Bitcoin Forum

Okay,

So someone posted this under one of my threads and said there was massive overheating due to a kernel log error that basically caused his R4 to combust into flames:

https://bitcointalk.org/index.php?topic=1769525.new#new

"Re: How to fix this R4 log?? [Fatal Error: DownOneChipFreqOneStep Chain[-1] has...]
February 05, 2017, 11:47:24 AM
Reply with quote  #5
If anyone else gets this error... whatever you do, don't leave your miner on in this state for long, I've had 2 boards burn themselves up literally.  The chips are heated highly, and the top right VRM chip started smoking in one."

This has me a bit worried about my 18-ish R4's would do the same thing at some point; I wouldn't be paying attention, and basically burn everything down.

Assuming electrical circuits (at wall and in the building) are wired OK, and surge/internal surge circuit breakers @15A are OK, and the main panel 20A is OK, and using EVGA-1000 G3's ATX continuous (w/ over & under volt) are OK....


Has anybody run into issues with ANY Antminer (S3, S5, S7, S9, R4 - though I'm only running R4's) burning into flames?? Or heard of that happening? Mostly focused on the newer R4's though...

Luckily I was around when the VRM started smoking in 2 of my miners.  For these, the hashboard reported 0 asic and I was trying to get them back to life by running only the bad hashboard (thus getting bad index error). A third hashboard that was reporting 55 asic has now failed in short circuit, luckily my PSU short circuit protection kicked in, but if using a shitty PSU I can see this resulting in a dangerous situation (either your PSU is damaged or something is going on fire..).

Unless you really fuck something up or there's something wrong with your miner, then there should not be anything bad that happens with your miner. Worst I've had is a power surge that knocked out one of my S1s (ALWAYS KEEP PSUS ON A POWER STRIP). Besides that, all I can think of is a cheap or just plain bad power supply.

Grounded power strips/surge protectors with 18x R4's and EVGA-1000 G3 PSU's - Not the most expensive PSU's, but certainly well-built...

The older 54 chip S7s were known to burn up when they lost internet connectivity and were on one of the older firmwares (fw dated 2015) but as far as the newer ones go I have not seen any catch on fire.

So... this just happened:
https://i.imgur.com/cOz44pC.jpg

Not quite sure what the cause is yet but the other hash board is still working fine..  The data cable was also burnt.

Please use caution and place these things away from any combustable materials.

So I replaced the burnt PCI connector, power and data cables and by some miracle the hashing board still works.

Does anyone know what would cause something like this?

Did you lose internet connection?
That is one situation when burns occur.

Wow, yes, looking at the mining history, I lost internet for 4 hours on March 28th, this is exactly when the miner stopped functioning.  I can't believe this known issue hasn't been resolved since forever.

I haven't had a miner fire since I stopped running Butterfly Labs gear.

From the view of an Engineer well versed in power systems, seeing just that one PCIe connector burnt is highly suspicious.
Either:
A) The board was powered by just that one connector. If so, of course it fried...
B) Very bad sockets in the PCIe cable feeding it. I do see that there is no solder fillet around the pins where they join the board. I will assume that there is good solder on the backside. Lack of a bead on top is not good but not conclusive.

One problem with "B" -- if all 3 connectors were being used, the other 2 should have picked up the slack resulting from the poor connections. Remember: All 3 of the PCIe connectors are in parallel. In no way could there have been enough current through them (bad connections) to burn it and the 2 good connectors show no signs of overload at all eg, turning brown.

Hi, all 3 pci-e connectors were plugged in and both hashboards were operational until recently.  One interesting thing is the data socket and cable was also burnt although not as bad.. so what ever happened also happened to the data cable.

Earth failed on all 6 conductors of the other 2 PCIe sockets. The earths on the data connector then started passing current because 3 * "normal" remaining cables and connectors heated up (higher resistance).

Cheap PCIe cables, undersized PCIe cables, would highly doubt it's a board fault.

Don't use cheap PCIe cables! :)

^^ Yes, that would explain it.
Query: The PSU that was tied to it, was it new or has it been used on other miners? Reason asked is because sockets/pins (mainly the sockets) wear out rather fast - not from time but from being plugged/unplugged. Even top-grade gold or silver plated contacts cannot not be plugged/unplugged more than a couple dozen times before they start showing increased contact resistance.