Chipset Cooling

How-To – Jason Choong

Pic

Is it worth the effort cooling a chipset?

Overheated system chipsets can cause instability, causing unpredictable system crashes. It tends to happen when an inferior wafer/die is used in manufacturing. If the power dissipation exceeds specification, cooling is required or it will kill the chip. That’s why el-cheapo VCD players always break down – so you have to buy again.

Many chips didn’t get to see the open market – they died in the assembling plant. Those that made it are tortured to death by extreme heat. As a rule of thumb, if we lower semiconductors temp by 10ºC, we can double its lifespan. How good are these chipsets?

Pic

Hitachi Super-RISC CPU – running fine after 30,000 hours continuous in an industrial environment without cooling

The VIA KT400 on my MSI MBM was way too hot; the heat sink was burning hot. After several failed attempts, I decided to superglue a Pentium 200 MMX heatsink on it. Wow! The heat sink was as cool as room temp. Then I used the original heatsink and superglued it on the Southbridge VT8237, which was not very hot.

I’ve tried mounting a heatsink to the power transistors of my MSI KT3 Ultra MBM with liquid superglue. It lowered the temp by a degree or two with fan cooling. It’s been at least six months and the heatsink didn’t drop off. At moderate temp, these non-heat-resistance super glues seem to hold. Once glued, it’s almost impossible to get it off. Using excessive force might rip the chipset’s plastic/ceramic mould. I used Loctite super liquid glue in my projects.

The VIA KT333 and KT400 release quite a lot of heat. The Pentium 586 heatsink has about 135 cm² of surface area. With a forced air convection, the KT400 chip was only a little warmer than ambient. If we use, say, a passive heatsink with only 96 cm² on the KT333, it gets quite warm. For the Southbridge VT8235 with 96 cm² of surface area, it is not required as it releases less heat. The first test was conducted after at least a few hours after bonding. Subsequent power ups were done after half an hour.

Pic

The SBlive chipset is quite hot; with 10 cm², the heatsink itself is still bloody hot! I used 48 cm² heatsink. These heatsinks can be obtained from an electronic surplus shop. We can also dismantle some old electronics gear to get some useful heatsinks. Old Celeron heatsinks are small and efficient enough for Northbridge chipsets and hot AGP chips, but you may have to sacrifice a PCI slot.

Pic

If you decide to make removable mounted heatsinks or testing a heatsink, there’s one way I’ve done it¹ – this is important if the product is still under warranty.

Pic

First put an aluminium tape on the chip, then glue the heatsink on. With enough force, the heatsink will come off. The most it will screw up is the tape, not the chip. Gluing a heatsink on a hard disk is not going to do too much. The best option will be to blow the HD with a fan, preferably from the bottom, cooling the board as well. Some chips on the HD are quite hot, so cooling it may be a good idea.

I told my friend about it. He went to the extent of gluing the Athlon-like Intel chipset. I tried to say no but he did it anyway. The heatsink seemed to hold. It solved his overclocking problem – sometimes it’s the chipset that cannot tolerate the overclock, not the CPU.

Pic

Making the case temp equal room temp

Pic

Cooling everything!!

¹ED NOTE: A simpler alternative is to use thermal tape for heatsink mounting – adding small pieces of aluminum in electronic gear could lead to some problems.

Tags:

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Discussion
  1. Correct me if I'm wrong, but the author seems to be in error concerning his ideas about where heat goes, and why it goes there.

    Heatsinks, ideally, should be just as hot as the chip they're connected to. The idea is that the waste heat generated by the chipset will, through conduction, bleed into the heatsink and will, through convection, bleed out into the surrounding air.

    A cold heatsink means one of two things. Either the chipset isn't generating any waste heat (which would be nice) or the heat from the chipset isn't bleeding into the heatsink (which isn't nice at all).

    I'd rather my case temps be a few degrees above ambient, rather than watch my north/southbridge fry. Be afraid of cold heatsinks. Be very afraid. :)


    :beer:

    Why is this article even published?
    Correct me if I'm wrong, but the author seems to be in error concerning his ideas about where heat goes, and why it goes there.

    Heatsinks, ideally, should be just as hot as the chip they're connected to. The idea is that the waste heat generated by the chipset will, through conduction, bleed into the heatsink and will, through convection, bleed out into the surrounding air.

    A cold heatsink means one of two things. Either the chipset isn't generating any waste heat (which would be nice) or the heat from the chipset isn't bleeding into the heatsink (which isn't nice at all).

    I'd rather my case temps be a few degrees above ambient, rather than watch my north/southbridge fry. Be afraid of cold heatsinks. Be very afraid. :)
    ok, so its not the best of ideas, but its not the worst. i guess just an example of what can be done in a pinch. and take it with a grain of salt. i thought it was a kinda neat idea though, although i do believe im the only one.
    With all disrespect to the author of that article, I find this advise to be poor. Maybe I am missing something in regards to that foil's installation (yes I read The_Jizzler's comments). This foil is thicker than say a quality thermal tape, and that Loctite glue is a thermal impotence and may act as an insulator.

