Anyone here found a good thermal epoxy for T17+ heat sinks or similar? We tried that black epoxy from zeus mining, but it made the chips overheat right away. We're looking to avoid soldering the heat sinks since it risks losing solder balls that can lead to shorts and failures. The original solder paste used on these boards seems to be low-temp, which makes the heat sinks loosen during operation. So, we're on the hunt for a better solution. Appreciate any help!
Maybe try Arctic Silver thermal adhesive?
http://www.arcticsilver.com/arctic_silver_thermal_adhesive.htm
Please look very carefully for the information at their site, it's very strong stuff and requires careful storage.
I have no experience myself about re-installing loose T17 style heatsinks.
From the 'been there done that' department remember you are better off with heat sinks falling off then ripping the chips off the board when they take an impact.
The Arctic Silver that HagssFIN recommended is good as is Ther-O-Bond stuff:
https://www.mouser.com/c/thermal-management/thermal-interface-products/?tradename=Ther-O-Bond
Although it probably will not matter in a miners lifespan, the more thermally conductive it is the quicker the epoxy / adhesive breaks down.
It's just chemistry the more conductive stuff in it, the less adhesive you can put in there....
-Dave
If you try another adhesive, please post your results. Many people have talked about using Arctic Silver, but I have not seen a report of it actually working. Thanks for your results on the "black stuff" that was used on S9s.
The Arctic Silver adhesive, and other thermally conductive epoxies I found specs on, are all around 10X worse at moving heat than the low temp solder. So I have my doubts about it working.
It took me a while to work out a good procedure for attaching the heat sinks with the solder, but now I do it routinely and rarely have issues. If I get motivated enough I'll take some photos or maybe video and create a post showing my method.
Here are the basics of what I do. I clean the solder off the heat sink by applying flux, heating with a hot air tool, and then wiping away the solder with a cotton swab. Then I add a bit of solder paste to the top of the chip along with some flux, and use a soldering iron to flow the solder evenly and remove any extra. I then clean the heatsink and the chip with isopropyl, apply a small amount of flux to the top of the chip, place the heatsink on the chip, then heat with a hot air tool until the solder flows. You can adjust the position of the heatsink for only a small amount of time after you start heating, otherwise, you risk moving the chip. I don't touch it after about 5 seconds with the heat.
I believe the solder they use has a 138deg C melting point, so pretty far above where the chips can operate. If heatsinks are coming off because the solder is melting, I think you've got other problems ... that would probably require taking the heatsink off to fix anyway...
re: solder bonding of heatsinks vs using thermaly conductive epoxies.
There really is no comparison in thermal performance and I'd bet 10x is a low guess. Any sort of direct metal bonding is far superior for transferring heat. The mechanical aspects however obviously can be problematic.
Here are some #s
So comparing Bi58Sn42 to EP30TC from Masterbond, the solder is 6.6X better at conducting heat. That Masterbond epoxy is the best I've found so far, I think most are much worse than that, closer to 1.0 to 1.5 W/mK. I wasn't able to find a price for it anywhere... I'm suspecting it's one of those "if you have to ask you can't afford it" kind of situations though...
How thin you can get also linearly affects the temperate rise across the bond so if you use too much adhesive you can easily go from say 0.1mm thick to 0.5mm thick and get 5X the temperature rise over the material.
I don't have #s handy for the T17+, but for the S17 running at 2400W total, that would be about 16.5W per chip. Also I'm estimating the chip to be about 8mm square, so interface area to the heatsink of about 64 mm^2.
Using the 2.88 W/mk adhesive, the rise across the adhesive would be about 22degC if the adhesive is 0.25mm thick.
Using the 19 W/mK solder with the same thickness, the rise would be 3.4 degC.
So your chip with the adhesive would be running around 19 degC hotter than the ones with solder. If you can get your adhesive down to 0.1mm thick, then it comes down to around 8 degC hotter.
For a "normal" thermal adhesive with a thermal conductivity of 1.5W/mK (which I think is what Arctic Silver would be), the 0.25mm thick temp. difference rises to 39 degC, and the 0.1mm thick goes to 16 degC.
T17 and T17+ run fewer chips at higher frequencies than their S counterparts, so the difference would get even more in favor of the solder.
If you have replaced your T17+ heatsink with a one-piece heatsink, you can use thermal grease to connect the chip with the heatsink. Here is an introduction to how to use it: https://youtu.be/nUITyaE_aI4
No you CANNOT!
Thermal grease and pastes are NOT adhesives. Ja they are tacky but they will not hold for very long plus without some sort of mechanical clamping they will eventually start to lift off.
Well, I suppose yes you can as long as the miner is laid on on its side with the heatsinks facing up... In any other orientation gravity is going to do its thing and in fairly short order the sinks are going to slide & drop off...
And then the whole assembly is held to the board with anchors which are soldered down, so its not like you can get away without soldering something to the board. At that point why not just solder the heat sink to the one above it or below it and then use the grease?
I received my thermal epoxy last week which is now cured and will be Age tested tomorrow, Ill let you guys know how it goes. This S17+ board ran PT2 test successfully with heat sinks held down with thermal pads and pressure.
Not one single issue, running solidly cool If not cooler than some of the other chips with original solder on them. I only use epoxy on x17 ASIC that I have lost their copper coating, in other words if I have another problem with this chip Ill will be fine with the fact it will likely not survive being separated from its heat sink.
Im using 8349TFM thermal adhesive from MG Chemicals (https://www.mgchemicals.com). After its been sitting on the shelf for a while You should put it in a cup of steaming hot water for a few minutes before first using it as the B component can crystallize(thicken), actually had to ask customer support about this.