I am using this one: https://www.amazon.nl/-/en/gp/product/B07MPBFRRH. Worked perfectly for putting a loose heatsink back on a 468 processor. Also used it for some chipset heatsinks without problems.

Tape is better. Thermal glue may be more difficult to remove.

Yoghoo wrote on 2021-12-11, 00:24:

I am using this one: https://www.amazon.nl/-/en/gp/product/B07MPBFRRH. Worked perfectly for putting a loose heatsink back on a 468 processor. Also used it for some chipset heatsinks without problems.

Thanks for the tip!
Allthough this is exactly one of those items I've read various comments on, like :

As double sided adhesive tape this is fine.
When it comes to thermal conductivity, not so great.

Which makes me wonder how well it actually works for the cooling part of its job, and not just the sticky part... 😄

As for tape vs glue, I am not that worried. If I put this (small) type of heat sink on something, I plan to let it stay there.

I use these: https://www.coolsierra.com/collections/ec360/ … ucts/ec360-tape

Anybody up for some math?
thermal tape 3.0 W/mK conductivity, 0.15mm thick
cynoacrylate adhesive (crazy glue) 0.2 W/mK conductivity, thickness of film (The really runny stuff, not the gap filling stuff) on the order of a human hair, 0.001mm

3/0.0015m = 2,000W/K per unit area at that thickness
0.2/0.00001m = 20,000W/K per unit area at that thickness.

Presuming the irregularity of the surfaces joined is no greater than a human hair, I pick crazy glue.

edit: Ooops, the figure I got for film thickness of cyanoacrylate was 100 to 1000nm, 0.0001 to 0.001mm but I misread thickness of hair at 0.001 inches or 100microns to be equivalent to that, hairs be thicker.

Thing is.. crazy glue will not stay adhered due to the thermal cycling and different thermal expansion properties of the chips vs the heatsinks vs the crazy glue.

It used to be common to use thermal compound along with tiny dots of crazy glue on the corners to adhere the heatsinks to different chips. Works fine if the heatsinks are on top of the chips but if they are hanging either upside down or perpendicular they had a good change of randomly popping off and falling down onto whatever was below them.

As far as thermal tape goes, I tried for a few months to find any that actually had good thermal properties. I even contacted the places that make it for industrial applications. Basically even they don't have anything that is worth using.

Look at the low star reviews of all but the insanely expensive ones on Amazon. The people who actually seem to know what they are doing all say it is crap.

Same goes for the currently available thermal glues. The last good one was the Arctic Silver thermal epoxy which they stopped making years ago.

3M has many thermally conductive tapes with proper specs.
https://www.3m.com/3M/en_US/p/c/electronics-c … nductive-tapes/
Just be careful and check twice what you receive if buying from ebay.

Here's what I use for passively cooled chips: https://www.amazon.co.uk/akasa-AK-TT12-80-The … e/dp/B001GIM9V8 Not the best, but it holds, transfers heat solidly and isn't expensive.

Having to use tape implies that there is no need for high performance cooling, thus good quality tape such as hpfix should be more than enough.

I used glue to fix a fan to a heatsink, but I would not use it directly to a chip.

If the plan is to cool video cards, I found it more convenient to use a PCI fan holder, with 2 x 120 mm fans. Much better air flow and less noise, compared to smaller fans. I am using such setup to cool a Voodoo3, and it does a good job for the entire PCB and components.

cyclone3d wrote on 2021-12-11, 06:44:

Thing is.. crazy glue will not stay adhered due to the thermal cycling and different thermal expansion properties of the chips vs the heatsinks vs the crazy glue.

Times I've seen thermal tape let go, 7 or 8 (Tapes supplied with heatsink kits or coming on hardware with small sinks installed on RAM etc) seems to age poorly, goes the same as glue on other types of adhesive tapes, gooey, softer and prone to creep. Heatsinks on stuff not horizontally mounted tend to fall off when warm and tape old.

Times I've seen crazy glue let go, 0 .. though you can get it off with mechanical shock. I am not using it on larger packages like chipsets though.

JB Weld has more sense behind it than you'd think too, rather than dismissing it by gut reaction as redneckery... Let's consult some thermal engineers about it's properties... https://www.sciencedirect.com/science/article … 214157X17301156 ... 7.47 W m−1 K−1

JB Weld states it is electrically non-conductive, but that might be a more macroscopic assertion, I'd still be real careful, the tiny gaps between some pins might be thin enough to give you say a 10,000 ohm pullup or pulldown where it's not needed.

But anyway, application technique is the thing, you can see that if you just blodge it on and sit the sink on top so you end up with a half millimeter in there, it's gonna be worse than the thermal tape because it's three times as thick, with only 2.5x the thermal performance. So, apply very sparingly and use A LOT of mounting pressure.

The reason it's steel filled is not really so much for strength, glass bead filler would do that, but to more evenly match its expansion co-efficient to metals, so it stays stuck to them. That means that it will probably mediate well between high expansion of aluminum and very little expansion of epoxy or ceramic encapsulation.

BitWrangler wrote on 2021-12-11, 05:42:

Anybody up for some math? thermal tape 3.0 W/mK conductivity, 0.15mm thick cynoacrylate adhesive (crazy glue) 0.2 W/mK conductiv […]
Show full quote

Anybody up for some math?
thermal tape 3.0 W/mK conductivity, 0.15mm thick
cynoacrylate adhesive (crazy glue) 0.2 W/mK conductivity, thickness of film (The really runny stuff, not the gap filling stuff) on the order of a human hair, 0.001mm

3/0.0015m = 2,000W/K per unit area at that thickness
0.2/0.00001m = 20,000W/K per unit area at that thickness.

Presuming the irregularity of the surfaces joined is no greater than a human hair, I pick crazy glue.

edit: Ooops, the figure I got for film thickness of cyanoacrylate was 100 to 1000nm, 0.0001 to 0.001mm but I misread thickness of hair at 0.001 inches or 100microns to be equivalent to that, hairs be thicker.

The way I read your results: performance will be limited by the thermal conductivity of the heatsink; both tape /glue can transfer more heat than the heatsink (made of aluminum / copper) would be able to dissipate.

yah, but there's a millimeter of epoxy over the die that's probably only 0.3W/mK, so that's the bottleneck, plus whatever bulk of the die is in the way to the hotter part of the chip.

As for Arctic Silver, they apparantly(?, page dated 2017) still MAKE it, but not for sale to the general public...

http://www.arcticsilver.com/ta.htm

OEM/commercial: http://www.arcticsilver.com/epoxy.htm