Bitcoin Forum

A recent post by jeppe, a bitcoin forum member and miner, reminded me of the potential hazards of bitcoin mining.  

His post is here https://bitcointalk.org/index.php?topic=517976.0

Bitcoin miners do get quite warm/hot when mining and not everyone who mines is knowledgeable about electricity or its components.  My personal knowledge of electricity does not go too far beyond calculating W=V x A and basic knowledge of circuits.  There are plenty of mining setup pics created by amateurs that I thought were neat and interesting but after assessing the same pictures with an eye for safety, they seemed ridiculously dangerous.  I hope no one lost their homes to fires from bitcoin mining.  Since bitcoin miners are now using serious wattage, I think the point in bitcoin mining to talk about safety has arrived.  If you have any advises or stories of mining safety, please post them here to promote safer mining.

Always wear safety goggles!  ::)

If you are going to mine, and you have not created a map of the circuits in your home, prepare to burn it down. Map, as in every outlet, non-moveable load (lights, fans, bathroom blowers, garbage disposals (bonus, those are usually branch circuits, extra current for free!)) and static loads. If you have a TV on the same circuit, and you plan to use it, take it into account on calculations.

If you are going to mine, and daisy chain a half dozen 750W supplies on a "surge protector," you're going to have a bad time. Pay attention to W=VxA and keep things in spec to the map you've already made. "surge protectors" are not bad, but misuse is.

On the other hand, pay attention to the other side of your power supplies. W=VxA still applies! Don't try and run 7990's on daisy chained 20ga molex connectors, as an example. Use this, if you don't know gauges yet.

http://www.bulkwire.com/wireresistance.asp

Understand how to properly purchase a power supply (multi-rail vs single, 12V wattage compared to total, outputs available, and how they map to separate rails etc..) Realize that properly loading an outlet and power supply does not mean using it to 100% capacity, try and stay within 80% on both. 15A 110V outlet? No, it is a 12A circuit! 1000W power supply? I'd consider it 800W, but that's just me.

Of course you should use quality cables between the wall and power supply. 14G is common, but 12G is best. Just because it's a fat cable, doesn't mean it's high gauge. Verify, and then check, check, and check again once mining actually starts. Be especially mindful of the plugs on either end, as those will likely be the failure point that gets hot. Even if it's brand name, watch it closely for some time. I really like non-contact IR thermometers for checking wiring, plus pegging cold/hot GPUs.

Understand that a power supply is capable of providing too much power sometimes, and that can really screw your calculations up due to efficiency curves. You won't really notice this without measuring, a kill-o-watt is a very wise investment...

If you burn your home, or god forbid apartment complex, down while mining there will likely be a sea of legal issues for you. Especially if you somehow daisy chained a bunch of supplies together, blatantly ignoring electrical code. Ignorance is no excuse, as with most things.

Be careful everyone!

I currently have a Terraminer IV running at full power stepping (setting 9) which I believe pulls ~2.2-2.4kW from the wall. Both PSUs are plugged directly into separate outlets and not daisy chained or connected to power boards. Where I live has ~220-240V power however so it'll be using ~10A per outlet if my math is correct. Should I consider investing in some kind of UPS to regulate the power and make sure it's getting 'clean' electricity or would this be a waste of money in my situation? My knowledge of electrical engineering doesn't really extend past W=V*A.

A waste of money, nearly all ATX power supplies made use what's called active power factor correction. I believe this came from wikipedia...


If your power supply is gold rated, or even 80+ rated, chances are it uses active power factor correction. Basically, a switching power supply that is able to adjust to nearly any input voltage. At any given moment in time, the power supply switches it's total input current depending on the voltage. You can hear this "switching" if you listen closely sometimes  8)

Ripples in input voltage are not going to bother any good supply, but they will bother an active PFC supply less.

You may have seen passive PFC supplies, with the little red 110-220 switch near the line voltage socket. Those guys are not as efficient, so they are not as popular any more.

if anything has a fan in it, it's helpful to blow out the dust bunnies that build up in the fan grilles, ducting, and in and around heat sinks with dry, compressed air.  the accumulation of dust not only restricts airflow and increases temps, it is also flammable.  too much dust can and has caused fires.  

Active Power Factor Correction does not directly reduce power consumption. Most old/cheap power supplies appear to use passive power factor correction. (They get a power factor of like .67 instead of 0.5). Where PFC reduces power usage is in transmission losses.

I have an 80+  Power supply that does not appear to use APFC. It is also auto-switching. In general, switch-mode power supplies are tolerant of power (voltage) sags (but will draw more current to get the same power). The reason is that they first convert the AC to high voltage DC, then use an inverter (at about 20kHz instead of 50/60 Hz) to get the low voltages.

GPUs can generate a lot of heat. And if there isn't a way to get rid of this heat, it will build in the residence and make it difficult to mine.

The fire safety aspect is also of concern.

Cooling is important for several reasons. The main reason is obvious: to keep things from overheating. But there's a second, less known reason.

In electrical circuits such as computers, mining hardware, etc... the voltage applied is constant, but the current depends on the resistance of the device, as illustrated by Ohm's Law: I = U / R (I = current in Ampere, U is voltage in Volts, R is resistance in Ohm). The resistance is typically not constant, but depends on the temperature: a higher temperature means a lower resistance. This behaviour is specific for semiconductors, regular metals exhibit the inverse relation between resistance and temperature.

