S7 Fire

[pimpin]

I unplugged two of my S7's today and they caught fire a few hours later?  I had them connected to Delta PSU's 2400W each.  I  unplugged the the 6 pin plug from the main board on each machine and left the rest plugged in.  The machines appeared to shut down fine.  Would the chip boards continue to heat even though the main board was unplugged?  This is the only explanation I can think of.  Would greatly appreciate any input as this was a nasty fire.  I keep them on steel racking so the fire started and stopped there BUT obviously I don't want this to happen again.  Any help would be appreciated.

[sidehack]

The 54-chip boards (4.86TH models) would still be live at full voltage; the 45-chip (4.73TH) models would be operating at a hardware-default lower voltage. With voltage still applied to the chips, even without work, they're still going to be pulling some power. For reference, my 2Pac at stock voltage pulls 360mA at plug-in (because the chips default to 200MHz core clock even without work) compared to about 800mA at 100MHz.

Without fans, whatever heat is building up inside (still probably less than 150W) wouldn't really have anywhere to go. With the extreme solidity of cross-section in an S7 due to the heatsinks, even convection wouldn't clear much heat. So though I haven't heard of anyone's S7s burning up like this (though I've also never heard of someone unplugging just the controller and leaving), I don't believe it's impossible. Bitmain probably specifically says not to do that, somewhere in the fine print.

With 4.73TH miners, the mainboard gets plugged in first because the boards' main regulators use the 3.3V line supplied by the controller to power on the parts used to set the output voltage. At least that's part of the reason; there's probably other things having to do with chip detection and stuff too, as well as making sure fans are running before heating up the hashboards.

[pimpin]

Ouch... Per Bitmain's explicit instructions (when using multiple PSU's but applies here as well) is that the main board is always supposed to be powered on first followed by the hash board power. That strongly implies that powering only the hash boards is verboten....

     This was two S7's with two separate PSU's which caught fire at the same time.  To me it looks like the boards over heated then the wires melted which caused the PSU's to short and the whole thing melted down.  I wasn't however aware of this main board issue obviously or I wouldn't have done it.....  Thank-you for the input!  Greatly appreciated.

[tim-bc]

the main board is always supposed to be powered on first

Usually the advice is (when there are multiple PSUs) to make sure the controller board is powered on last, is it different for the S7? Even the user manual for the S7 seems to indicate you need to power the controller last.

https://bitcoin.stackexchange.com/questions/56626/antminer-s7-dual-power-supply

[NotFuzzyWarm]

I checked and stand corrected. Yes when using multiple PSU's the one feeding controller turns on last.
Finally got into the Bitmain site to dl the s7 manual to confirm it.

Never really needed the info as I've always used single server PSU's for my farm. Good catch and Merit for ya.

That then mainly leads to the s7's propensity to overheat/burn when coms go down as the culprit. Removing power from the controller could amount to much the same thing as far as the hash boards are concerned. That and as Sidehack said - no controller power means no fans and even idling the hash boards pull a fair bit of power...

[fanatic26]

The documentation for testing s7 boards from bitmain clearly states never leave a hashing board plugged in for over 5 minutes even when at idle or testing, and I can confirm they will melt right through the table.

I accidentally did it during my bitmain training....whoops!

[lightfoot]

Ouch. I've been gone awhile but never thought to mention that when S7 and S9 boards are powered up the chip voltage regulator goes to a default which the chips themselves consume and turn into heat. Heat which sits on the sinks and does not dissipate without forced air.

This is more dangerous on the S9's: I was testing one of my test boards outside of the box and plugged into a supply (measuring signals at rest) with no controller. I walked away from it and forgot to power down. When I came back 30 minutes later the heat sinks fell off the bottom of the board (the board was at solder melting temps) and the thing scorched my work bench. Had I gone away longer there might have been a fire.

Moral: Respect these things.