Experimenting with 1 board t21 miner at 120 volts between 500 and 1000 watts

I've got 4 T21 miners that I modified to run on 220-277 volts single phase. They were having random failures, so I picked up 2 S21 XPs to swap out the two worst T21s. Now I'm trying to build six 1 board miners running at 120 volts using the six boards. However, I'm facing an intermittent chip overheating problem. I suspect that the OEM thermal paste might be the culprit here. On the bright side, the t21 heatsinks can be easily bolted on. I'll share some photos and updates on the build and repasting process. Feel free to share your thoughts.

spacer I will photo a removed heatsink in a bit. I can tell you the paste is all over and has oozed away from where it should be and the heatsink screws are not as tight as they should be. I get paste tomorrow and will the boards will repasted hoping to fix this issue. I suspect the paste job is failing as they have run for six months and oozing occurred Look below temps are solid then boom it will say they are over heating . Loosing contact with chip to heatsink can cause this issue.

spacer paste is oozed out all over close up of paste ooze and by just picking up the PCB board with a tiny shake and slide look at oozed dried paste

spacer:  more to come so I repasted and temps are about 5c lower. I can run the rig at 16th-17th using chip tuning and it pulls under 325 watts I will check this in an hour

spacer I can crank it to 48th and about 965 watts. If you use the Loki chip you can chip tune but not voltage tune as the psu is the old school apw3++ putting out a constant 12.5 volts. I did find a different psu that does 13.5 volts which is able to run the board at higher freq the oem apw3 will not go to 50th but the one I have now is doing 48 th before the pasting. So with the new paste the numbers below are decent. I set it for less than 48th It is at 40-42 th freq of 300 and fans at 70%. This kind of makes me think that a lot of weird drop out issues on s19 and s21 are all about paste that is drying and filling. the chips lose contact with the heat sing due to fan vibration and go from normal temp to high temp pretty much instantly. Shutting down the miner. While that may seem good to know it is not if you want a lot of air cooled miners. 1 board used all of this paste https://www.amazon.com/dp/B0DNQ936MF?ref=ppx_yo2ov_dt_b_fed_asin_title that is about 9 bucks and first time I did it maybe 4 hours time. So the entire miner may take 27 dollars in paste and 8-12 hours time you can buy a mesh which will allow faster paste time .  Maybe 6 hours. Think if you have a container of these the repast would be 300 x 27 = 8100 retail (maybe 2000 retail for the paste) and 6 x 300 = 1800 hours labor. The paste job was needed in 6 or 7 months mining . yet another maintenence cost to mine. I have 4 t21 units. If I do them all I will spend over 100 on paste and maybe 24 hours time. Yet again more work.

https://www.amazon.com/gp/product/B07PHT3LWC/ref=ppx_yo_dt_b_asin_title_o00_s00?ie=UTF8&th=1 I ordered a cheaper paste 100g for 14 bucks Vs 16g for 18 bucks. It seems a board needs about 11g if you do both sides. This will save a lot if you try this. After getting these results on the t21 I am convinced my issue was the paste failed after 6-8 months. I am going to fully paste the modded t21 and run as a three board miner in a week or so. This may be good and bad news at the same time.

did the second board with the cheaper paste. I can clean and paste it in under 2 hours . uses 1 tube of cheaper paste  around 4 dollars. for a 25g tube. so 6 hours labor and 12 dollars to do 1 t21 miner. so a 1 board Loki chip 120volt board using the pricey but good epic controller allows 20th to 50th and 20% fan to 100% fan freq of 150MHz to 370MHz this is with the higher end 1600 watt 13.5 volt psu I found. not the stock apw3++ which can do 12.5 volts which means you have to go a bit lower in freq It is a nice piece of gear 50t-51th for around 1000 watts is good. and being able to drop it to 500 watts is also nice.