    Honestly, this article I cannot agree on anything the author suggests, especially when you can buy 2' of quality thermal tape for $3 shipped, which would be a better alternative. Obviously a thermal adhesive would be the best option. I only mention thermal tape is because the person implied that the warranty would be intact using this crazy method:
    "If you decide to make removable mounted heatsinks or testing a heatsink, there's one way I've done it - this is important if the product is still under warranty."

    This is also incorrect, as soon as the heatsink is added/removed, the product's warranty is voided.

    I have asked Joe to look into this and perhaps add a footnote to this article...
    lol, actually i just reread his article, and i had it all wrong as to what he was doing. now that i understand i think its brilliant! basically hes taken some of that heavy aluminum tape they use for HVAC duct work. like this http://www.globalsources.com/gsol/I/Industrial-tape/p/2000000003844/3000000149681/sm/1001890851.htm

    and cut a square and put it on the chip, Alum. side up. then take some ca-glue and apply it to the now alum. clad ship, just like in his article. glue on the heat sink. now if you need to remove it the only thing that ever touched the actual chipset chip was the sticky side of the aluminum tape. now tell me that isnt as slick or not better than any other thermal tape out there! i dont see how it cant work better than that spongy crap they sell. i got a roll of that stuff, i should do it up on a few things and check it out for my self.
    combos of super glue and thermal paste have been seen on these forums for years, so whats the big deal there? anyway, as far as the superglue, he clearly says ots temporary. a way of trying out a new design, without permannatly attaching it in the event your board has no holes/ loops or any other way to mount it other than glue. better than thermal tape, worse than thermal epoxy, but hey its removable.


    I believe the problem most of us here are having is the use of superglue and only superglue on the chips. Is it permanent ? No idea, and I'm not trying it out anytime soon.

    Yes I do agree that the superglue/thermal paste combo has been seen and used around the forums for quite some time, but the difference here is that we mainly use thermal paste at the center of the die/chip and only around 4 dots of glue at the corners to hold the hsf in place. However in the article the author did not mention the use of any TIM at all and only superglue, imagine spreading glue all over the hsf as the way you would a TIM and attaching that to the chip.

    How semi-permanent is that gonna be ? Not only is super glue not that a great conductor of heat but when u remove the hsf you risk damaging or at least leaving behind alot of glue on the chip, impeding heat transfer to whatever you're gonna put on it next.
    combos of super glue and thermal paste have been seen on these forums for years, so whats the big deal there? anyway, as far as the superglue, he clearly says ots temporary. a way of trying out a new design, without permannatly attaching it in the event your board has no holes/ loops or any other way to mount it other than glue. better than thermal tape, worse than thermal epoxy, but hey its removable.


    Were the author to have proposed the admixture of adhesive and paste, I might have lent his missive more credence. As it stands, however, the article suggests techniques which are below the standards of this community and thus should not have been posted.
    combos of super glue and thermal paste have been seen on these forums for years, so whats the big deal there? anyway, as far as the superglue, he clearly says ots temporary. a way of trying out a new design, without permannatly attaching it in the event your board has no holes/ loops or any other way to mount it other than glue. better than thermal tape, worse than thermal epoxy, but hey its removable.
    The thought of superglue coming near anything in my rig is laughable, especially chipsets. What a mess. I wouldn't even superglue a cable out of the way for management, that's what god made velcro for.
    I only just read it and I'd have to agree. This article is not up to Overclockers usually high standards.

    Following the procedures in that article could potentially be harmful to performance. Even stock TIM will perform better than superglue.

    If you really can't get ahold of artic silver adhesive, and dont want to use thermal tape, at least use whatever TIM you've got on the chip, and just put a drop if superglue on each corner of it, then slap your heatsink on. Much better performance, superglue holds it on, but it's still removable if you should be so inclined.
    I agree that the article is really not up to the standards of this community. The methods and procedures outlined are, at best, poorly thought-out and display a certain ignorance of modern materials and devices available. We're I the nit-picking sort, I'd even point out the terrible wire management shown in the photos...:rolleyes:
    I want to mention that a lot of chipsets are not flat. Their ceramic/ plastic molds are very uneven(people who had NF2s) know what I'm talking about. He should add some lapping temps to chipsets as well.

    I got to use a lower voltage (VDD) and noticed that the heatsink was actually wisking heat away after I lapped my NF2 NB and SB. Not to mention it fixed the SS Crackling.
    I expected something like coming a custom heatsink to the big boys and got nothing. It wasn't up to OC standards in my opinion. Should you always use artic silver alumina or could you just use a ceramique? Or does it depend on the hold down mechanism?
    Well if your going to buy aluminum tape and superglue, why not buy the products designed for thermal transfer and adhesion, you will most certainly get better results. The article should a least contain some kind of disclaimer. I for one would never try that method and actually found the article laughable, although at least it looks like you have nonfunctioning 'cool' chipset sinks great for bling factor.
    The glue goes on ALU tape in the instructions in the artical, ALU tape is actualy in contact with the chipset.

    Is that relevent? The glue will still give way and the sink will eventualy fall off is the concern here right?

    Just trying to clarify thats all :)