observed better performance by better pastes too frequently to entirely disbelieve it. It's not much, but in the old days when even 1 degree was worth fighting for, Arctic Silver delivered it where crappy white stuff didn't. Too many times for me to believe it was just coincidence.

As for Arctic Silver Warp Drive 7 being noticably better than Arctic Silver II , I don't believe it. Any decent quality paste seems to be OK.

 

[LiamC]

Tea, is that "crappy white stuff" that you applied versus AS you applied or "crappy white stuff" that clueless customer applied versus AS you applied?

Didn't mean to imply no difference, just not enough to fight another war over. I'll defer to your superior judgement in the crap v good area BTW.

Just anecdotely, I recently pulled the Zalman off of the 754 because it needed the dust removed, and I cleaned & replaced the heatsink with no grease. Ran it that way for a couple of months. Without grease, it idled (generally) high 30's low 40's. Cleaned HS & replaced the grease about three weeks before swapping over to the X2. The S754 idled at low 30's, so no grease is not a good idea. Meanwhile, the highest temp I've had the X2 up to is 36 C with it under full load, both cores for 40~50 minutes. The E revision cores really do run very cool.

 

[tazwegion]

Great post, illi. One point I disagree on: we have observed better performance by better pastes too frequently to entirely disbelieve it. It's not much, but in the old days when even 1 degree was worth fighting for, Arctic Silver delivered it where crappy white stuff didn't. Too many times for me to believe it was just coincidence.

Right on primate! :aok:

I wasn't aware of the DansData cooler review... but found in my own dabblings that AS5 works better period! for example lapping the h/sink surface + 40cfm fan + AS5, brought an OC'ed 2500+ (2.2Ghz) down 10c total, using a standard issue Heatsink with the AS5 achieved a -3c result on it's own... :mrgrn:

Whilst on the other hand a prebuilt Intel vomit box (Hardly Normal) belonging to my brother-in-law had mere 'vanilla' under it's heatsink... so I figured why not AS5 there too, once again a -3c result :aok:

The ONLY time I ever use thermal pads, is when I need to affix an after market chipset cooler and no mounting points are present, then the adhesive TP's are just fine

 

[tazwegion]

Just as an after thought LiamC...

Once a thermal compound has been heated it seeps into the 'pores' or surface imperfections (if you prefer) of the heatsink, and AFAIK remains there indefinitely... inherently effecting the performance on the next level of thermal compound to be applied (though only microscopically), this is why many people remove the T/Pad prior to primary installation

 

[LiamC]

Just as an after thought LiamC...

Once a thermal compound has been heated it seeps into the 'pores' or surface imperfections (if you prefer) of the heatsink, and AFAIK remains there indefinitely... inherently effecting the performance on the next level of thermal compound to be applied (though only microscopically), this is why many people remove the T/Pad prior to primary installation

I prefer science myself, to voodoo or faith :wink:

AMD's testing on thermal pump-out would seem to be at odds with your belief. The makers of Artic Silver would also be at odds with your belief as I read an interview with aforementioned creator on the release of Ceramique in which he claimed that Ceramique was "immune" to thermal pump-out--which implied that Artic Silver wasn't :eekers: He rationalised AS by saying that overclockers wouldn't leave a heatsink on/CPU unchanged for thermal pump-out to be an issue.

And as I mentioned in my post, older heatsinks have less than optimal contact areas, the exact place that thermal compound would work.

As for your claims about -3 C, you of course did a double blind test where:

* you did not know what thermal compound was being used
* all compunds were applied in the exact same manner, on the exact same system, in the manner prescribed by the manufacturer
* ambient temperature was kept completely stable
* CPU load was exactly constant
* temperature readings were taken at measured intervals over a suitably lengthy period of time (to eliminate hot-spots, outliers, random "glitches" etc)
* the measurements were then subjected to suitable statistical analyses such as I mention here:

http://www.realworldtech.com/page.cfm?ArticleID=RWT061902001047

to reach a conclusion that yes Artic Silver is -3 C better?

Now I'm not saying you are wrong, but you have not proven your case scientifically (unless you did do all of the above). If you read Dans comments on grease/reviews/benchmarking temperature you would know how hard it is to acheive repeatable results--and most people don't care.

