There are also a bunch of questions about what thermal interface materials are acceptable and which ones aren't.
I first want to say that it's your unit, you are free to do with it what you'd like. Prior to shipment we are testing each unit long enough for it to reach thermal equilibrium for probably 10-30 minutes. The time is still up in the air depending on how efficient the test/programming techs and setup are. We're using some fancy carbon fiber pads and Arctic Cooling Mono Plus coolers if anyone's interested. They won't be sent if they don't function. After you get it, you can use peanut butter and hot fudge if you really want to. All we can do is provide suggestions.
TIM Tidbit #1: Your best interface is going to be using a grease like Arctic Silver. You'll get a much smaller gap between the die and the cooler, and performance will be much better than any other interface.
TIM Tidbit #2: The chips might not all be at the same height. In testing with thermal grease ChipGeek and I have both had single chips that sat appreciably lower than the others. In my case, it was enough to cause that one chip to be 20C higher that the others when voltage was applied but it wasn't hashing, so it was only consuming a watt or two. That chip still stayed about 20-25C hotter at full load, likely because the other chips were well cooled and it was able to dump most of its heat into the PCB copper layers. Still, it is not an ideal situation and if you had a couple chips that were low (or worse, two that were higher and prevented good contact on the other six) you might have serious problems. If you want to try thermal grease I can't stop you, but it might not work well and while we will attempt to throttle based on temperature if there is a really bad contact the ASIC could overheat.
TM Tidbit #3: Thermal interface pads are bad. Even the good ones are bad, but that's the hand we're dealt. The
0.5mm thick 17W/mK is a very good thermal pad, but it still has a thermal resistance of 0.4(°C*cm^2/W) at about 50PSI. Since the BFL ASIC die is ~8mmx8mm, that means that with good pressure the thermal resistance of each chip is about 0.63°C/W. If you're dissipating 10W per ASIC, you'll have a 6.3°C temperature rise just in the thermal pad. The
0.5mm thick 11W/mK Fujipoly stuff is around 0.9(°C*cm^2/W) at the same conditions, meaning a 1.4°C/W temperature rise. Start going to the 6W/mK and 3W/mK stuff, and you'll find you're dumping a LOT of heat into the board. Chip Geek had our test board with no asics hooked up to a large programmable dummy load and was pulling 100A from it, he said while it warmed above ambient it stayed quite cool to the touch. If I use a 3W/mK thermal pad and pull similar power from it, the power supply components and PCB can actually get painful to the touch without good airflow over the board. Good thermal pads are bad, but bad thermal pads are
really bad. You can mitigate a lot by using more airflow (just look at all the little heatsinks on the BFL Singles and how much air over the board they push to keep things cool), but just using a really good thermal pad to start with is best.
TIM Tidbit #4: Thermal pads might be bad, but thermal adhesive tape is terrible. Please don't use it, you might as well just leave the heatsink off. It works in low power applications or where you have a lot of surface area, but not at these power densities. The
3M 8810 0.25mm thick tape you see mounted to the bottom of a lot of press-on heatsinks has a manufacturer rated thermal resistance of 0.90(°C*in^2/W), or 5.8(°C*cm^2/W). That's 6.5 times worse than the Fujipoly 11W/mK pad that's twice as thick. They don't even work nearly as well as a gap filler. Please please please don't use thermal tape.
Overall, you should just use a good thermal pad, and use a good backplate so you can tighten the heatsink without flexing the board too much. If the units coming out of the assembly house are good (and we're hoping they are) you might be able to get away with grease and it will work really good. If you get one that has a low chip and you use grease instead of a pad and blow it up, that's on you. If you use adhesive tape instead of a pad and blow it up, that's really all on you.
