Another method was to borrow from the design of 486 clip-on heatsinks. I cut off the clips from a black 5.25" bay cover. I then took a belt sander and made them a lot thinner so that the clips had some bend to them. Then JB-Welded them onto a chipset heatsink from a dead 2000's era motherboard. After drying, there's a clip or two which needs to be a good half-mm tighter, so I was think that I could either add some masking tape, or apply some more JB-Weld epoxy and file it down to size.

The problem with this design is that the heatsink can technically slip off from the two ends that do not have clips. On the 486 clip-on heatsink approach, the open ends have a raised lip to prevent this. I was thinking that I can either, a) make the clips sufficiently tight so that the heatsink is unlikely to slip off, or b) mask out two likes on the heatsink where the CPU will reside and lay down some epoxy to create a raised edge.

Due to the rarity of this CPU, I didn't want to use heatsink compound, thermal epoxy, or thermal tape due to fear that it will remove the surface font.

The fan is 5 V and 6.9 mm tall. I will add a connector to the wires to connect to the 5 V line of the molex.

I haven't done the masking out for the raised lip or tried to tighten the clips by adding an epoxy filler yet. I want to give the JB-Weld 24-hrs to dry first. What do you guys think? Is there a better way to do this?

Nice ideas 😀

Will come handy when I put mine into a fixed system... although will not be there for some time 😉

Just in case, I would advise against epoxy glueing the coolers to the chip, for two reasons:
- irreversible (cannot be removed)
- bad thermal conductivity (typically 0.2-0.3 mW/K)

I yesterday got a few sheets of Sekisui #5760. It is said to have 1.4 mW/K warmth conductivity, which is the best I could find when researching the technical specs of two-siided tapes commonly used/recommended.
It is very easy to handle and sticks way better than the also-frequently-used 3M dual sided tape (with which I had some bad fall-off experiences), thermally-wise and mechanically-wise. Things do not fall off by themselves.

Made a few photos yesterday when I retrofitted my 486 mobo with almost the same kind of coolers you pictured. Can post them later.

It is possible to get epoxied heatsinks off using a sharp wood chisel and a hammer. Usually one strike is all it takes. The problem is that the epoxy usually takes some of the lettering with it. You can also scratch the surface of the chip.

To prevent the chip from falling out, e.g. after some case vibration, I've epoxied in some stop tabs. I also decided that, rather than adding some epoxy to the retaining clips (to make for a tighter fit), I've used masking tape. Two clips received one piece of tape, another received two, and another three. It clips on quite tight now, but the perfectionist might want to dab on the epoxy to reach the thickness of the masking tape. It would require another day or two to dry and sand down the epoxy. Perhaps if there is some black masking tape, the quick solution would look a little cleaner.

I have an IBM SLC mobo that someone tried to identify by removing the heatsink from CPU with such a method. The result is missing CPU and 2 of its solder pads gone - because it was only 100 pin package and the bond with the heatsink was greater than force needed to shear the solder joints. So this approach migh work most of the time, but not always.

I'm using 800mW/mK tape that's 15mm wide and if I need a wider area I just put several strips of it side by side. It's a good tape, -20C to 120C rated, up to 180C short term, strong adhesive. You can get up to 3W/mK from the electrically non-conductive variety but these are expensive. And if 3W is not enough then there are aluminum-based tapes but these are electrically conductive obviously.

If you have a chip in plastic package then additional 0.2mm of adhesive tape between it and the heatsink is not a huge deal. The package itself will be limiting heat transfer. In general these chip run with no heatsink or are passively cooled anyway - so any additional surface area is helping as long as it's not mounted using a thermal insulator.

Ceramic packages can transfer heat more easily due to higher density but it's not such a big difference vs. plastic as you'd think. Thick enough ceramic brick is actually a thermal insulator. If you need to really cool things down there better be a metal plate - and then the choice of mounting method starts to matter a lot. A tiny bit of clear silicone grease will help with heat transfer as well and is usually easy to remove (and again, electrically non-conductive). Much like ceramics, such grease will prevent good heat transfer in a thick layer but a very thin one improves it by filling the plastic and metal pores.

As for solutions other than adhesives - the twisted wire idea is good but I've been using a thin nylon fishing line. A bit more difficult to work with than wire but won't conduct electricity at all so no need to worry about damaging insulation. You'll probably need to put the loops in a cross pattern to make sure the heatsink not moving around too much.