You would also know that in his testing, things such as a single hair or two, or sloppily (slightly too much) applied grease changed the reported temperature by more than changing grease.

Artic Silver is expensive no? Are you absolutely/able to be repeated every time sure that you aren't just taking more care applying AS than generic junk?

And as Dan said, there are generic white and yum cha--some are better than other, but just because it is white, does not mean it is no good. Vomit boxes are built to a price are they not? What makes you think that the builders take great efforts to apply grease in the correct fashion? It is highly likely that you applied it much better than Vomit box builder did. So how do you know, that it wasn't the careful application of thermal grease rather than AS. You are just focussing on the AS because that is what you used, rather than on of the other things that changed--that Dan Rutter at least has shown to have more bearing on CPU temps.

By all means keep using AS. It does work, and I myself use it, but I don't believe the hype.

</gratuitous Public Enemy quote>

 

[tazwegion]

Actually, I did use a scientific approach to the tests hell I even documented the process in two posts @ OCC, but owing to the phenomenon known as the 'hosting black hole'... most if not all the screen shots (that had been hosted on Mediapimp.com) are nothing but a memory, though the threads still exist (minus pictures) @
Project Stoker part I & part II

How is the concept that microscopic thermal paste particles become caught in the imperfections of a heatsinks surface, Voodoo or faith? if you were to look @ any apparently smooth object through a sub-atomic microscope (a really good one ), you'd see at some magnification all matter is not solid, thus having imperfections and holes

It's not like I sacrificed a sheep to the 'OverClocking Omnipotent One' and asked for a favourable outcome?

BTW I have since swapped over to www.Imageshack.us for my picture hosting :aok:

 

[Tannin]

Taz, you are right in at least this much: it is the microscopic imperfections of a heatsink's surface that give rise to the need for thermal paste in the first place. The physical structure of glass (AKA what a chip is made out of) is such that, no matter how carefully you polish it, only about 20% of the surface area actually makes contact with another flat object (such as a heatsink). With aluminium it's even less. (I used to know the figures, part of that AMD certification program again, but I've forgotten them. Does this mean that they will come and take my certificate away? They will have to find it first - I have no idea what it's filed under.)

Liam, I hasten to add, will be most unlikely to disagree with me here.

As to what real world effect having some Brand X goo stuck in those pores will have, that's another question. I'd vote for "not very much", but I'm only guessing. Depends, I suspect, on the conductivity of the paste in question. I'd imagine it would need to be really bad before it did much harm.

On a related matter: used goo removal tools. I use three:

* An 8MB 720pin SIMM. The square corners are well-shaped for this, and it's hard enough to scrape the goo off, soft enough not to scratch the heatsink.
* The roughened wood on the edge of my workbench. (Very thick chipboard, actually.) Rub it briskly over that for a few moments to polish it a bit. Being uneven, the wood works quite well, gets into little corners and stuff.
* Last but not least, any convient bit of flexible packaging foam for the final polish, the stuff that comes in motherboard boxes, for example. (Cloth would do, but needs washing. You have to throw gazillions of bits of plastic foam out every week, why not use them for something useful before they go into land fill?)

PS: Eeek! Did I really just say "gazillions"? Don't blame me it's Tea's influence.

 

[LiamC]

How is the concept that microscopic thermal paste particles become caught in the imperfections of a heatsinks surface, Voodoo or faith? if you were to look @ any apparently smooth object through a sub-atomic microscope (a really good one ), you'd see at some magnification all matter is not solid, thus having imperfections and holes

This is not what I disagreed with. Nor did I say that matter was not solid, though I'm fairly certain :wink: that thermal grease does not operate on the sub-atomic level. But what has this got to do with thermal grease? Thermal grease works by filling in the air gaps between two mated surfaces. The grease has better conductivity than air, but its still a hundred times worse than copper or aluminium--but as you alluded to (or at least I think it's what you were alluding to), flat seeming surfaces aren't indeed flat and it is hard to make any two surfaces mate with any degree of success--without spending inordinate amounts at any rate. If you could get any two surfaces "flat" and mate them, chances are the resulting stiction would make it almost impossible to separate them.

