The default values are correct. Bitmain changed the way S9 reports temperatures around 2017, basically subtracting 15°. The oldest firmware reports temperatures exactly the same as Braiins OS: as the sensors send it.
What warning? If it reaches hot fans go 100%, if it reaches Dangerous hashing stops. You can change them if you wish.
In the Bitmain specs the max chip temp is written as 135°C. Someone mentioned that factory fw stops hashing at 130°C. Braiins OS is far more conservative by default.
The default requires 1 fan plugged, if won't hash until you enable "Immersion mode" which means minimum fans required: 0, and nothing else.
Thanks for the explanations!
Very interesting to watch the auto-tuning in progress. Takes quite a while!
I notice particularly that one of the 2 connected hashboards (1 of the 3 is cactus and not connected) always had a higher chip temperature than the other one, but after about 8hrs of tuning the chip temperatures are now even.
After tuning is complete, I'm getting 11.2TH/s, that would be 11.2TH/s / 2 ×3 = 16.8TH/s from a 14.5TH/s spec miner. A 15.8% increase.
Estimated efficiency as per BraiinsOS+ is 96W/THs, and the measured efficiency at the wall (including PSU and 8'' Phresh fan and ethernet hub thingamabob) is 1062W/11.2THs = 94.8J/THs.
Regarding the 'missing warning', I meant this: I powered up the (2/3rd) S9J without it's stock fans, with the brand new BrainsOS+ SD card and cooled by a Phresh 8'' ducted fan (with manual control dial to adjust the speed). BraiinsOS+ had no way of knowing that a big manually controlled fan was taking care of cooling, and the 'Enable Immersion Mode' was not ticked, but it did not warn something might be wrong. It did however show Fan Monitor 0% 0% 0% 0%
I have since ticked the 'Enable Immersion Cooling' box and now the non-existent fans are shown in Fan Monitor as '100% 100% 100% 100%'.
Questions:
1) Does it matter if I manually adjust the fan speed during the tuning process, or if the air intake temperature fluctuates? It seemed that setting the 'Enable Dynamic Power Scaling' caused trouble. Because I did not realise that cooling requirements do fluctuate so much during tuning, and because I was trying to keep the noise low, the power reduction got triggered a few times by reaching the 'Hot 100C' temperature. Then I turned 'Dynamic Power Scaling' off and tuning appeared to progress more orderly.
2) Can the results of the auto-tuning be stored and then implemented quickly after a restart, or must it go through the whole tuning process every time? I ask because I plan to run comparatively inefficient miners like this one on 'free/stranded' electricity from solar panels. The miners will need to turn on and off frequently, or maybe ramp up and down their power use depending on SOC of solar battery, strength of solar irradiation and grid feed-in tarrifs, or when max allowed feed-in power is reached.
3) Will it make a difference to the tuning results (or to the tuning time requirements) what 'PSU Power Limit (for all hashboards)' I set? I am using a HELA 2050 Platinum PSU (up to 2KW) and I noticed that the auto-tuning starts at different power levels and frequencies when I set different PSU Power Limits.
4) What is the best / fastest / most efficient way to find the most efficient J/THs setting?
5) What is the best way to find the lowest possible power setting for a miner?