What is the problem with cheap power supplies?

[aknnig]

Hi.

I very often read that it is wise to invest in very good power supply, due to their efficiency, durability, stability... Dogie had a nice review about good power supplies, https://bitcointalk.org/index.php?topic=486121

However, to introduce bitcoin mining to the masses, sometimes we must consider prices, noise and form factor (shape) over the other points mentioned above.

Is it possible to use cheap 12V DC power supplies like this one http://yqunique.com/products/s-201-singe-output-power-supply-ID584.html#.Vv0nnNWLSVM  or other similar instead of ATX alike power supplies?

Are their specifications too far away from standard ATX specifications that typical 12V ASIC miners will not work, or even worse they will burn down the miner? Here I have assumed that the power supply can give the necessary output.

If so, what is the minimum requirements a DC power supply must have to be able to drive the miners?

Comments from you will be highly appreciated.

Best regards,
aknnig

[1715157700]

1715157700

[1715157700]

1715157700

[1715157700]

1715157700

[1715157700]

1715157700

[xhomerx10]

That particular supply is only capable of 16.5 Amps which isn't enough.   Also, the price isn't even mentioned... what makes you think they are cheap?

[Finksy]

The biggest problem with that one is efficiency of 80%, that is less than Bronze. And as mentioned above, it is only a 200W power supply.  If you pay for electricity the savings in electricity are obvious from efficiency, but even with free electricity going to a 90% efficient PSU (Gold rated) allows you to host 10% more miners.  

If you want to save money over ATX power supplies, look into server PSU's.  You can either hard-wire them yourself (you can find many resources here and on RC forums), or find one of many suppliers on here that sell breakout boards to adapt to the common (and cheap) server PSU's available.  HolyScott, J4bberwock, sidehack, quakefiend and pmorici on this forum all carry breakout boards for server PSU's in the 800-2000W range, I sell breakout boards for 2880W-4000W server PSU's.

If you want to do it yourself and save the most amount of money, browse through this thread on RCgroup: http://www.rcgroups.com/forums/showthread.php?t=1005309  It has 200+ pages of configurable 12V power supplies, in many cases with pinout and pictures.

Edit: To answer your question more specifically, ATX is simply a standard for motherboards, including the PSU and how it supplies power to the motherboard via 24-pin connector and 4/8-pin EPS (CPU) cable. The ATX form factor has no real relevance to bitcoin mining at all, since we just use jumpers to power on the PSU anyways, and only make use of the PCIe cables on them.  Also -as you have surely noticed- as you get into larger ATX PSU's, you get into a smaller, performance-oriented market (read: more mark-up by retailers).  With the power demands of bitcoin miners seemingly always going up, server PSU's make more and more sense now.  Even Bitmain began offering their own in-house designed/marketed server-style PSU for the S5+ and S7.

[sidehack]

Things to worry about from cheap PSUs:

Low conversion efficiency - 500W from an 80% supply costs 1.67KWh more per day to run than 500W from a 90% supply. At ten cent power that's five bucks a month. You may not care about five bucks a month, but that's sixty bucks a year. If you expect the PSU to be around for more than a year, spend sixty bucks more and get a higher efficiency one.
Additionally, if you require air conditioning to cool the location your miner is in, that's an extra about 240BTU you need to clear, which you then have to pay for in power costs again.

Low longevity - If the PSU is cheap, it was probably underbuilt or made with cheap parts. It probably shouldn't be run at close to capacity for very long. Some good PSUs will run at 110% rated load for years, and some cheap PSUs will fail at 90% rated load for more than a couple hours. If you're going to be pushing a cheap PSU, you need to make sure it's way overrated for the continuous load you expect to draw - which will invariably add to the price.

Poor PFC - if you run a high load on a PSU with bad Power Factor Correction, you'll be putting a lot of nonlinear current loads on your mains wiring. This can cause some heat strain to your wiring, and undue transmission losses that your power provider has to eat and they won't like you very much. A good PSU will have effective line filtering, and a really good PSU will have an active PFC on the input - which will give you a near-ideal power factor and typically handles input voltage changes automatically (so no 120/240 switch like on your old PC PSU).

Risk of catastrophic failure - There are any number of different ways for a PSU to fail, and it's unlikely but possible for a PSU to fail with a high output voltage condition. A cheap PSU will probably have fewer protections built into it to watch for, prevent or manage failure conditions like output overvoltage. If your cheap PSU blows and overvolts the miner, you could lose the miner as well. A well-made PSU's protection circuitry costs a bit more but it's kind of an insurance policy against exploding electronics.

[aknnig]

@sidehack and Finksy

Thanks for the useful comments.