But, the grease itself would have to be composed of extremely fine particles--much finer than most greases--particularly AS--are made with today to really get into the gaps between todays reasonably flat heatsinks--and by gaps I mean the gaps that can't be seen with the naked eye. The little pockets you see when lifting a HS off of a CPU are macroscopic rather than micro. Next time you do, and if the grease is a little on the dry side, rub some between your fingers, you'll be surprised by how "gritty" it is. This stuff is way too big to be doing anything on the microscopic level. If you are attributing increased performance from repeated applications of grease, try lapping the heatsink. Metal to metal is much better than metal-grease-metal.

..."this is why many people remove the T/Pad prior to primary installation"...

And I thought is was because they did believe the hype.

 

[time]

Differences between thermal transfer compounds are mostly down to thickness of application. Certain compounds can be spread much thinner than others, but in addition, some compounds are a little more tolerant of extra thickness.

Consider that heat conductivity ranges from 200 to 400 W/m°K for aluminium and copper, but a puny 1 to 5 W/m°K for paste. Obviously, the thinner you can spread it, the better.

The problem is, if you make it too thin, you may create hot spots on the CPU surface. Of course, this is only really a problem with Socket A CPUs - the Athlon 64 and Pentium 4 copper 'heat spreaders' make them far more tolerant.

The phase change pads that AMD recommends are guaranteed to be uniform and therefore achieve a good result without risk. Exotic compounds may leave gaps - this leads to instability and premature failures.

Incidentally, phase change pads also "seep into the 'pores' or surface imperfections", but are designed to do so after reaching the phase change temperature, typically around 50°C. Given that Athlon 64s run so cool, I wonder if builders should make a special effort to raise the CPU temperature after installation to ensure this?

IMO, DansData is still the most reliable source of information for all this on the web. Notice that his observations tally with what the chip manufacturers recommend. Bill, you're well across your topic, as is Tannin.

Not too sure about the ape, though ...

 

[Gilbo]

Damn you LiamC for beating me to bringing up Dan's thermal goop comparison! At least you didn't provide a link, as far as I can see, so I'll still get a little of the glory.

Dan's Goop Groundup compares the advantages of, among others, Arctic Silver vs Vegemite vs toothpaste in a rambling entertaining style that's well worth a read. I won't ruin the suspense by giving away the ultimate champion, but there's really no doubt at all.

Mmmm, vegemite...

 

[tazwegion]

..."this is why many people remove the T/Pad prior to primary installation"...

And I thought is was because they did believe the hype.

Nah... it's because they didn't want thermal (phase change) pad remnants remaining within the heatsinks (pores) surface, which would be counter productive to the AS5 characteristics

Personally I'd prefer AS5 in those imperfections than T/pad remnants any day :aok:

Next time you do, and if the grease is a little on the dry side, rub some between your fingers, you'll be surprised by how "gritty" it is. This stuff is way too big to be doing anything on the microscopic level.

Yeah a proportion of the substance can become gritty... after all it's supposed to be a dense paste, but not all... remember everything exists on a microscopic (and as you pointed out macro-scopic) level, and it this quality/property which allows AS5 to seep in to the 'pores' without which it would be solid and ineffective

 

[tazwegion]

Tannin how nice of you to share your 'hi-tech' goop removal solutions & techniques


* An 8MB 720pin SIMM.

* The roughened wood on the edge of my workbench.

* Last but not least, any convenient bit of flexible packaging foam for the final polish

That's certainly more 'leading edge' than my cloth & methylated spirits approach :eekers:

Gazillions? really... next you'll be using the word Squillions as a volumetric description! :lol: we really must congratulate Tea for the positive influence

 

[tazwegion]

Hmmmm... a 720 pin SIMM? :eekers:

 

[Santilli]

Didn't I see this somewhere, argued about how many fairies can dance on the head of a pin?

:mrgrn:

g

 

[tazwegion]

I guess it would depend on if they were fully grown or not eh?

 

[sechs]

I've yet to be convinced that thermal paste needs reapplying every year or two. Have you had situations where it's made a significant difference? In my current machine (XP 2200+), I've used the existing thermal past for 2+ years. I only recently changed to AS5 because I switched my HSF. The existing grease looked fine, it wasn't baked on.

I have. I recently refurbished the ol' 2200+ for reuse. Replacing the paste gained a 5C or so reduction in CPU temperatures.

I suppose that there is likely some dependence on what the old and new materials are. Any material that's thermally cycled will have a tendency to not stay exactly as you put it in. Whether this is physical movement, or chemical or physical change can have a major bearing on how bad it gets over time.

