Topic: Tips for powering off your miners (Read 918 times)

That means it should work fine at 100 degrees F.  Like other semiconductors, it must not suffer a hardware failure until temperatures are hundreds of degrees higher.  35 degrees only says software crashes may occur when the room exceeds 100 degrees F.  Which means semiconductors that are even happy and normally at 100 degrees C do not fail.  At those temperatures and due to insufficient design, some semiconductors can cause software crashes.  That does not damage hardware.  Cool semiconductors and it executes same software just fine - for Intel, AMD, FPGA, GPU, or ASIC designs.

Hardware damage never occurs at these near zero temperatures.  These normal operating temperatures include a room at 100 degrees F, semiconductors at 100 degrees C, and no hardware damage.

Hardware damage does not daisy chain into other parts - except when damage is flame.  Flame must never happen in any electronics.  Flame only occurs when a design is overtly defective due to technical ignorance or cost controls.  Flame means a complete design failure in that product and probably others from the same manufacturer.  Flames must never happen no matter what environmental act caused that failure - even a direct lightning strike.

No burn indication is in the picture.  Did it spit flames or just smoke - a major difference that many overlook?  Smoke is acceptable.  Flame means all products from that manufacturer may require a major recall.

Eight signal wire connections exist - six power connections are not visible.  Failure of one capacitor is typically a manufacturing defect - either due to a defective part or due to a cost controller (ie business school graduate) doing the design.  Cost controls actually increase costs as well as reduce product quality.

Turn off hottest hardware without any fan.  It never gets hotter.  Hottest spot - where failures happen - only get cooler when power is removed and no fans spin.  Nothing gets hotter when power is cut.  No special care is necessary for power offs.  A chip at 100 degrees C does not increase to over 300 degrees C when power is off - so no hardware can be damaged.  All hardware has no damage even if the power off is due to the nearby nuclear power plant shutting down.  However unexpected power loss can harm unsaved data.

Most all failures are directly traceable to manufacturing defects.  If hardware fails (when new) at maximum temperature doing maximum work, then the hardware is 100% defective (even though it works fine at lower temperatures).  That is true for all hardware - AMD and Intel, GPUs, and ASICs.  A highest temperature failure today means to expect future failures (months or years later) at lower temperatures. How to keep defective hardware working?  Lower temperature of the incoming air flow.

Most destructive failures are due to manufacturing defects - not due to bad ground, dirty power, or bad PSU.  Those three suspects only cause software crashes - do not damage ASICs.  Fix those defects and all semiconductors and capacitors must be just fine.

In that one failure, informative would be a picture of the actual destroyed C456 capacitor and the adjoining (electrically connected) parts.

Typically I've not gone to an effort of finding root cause on the majority of the blades which died so you could call it speculation but given the environment of where the miners operate is controlled, my best guess for reasons of failure educated only by common factors and practical experience are cheap material, low build standards and possible miss use by previous owners if the product is second hand. I'm also aware that because some ASIC chips are daisy chained designed, any failure to one chip will instantly disable the remaining in that row.


Unfortunately 'properly designed' is open to interpenetration and in my opinion with the accelerated evolution for purpose built mining hardware, build quality is easily sacrificed for faster delivery and lowered expenses.


Pentium chips and mining chips. Apples and Oranges perhaps? But for your information at the moment I only operate Antminer S5 products which I understand operate from 0C to +/-35C. Right now I don't have any way of monitoring temps remotely but I have configured the inbuilt safety feature to power of the board if temps hit the +80C thresh hold.



I don't have the electronic experience to know how my miner sparked and sent blazing glory out the rear fan but out of curiosity I did dismantle the Antminer S3 which managed to 'halt and catch fire' and I found the part marked on the inside board as C456 was where the initial fire incident occurred. Refer this image and see below the 6 power connections and directly to the left. https://farm3.staticflickr.com/2900/14713993833_d214627e56_o.jpg
Here one could have learned and taken away not to buy from Bitmain again but limiting my options in a ever shrinking market isn't an option. Also given I had already experienced chip failures with other manufactures and read about fire related chaos for other miners then this although worrying came to me at no surprise.

  Any conclusion only based in observation is classic junk science.  Best evidence is a dead body.  Exactly what blade part died.  If you do not know that, then nothing was learned from the experience.

Nothing in a properly designed computer can catch fire even if fans are not blowing.  Heat (temperatures that low) do not cause fire.  And do not damage semiconductors.  In fact, Intel hardware simple modifies its operation to create less heat and no crash if overheated.  Heat causes AMD hardware to software crash - without damage.

Using pentiums, hardware was heated to maximum temperatures to learn how hot that CPU could operate - slightly above 300 degrees F.  Hardware damage does not start until temperatures exceed 400 degrees.  Operating a CPU at higher clock speeds only means software crashes occur at lower temperatures - typically above 100 degrees C.  Even that is not hot enough to cause hardware damage or fire.

So again, did you learn from damage?  Or just speculate?  Exactly what part burned - and why?  That study may identify a defective manufacturer.  Defective hardware would only be a symptom.

No problem with ASIC, GPU or CPU mining when shutting it down instantly. Though i would not do SHA256d on gpuss.
If your things surge and catch fire, its most likely improper ground, surge protection or bad PSU. I never, ever had a problem with this whatsoever, ever.

Mining with any other hardware than an ASIC is risky, as they are not designed to run at full capacity 24/7.
As for your blades breaking, it might be because of the environment you are running them in. Otherwise, it's just plain bad luck I think.

Maybe I've had some bad luck in the past, I've definitely had a few too many blades randomly die and I even had one miner catch on fire!

Also I remember reading a long time ago some one was going to the effort of using ssh on each device and running sudo poweroff

This never worked for me as I typically never had a permanent management computer near to any of the miners when it came to a decom.

Does anyone else have tips to share perhaps someone GPU mining?

I'm pretty sure your chips won't heat up if you just flip the switch, so yeah just flip the switch. Its sensible to "properly" power down a computer to prevent lost of data and some shutdown protocols.

But for miners i just flip the switch. I very much doubt there is enough power held in the miner to cause any damage as it dissipate.

OK just checked and you are right, after 4 Minutes or so after loss of Internet an S5 does stop hashing. However I will still just cut the power when I want to shut down, because there is no more heat being generated by the ASIC's when power is cut and Life is too short.  Let's see what others have to say?

Rich

I've had experience with the S5 (and hopefully the S7 soon) and found disconnecting the Ethernet worked as with it's predecessors.
Typically after disconnecting the antminer they will continue for some time running at full speed until the target pool is no longer recognised as responsive triggering the beeper.

Understand what you are doing & why. However I just throw the mains switch. The ASIC's will not be generating any more heat so I suspect no harm will be done.

Also on the S5 & I suspect S7 even when you remove the Ethernet the chips keep hashing, so no benefit.


Rich

Hello and welcome.

I thought I would start a little thread to share tips on how users power down their miners, this could be your ASIC, FPGA or GPU.

I've personally had a few miners over the years. I got playing with mining on GPU's back in late 2012 and jumped all over buying ASIC's as they became available.

Through that time I've learn't to stay in front you should buy, mine and sell ASAP.

But when it came to shipping the miners off I was unsure of the best strategy to safely power off the devices so the chip wouldn't burn after loosing power to the fan.

Anyway since this occurred to me I've always disconnected the Ethernet to stop work from being pushed to the miner, waited 10 minutes or until I felt the heat had dissipated enough then pulled the power plug.

What are others doing to safely power off their miners?