I am considering the possibility of putting together a small nearly noise-free miner ~200W for the mass-market. To make this miner available for the mass-market it must be very cheap, of course on the cost of quality and risk of failures. I wonder if these cheap/bad power supplies with cost of 10-15 USD is sufficient.

It might be worth a test. So I am looking for the most important factors I must consider for the test.

Best regards,
aknnig

[sidehack]

You have my attention. Not so much with the "cost of quality" part but the "~200W miner for the mass-market" part.

[Finksy]

^ what he said, what stage are you at in development? Do you have a source for chips?

Also, being that most individuals in the market have PSU's left over from earlier generations, it would make more sense IMO to just have a standard PCIe plug on the 200W miner (or 2x if you plan on going higher than 250W) and allow everyone to source their own PSU.  For example, if someone bought 4 of the 200W miners, it would make more sense to use a single 800-1000W PSU than to buy 4 miners all equipped with individual 200W PSU's.  In the server PSU field, it's typically an "industry" of scale.  When you get up to 2000W+ server PSU's, you're buying them ready-to-go for as little as $0.04/W, which would be equivalent to $8 for a 200W PSU (which you are unlikely to source @ that price).

In regards to sale price and ROI, it will be cheaper for everyone involved if 1000 people were to source their own PSU's for as cost-effectively as possible, than for 1 person to source 1000 identical PSU's and include them in the sales price. Also makes for one less expense when building the miner.

[notlist3d]

A good example is over in A6 thread - https://bitcointalk.org/index.php?topic=1257327.msg14328111#msg14328111 .  If you like your equipment don't go with cheap supplies.  Failure is a possibility, and it could be very bad.   

But good "cheap" tend to be server psu's anymore.  I have both server and ATX as I have ROI'ed on them and kept for some miners.  But server PSU's have kinda taken over due to price and quality kits.   

I also am interested to hear if your 200 watt miner is something on drawling board, or has substance such as chip supply.

[sidehack]

Admittedly, for n00b gear, having an already-provided or internal PSU option would be nice. Some people would rather get something to just plug in and go (like the X1) than something needing a bunch of other stuff to make it run (like New R-Box). Of course that makes it more expensive, which makes it less viable as a miner, but if it's already fairly cheap and the best thing in that market sector (which would, at this point, be mostly for n00b introduction and learning rather than very profitable mining) you'd still see sales.

[NotFuzzyWarm]

At the 200-300w level there are many options for a PSU but 1 question first:
Is it to be a standalone box that is plugged into the wall with the PSU inside of it or a 2-piece setup as is most common (miner box and a separate PSU)?

If it is internal then is always the open-frame option which can be very easily and cheaply sourced. Small too.

If you are looking at doing this as a sizable business targeted at the proverbial Home Miner you better be looking at the electrical safety hazard part of this. No exposed line voltages allowed unless opened.

[FUBAR-BDHR]

It's not only if the power supply is cheap but what it is capable of.  You can have a good PSU that was intended for older systems that has overall power output that you would think is capable of handling a miner but in reality it can't put out the power on PIC-E rails the way is needed for a miner.

[Finksy]

It's not only if the power supply is cheap but what it is capable of.  You can have a good PSU that was intended for older systems that has overall power output that you would think is capable of handling a miner but in reality it can't put out the power on PIC-E rails the way is needed for a miner.

Eh?

[sidehack]

Possibly he's talking about a mulit-rail PSU, one with 18 or 20AWG leads, or one with a high total watts but low 12V watts (mostly 5V/3.3V for older P2/P3 era machines).

[NotFuzzyWarm]

The internal (open frame) supplies I'm talking about are such as http://www.trcelectronics.com/View/Mean-Well/MPS-200-12.shtml. Industrial/consumer rated ones like that are available from gazillion of suppliers. Definitely rated for continuous maximum load.

Should also look at power brick style supplies along the lines of http://www.newegg.com/Product/Product.aspx?Item=N82E16824999631R&nm_mc=KNC-GoogleAdwords-PC&cm_mmc=KNC-GoogleAdwords-PC-_-pla-_-POS+-+Accessories-_-N82E16824999631R&gclid=COuOvJLa68sCFQUMaQodciABdQ&gclsrc=aw.ds Bigger rating of course. I've seen them up to >250W.

[FUBAR-BDHR]

Possibly he's talking about a mulit-rail PSU, one with 18 or 20AWG leads, or one with a high total watts but low 12V watts (mostly 5V/3.3V for older P2/P3 era machines).

Exactly this.  Some have separate rails for each 12v PCI-E and some lump them on one rail.  Depending on the capacity of the rails even if the PSU is rated high enough it may not be able to deliver that power in a way the miner needs.