I'm always awed by how much a little dusting can improve computer cooling....

 

[tazwegion]

I'm always awed by how much a little dusting can improve computer cooling....

Yeah 'DUST BUNNIES' are an overclockers worst nightmare

 

[mubs]

My wife showed me a department store catalog yesterday; fancy stainless steel cookware with copper bottoms. One of them said "copper bottom impact bonded for better performance. I guess the co-efficient of expansion must be similar enough among the two metals. And of course, they must be using quite a bit of ArcticSilver/I (industrial).

 

[Will Rickards]

Another semi-stupid question, do you guys think I should put the processor/heatsink/ram in before I mount the motherboard in the case or after? I was thinking before.

 

[Buck]

Another semi-stupid question, do you guys think I should put the processor/heatsink/ram in before I mount the motherboard in the case or after? I was thinking before.

If possible, before.

 

[Corvair]

Processor + Cooling: Yes -- at least the processor (so that you can observe the pins going into the socket).

As for the cooling assembly -- even though it's not really that big of a deal, why bother wrestling with the installation of the heatsink and cooling fan after the mobo goes in?

About the only thing that could cause problems for you by installing the cooling assembly (as I call it) on the mobo before it goes into the chassis is if you are using a *giant* freekin' heatsink like my Swiftech, which is basically a copper slab with aluminium porcupine bristles on top, and then large 92mm fan on top of all that. It was actually easier for me to install the cooling assembly last.



RAM: It doesn't matter. In fact, you had better be able to easily install or de-install RAM with the mobo

 

[Corvair]

...or de-install RAM with the mobo in place. (finishing sentence from above!)

* Because Arctic Silver is expensive and...

Arctic Silver Five is really not expensive here in the USA -- unless US$5.00 = expensive.

Arctic Silver 5 is well worth it if you are using one of those recent Pentium 4 thermonuclear devices for number cruching.

 

[Handruin]

I've always installed RAM and CPU after the motherboard is mounted in the case. Must be my personal preference, but i find it easier.

I just built a new sempron 64 2800+ tonight and I assembled it this way. I find it easier to click the RAM in when the board is on a solid surface (i.e the case).

No thermal pad on this CPU. I washed away the AMD gunk and applied a tiny amount of AS5. She's running between 28 and 30 degrees celcius on the stock cooler.

 

[tazwegion]

Another semi-stupid question, do you guys think I should put the processor/heatsink/ram in before I mount the motherboard in the case or after? I was thinking before.

I guess it'd depend on the space constraints within the case, also the size/type of HsF combination you intend to use and lastly the degree of difficulty associated with the retention mechanism of your cooling solution

As for RAM though... I wouldn't imagine it mattered either way, a personal preference perhaps? (though I don't)

BTW it can make life easier mounting your storage drives prior to installing M/board etc. IMHO

 

[Tannin]

In general, it is much easier to mount the CPU and HSF before you put the mainboard in the case. Quite often you can't get hte space you need to lever a HSF clip on/off once it's in the case unless you remove the power supply.

RAM? Who cares? But it's a good practice to mount te CPU, RAM, and video card you intend to use first, fire it up on the test bench, and then install it in the case. (Of course, to do this you have to have a test bench.)

We quite often skip this step, seeing as modern Gigabyte mainboards and Geil RAM are so trustworthy and practically never play up, but with older gear, or anything in the least bit questionable, we always do this. Test first, bolt in once is much better than bolt in first, test later, unbolt, fiddle, bolt, test, unbolt, fiddle .....

 

[Handruin]

I think not having a test bench is a key component to why I do not mount those items prior to installing in the case. I try to minimize the amount human handling of the motherboard. The faster I get it mounted in the case, the less I worry about killing it. If I had a test bench where I could easily mount/unmount a board, then I'd probably be installing CPU/HSF/RAM before installing in the case.

As for the hard drives, I find those easier to install after the motherboard. They tend to get in the way while the board is leveraged into the case. But again, it depends on the size of the case and board.

 

[Tannin]

That makes good sense to me, Doug. I do the same with hard drives. Besides, getting the motherboard in can sometimes require a fair bit of banging and sudden shocks (get in you damn thing!) so it's better not to subject the HDD to that if you can avoid it.