Bitcointalk world! Need some help understanding my burning question about S9 temperatures (pun intended). Done some exploring on the board and only found pieces on info here & there. Any info / insight would be greatly appreciated.
I have built and air cooled box for my 3 S9s. Going to be doing some minor tweaks to move evenly distribute cool air, but here are my current miner stats:
Chip1 Chip2
Chain 6 - 73 85
Chain 7 - 81 89
Chain 8 - 78 92
Chip1 Chip2
Chain 6 - 79 86
Chain 7 - 76 84
Chain 8 - 76 84
Chip1 Chip2
Chain 6 - 67 84
Chain 7 - 64 73
Chain 8 - 61 75
Should I be worried about the temps on miners 1 & 2?
Also, when looking at the kernel log it seems that 2 of my miners have more than one temp monitor. Looks like they are on chip62 and chip32 - what should I be looking for here. In the case of miner #1 above, it looks like what is displayed on the miner status page is chip62. And temps on chip62 always seem to be higher than chip32.
Kernel log from miner #1
do read temp on Chain[5]
Chain[5] Chip[62] TempTypeID=55 middle offset=24
Chain[5] Chip[62] local Temp=76
Chain[5] Chip[62] middle Temp=85
Chain[5] Chip[32] TempTypeID=55 middle offset=23
Chain[5] Chip[32] local Temp=58
Chain[5] Chip[32] middle Temp=73
Done read temp on Chain[5]
do read temp on Chain[6]
Chain[6] Chip[62] TempTypeID=55 middle offset=26
Chain[6] Chip[62] local Temp=78
Chain[6] Chip[62] middle Temp=89
Chain[6] Chip[32] TempTypeID=55 middle offset=24
Chain[6] Chip[32] local Temp=60
Chain[6] Chip[32] middle Temp=81
Done read temp on Chain[6]
do read temp on Chain[7]
Chain[7] Chip[62] TempTypeID=55 middle offset=26
Chain[7] Chip[62] local Temp=80
Chain[7] Chip[62] middle Temp=93
Chain[7] Chip[32] TempTypeID=55 middle offset=24
Chain[7] Chip[32] local Temp=56
Chain[7] Chip[32] middle Temp=77
Done read temp on Chain[7]
Thanks!!!
im not sure why this isnt being answered when people ask but the temps are fine.
the reason the temps are different are:
1. they mix and match boards to hit the frequencies needed to hash at whatever speeds they listed the miner to have plus or minus 10%. this means they can add under performing boards that run at slower speeds and 1 awesomely performing board that can handle much faster speeds. they find the 3 boards...add their numbers up....slap a sticker on the tupe and send it off to the customer. not every asic/board hashes at the same speed. some batches of chips are worse than others and some chips in a batch can handle things better. so since the autotune firmware tunes the boards to the max speed they can handle to find the sweat spot some boards may hash at 500mhz and others will be 600+ and that 600+ board will be warmer than the 500 board because more power is going through it.
2. the heatsinks are perfectly set. look at other pics of miners....you will see alot of heatsinks that have fins that dont line up with the heatsinks next to it for whatever reason. maybe it got bumped or moved before the thermal glue set and hardened....who knows.
3. the position of the board matters alot too. the fan blade part of the fan moves the most air. that means at the far end of the fan housing and the center of the fan moves the least amount of air over the heatsinks. this is the same with each and every single fan ever produced lol. anyway since the airflow is lower in those spots the temps are warmer which reflects the differences shown in the status page. if there was enough room for another fan this would probably change quite a bit but miners want the smallest solution possible so they can pack more miners into a small space so you have to compromise a little.
4. the thermal glue is shit lol. if they had large 1 piece heatsinks like they did on the old s1's the fans would be quieter and the temps would be lower because there will be much more surface area than they have now (between the heatsinks and 1 solid piece spreads the heat out from hot spots to cold spots keeping the temps about the same). they chose to glue the heatsinks down so not much we can do there unfortunately but the glue they use is probably not the best solution available. the amount used also makes a difference....since we cant see under the chip we dont know if the glue is covering the whole chip or just a dot in the center or if they pushed the heatsink down or just let it rest ontop of the dab of glue filling all the space on the heatsink and chip. there are alot of factors effecting temps with just this process alone.