So: Higher temperature means lower resistance, which means a higher current. And since the power consumption (and therefore heat production) is given by P = U * I (P = power in Watt), an increase in current means an increase in power consumption and heat production (which in turn leads to a further decrease in resistance, etc... Fortunately this behaviour is asymptotic and the effect doesn't blow up).

Excellent lesson in thermal runaway, good job :)
 
There is also two other important factors of temperature:
1. The lifetime of the component (especially the passive components) is significantly reduced by each degree the temperature goes up.
2. The transition time and propagation delay transistors in the silicon wafer of the GPU is also thermo-dependend, they are slower with higher temperature, which can caused timing errors in the logic, should not be a problem is long as the die temperature is within specs. 

This
https://web.multco.us/sites/default/files/imagecache/large/images/dust_off_canned_air.jpg

Don't hold a duster can upside down, especially if you are spraying hot hardware. I'd imagine it would be easy to put stress fractures in heatpipes and heatsinks from the thermal shock.

That stuff is flammable, too. Hardware has the potential to set it off, ask how I know  ;D

I recommend leaving hardware off and room temperature when using compressed air.

phillipsjk, you pretty much summed up what I said from what I can see. The passive PFC supply will loose efficiency when the line voltage is dirty, since it does not have the extra switching circuitry. From my understanding, the point of the APFC was a boost converter between the input capacitors and rectifier. Because of the input capacitors and the boost converter, the "first-stage" switching supply appears totally resistive, or nearly perfect power factor. To do so, it needs to have a pretty close eye on the input voltage and waveform, so since it's always adjusting, trying to match, it ought to be more efficient adjusting right along with "dirty" input voltage, right?

My logic is the "two-stage" SMPS should be sufficiently clean, no matter the input voltage, being essentially "double-filtered" by nature. Even a passive supply will be fine, still, though not as efficient as the active PFC supply in the same conditions. Basically, any "line voltage" correctors are not a worthwhile investment solely for the protection of computer hardware.

You seem to know a bit more than me in this area though!

So if I can the the Ohm with a multimeter then I can figure out Amps with I=U/R, correct?  Have not seen or thought about that formula since college...

I think my main objection was your claim that most power supplies (with automatic voltage switching) use APFC.

When I bought my "green" power supply (80%+ efficiency rating), I assumed, without checking, that it had APFC. My Watt-meter tells me it uses passive Power factor correction. (And the label does not make any PFC claims.) The PSU out of my old dell Optiplex has very good power factor: They may have used APFC to remove inductors to save space.

As I said, bad power factor does not directly effect the efficiency of the device. You draw the same wattage with good and bad power factor. The difficulty is that bad power factor requires more current to transmit the same amount of power (sometimes called "apparent power"). If the power  rating is measured in VA (Volt-Amps), that is an "apparent power" rating. Real power is measured in Watts.

With a resistive load, those will be the same. Power factor is a measure of how different those two numbers are. BTW, the Wikipedia article helped my understanding too. That is where I learned all those PSUs with crappy 0.67 power factor actually do have (passive (in the form of inductors)) Power factor correction. I know my vacuum cleaner has (probably passive in the form of a capacitor) power factor correction. Power factor is more important for large loads like vacuums or mining farms.


Simple multimeters don't work well on non-linear components like semiconductors, capacitors, and inductors.

Typically what is done to measure current (in the range of milliamps-amps) is to insert a 0.1 Ohm sensing resistor (keeping in mind that at 10 amps, that would drop 1V). For higher currents, you may want to look into clamp current meters or simply measuring the temperature rise to monitor power usage. Edit: my watt-meter just measures at the wall.

If you're using surge protectors, read what kinda load they were meant to take and don't go above 80% of that.  You may need to buy a few more, but at least you won't blow shit up.  If you're using ones that you had laying around, look and make sure the protected light is on.  If not, it's not gonna do a damn thing if (when) you get hit by a surge.

"My personal knowledge of electricity does not go too far beyond calculating W=V x A" haha it's a very good description

Do you really think the point in bitcoin mining to talk about safety has arrived? In my opinion there is absolutely nothing to worry about

I hope you plug all your miners into one circuit and the wiring catches fire then.

So many noob trolls nowadays...

Electrical safety is paramount even if you only run a small rig.

1. Test your electrical source and ensure your outlets are rated to handle it. On home circuits this can be tricky, as each leg of the grid can only handle so much power. Too much load on one outlet can cause a fire!

2. Use a server rack grade (fused) power distribution hub. Some of these are even smart that can track your power usage with a web browser, and can have fault-tolerance built in to switch power sources. These can be picked up for around $100 for the more basic ones.

3. Deploy an uninteruptable power supply (UPS) that is rated to supply your farm for at least 5 minutes. If the power goes out, you have some time to properly shut down the farm. Many of these support USB connections to automatically initiate a soft powerdown if the battery starts running low, if you have a computer of some kind running your miners.

4. Do your homework on local fire codes! At least in my jurisdiction, the mandate is for server racks to be at least 2 feet from any walls. You want a buffer between your miners and any walls or furniture so any exploding gear won't start a fire. Ideally your miners are not on carpeted surfaces either. Ensure your miners have adequate an open airflow.

5. For larger farms, you really should be running them in a space with a fire suppression system rated for electrical fires.

6. Always keep a fire extinguisher (rated for electrical!) nearby and visible. That is good advice for any household anyway. Many municipalities will recharge extinguishers for free from the fire department.