Deunan wrote:

In general these chip run with no heatsink or are passively cooled anyway

Which chip? The SXL2-66? The PGA-168 version comes with a heatsink, but I've never seen what the PGA-132 SXL2-66 interposer board looks like in the wild. Anyone know? I've never seen it with a heatsink, which seems strange since PGA-168 unit came with a heatsink. There are so few of these left in the world so it would be difficult to generalise. And J1 on the interposer is 5 V and GND, which I suspect is for a fan. Luckily the fan I chose for this project is 5 V.

Deunan wrote:

II'm using 800mW/mK tape that's 15mm wide and if I need a wider area I just put several strips of it side by side. It's a good tape, -20C to 120C rated, up to 180C short term, strong adhesive. You can get up to 3W/mK from the electrically non-conductive variety but these are expensive. And if 3W is not enough then there are aluminum-based tapes but these are electrically conductive obviously.

Wow... sounds good!
Wanna have!
Could you please hint what brand/types I should look for?

In general if the chip has a plastic package it's not supposed to heat up a lot. Obviously though the manufacturer might have just went the cheap route and you do need both heatsink and a fan - my point here is, the package will limit you way more than proper thermal adhesive tape. So there's little reason to be spending on expensive materials, you'll pay 2-3x the price and will get maybe 5C less on the chip in worst case scenario. But if you feel it's worth it, you can do just that.

The original SLC from TI (not SXLC) I have is for example packaged in what they've named "thermally improved package". Still plastic top like on any other PQFP but the bottom is a metal plate, and it's not for soldering ground connection like modern chips - just an improvement to allow better heat transfer to the PCB. The SXLC is fully plastic on the other hand and I've put small heatsinks on mine, using the very tape I mentioned.

The tape I got was this: https://www.tme.eu/de/en/details/tt707-15mm-2 … nductive-tapes/
It seems to be sold out now though, I guess I bought it on clearance sale? They do have other ones in reasonable prices though: https://www.tme.eu/de/en/katalog/thermoconduc … _stock=1&page=1

there is a socket3 holder in thingiverse if you happen to have a 3dprinter
it fits peefectly on the bottom borders of the cpu. the clips attach to a heatsink, was made for thin heatsinks but the clips can be modified if you know hot to edit the .stl

its way better tp do it that way you just need to experiment with the height of the heatsink

Maybe it's easier to just use a thermally conductive glue like GD9980 and a north bridge heat sink. It can be taken off by melting the glue at high temperatures and cleaned with some solvent like kerosene or acetone.

Attached is the final product. Perhaps it will help someone else with a non-standard CPU on PCB. I wasn't really wanting to get into thermal tapes in this thread as there was another thread for that, which I had started thinking that I would use thermal tape for this application, but decided against it. Thermal tape for heatsinks, particularly for DRx2-66 and SXL2-50+

I connected the 5V fan wire to J1 on the interposer and left the masking tape in place on the clips since it is barely visible. These chips will clock well to 80 MHz, but I'm not daring enough to test for 100 MHz. The system it is going in can only handle up to 70 MHz on this chip - at higher frequencies, the CPU won't clock double for some reason. If anyone has any input on that anomaly, please refer to this thread: How to get 486 SXL2-66 to clock-double w/AMI Mark V Baby Screamer motherboard

This chip is destined for my AMI Mark V Baby Screamer 386 with ULSI DX2-66 FPU, a SCSI2SD HDD, SCSI2SD - Comparison of SCSI SD, CF, and HDD performance , and native PS/2 mouse adaption, Re: Native PS/2 mouse implementation for 386/486 boards using the keyboard controller . Unfortunately, the CPU sits right under the HDD mounting bay and I estimate that the distance between the top of the fan and the mouting bay is 0.6 cm. I hope this is enough distance for the fan to operate properly. The fan runs at 7500 rpm and pushes 3.7 CFM.

The two washers under the fan are used as spacers because the centre of the heatsink is every so slightly taller than the surrounding fins, roughly 0.25 mm taller.

I was also wondering if having an inductive load (the fan) connected directly to J1 is the best approach. Would this introduce some distortion to Vcc? But on second thought, the CPU is being powered from a 3.3 V regulator on the interposer which should help in this regard.

Is anyone interested in creating the PCB for the PGA-168 version of the SXL2-66 to run on PGA-132? If so, Custom interposer module for TI486SXL2-66 PGA168 to PGA132 - HELP! . This way, just about everyone who wants to run an SXL2 on a 386 at 66 or 80 MHz can do so.

Finally got around to installing the CPU. Not sure yet if I want to push the FPU to 70 MHz.