[aknnig]

@ sidehack
I have a tested an indirect proff of concept system/miner that is able to run with nearly 200W with absolutely no sound. Such a system if made easy, neat and cheap can enter the mass-market as a living room heater/miner. Then the electric power for mining is "free" or vise versa the electric power for heating is free, due to the bitcoins created. With mass-market I really mean customers that has little or no technical insight. They have no clear idea of what a power supply is.

So I am in the process of raising money to build ~10 prototypes for testing. I will need to make contact with integrators when the money is raised.

The prototype should have the following requirements:
1) They look like standard room heater
2) No sound or virtually no sound.
3) Plug and play with wireless connection, so the only technical thing the user needs to do is to type in their wireless WPA-key. An online bitcoin wallet should be pre-created for each miner.
4) Not so much more expensive than conventional space heater

Point 1 and 4 above demand some special form factor for the miner board and power supply, in additional to their price.

Sorry for being brief on the explanation, the idea is fairly simple so I want a head start first before coming with more details :-).

Best regards,
aknnig

[sidehack]

You'll definitely want to look at the open-frame Meanwells then.

How should I understand "indirect proof of concept"? Does this mean you have run an existing miner fanless at 200W? Or did you build a dummy load approximately in "space heater" formfactor, that runs fanless (or with a silent fan) and maintains a heatsink temperature below 80C at 200W dissipation?

[NotFuzzyWarm]

Now since you are talking about being used as a space heater, it can also be extended to being http://st.hzcdn.com/simgs/03d1b7a90617ed39_4-8698/contemporary-space-heaters.jpg

You want silent then either an oil or other fluid filled heater style would be perfect. I have one I used to run in my basement that pulls/delivers 1200w of heat with max fluid temp of 150F (66C) so it is always 'safe to touch' in case tiny fingers go prying around inside the louvers. Just convection drives the fluid flow - no pumps. No reason that the heating element in them can't be replaced with ASIC's. In my case it was replaced with 5.2kw of S7's running down there...

btw: that Mean-well open frame I linked to is just to show what they look like. That particular medical-grade supply is of course way overkill. Should be able to find that style in consumer/industrial grades a heckuvalot cheaper, especially from Asia.

[aknnig]

@sidehack and NotFuzzyWarm

I have run a fan-less s5 submerged in oil for some time, clocked down so it uses only 200W. Then pumped the oil through a conventional raditor heater which was able to keep the whole system around 60 degrees Cellsius.

Next step is to design a miner board that can be inserted directly into an oil-filled radiator, replacing the conventional heat elements. The whole system will then be controlled by a Raspberry Pi Zero alike computer with Wifi and power supply outside.

One important requirement is the shape of the miner board and the power supply, so they fit into conventional radiator heaters. The other important requirement is that the whole system will be able to pay for itself after say 2 years, provided that the electricity is for free.

Which such a system, I believe that we can introduce bitcoin to the masses living in the cold part of the world. It will be profitable for them to buy this bitcoin miner instead of a conventional radiator heater. But people are conservative. They may only do the transition if it looks like a radiator heater :-)

Questions:
1) Do you have an estimate of how much money an integrator will need to take a working ~200W board and redesign/reshape it to a 5cm x 40cm board?
2) What is the estimated cost price for such a board, if it was populated with Bitfurry 16nm or BM1385 or other similar chips?

Best regards,
aknnig

[Finksy]

I have always thought that would be a good idea, finding practical solutions for asic heat.  My only concern has been with maintaining proper fluid temperature.  In typical rads, the elements are cycled on and off to maintain the set temperature.  If ASICs were always being re-booted, I feel like you would be getting a lot of rejected shares, etc.  I wonder if there would be a way to write a highly variable clock-speed setting into CGminer, closed-loop based on the fluid temperature, instead of completely power cycling it?

I would look into the hot water boiler system called “Hotmine Smart Heater-Miner”, it would have to use similar controls to maintain temperature along with safeties.

P.S. WTH is with all the b!tcoin and other currency names being written upside or differently? It's forum-wide.

[sidehack]

I'm assuming that's an April Fools thing.

Also, it'd be nice to put in both frequency and core voltage adjustment for thermostatic control.

[aknnig]

@Finksy
In the test I did, the fluid temperature reached a maximum at 62 degree Celsius at a room temperature of 23 degree.  So the miner can run at full power most of the time in cold places.

“Hotmine Smart Heater-Miner” is a similar concept yes, but I feel that it has a smaller market compared to a space heater.


@sidehack
No, this is not a 1st. April's fools joke :-)

I think that these kind of ideas may bring the bitcoin mining to the masses, reducing the dominance of those huge mining farms placed where the electric power is cheap. So the everyday man will not be able to compete. These "free" electric power concept may level out the field and make the bitcoin ecosystem more decentralized.

It will not be easy to create new, step-change bitcoin mining chip after reaching 16nm. So the bitcoin chip development speed will slow down soon.

Proof of concept showed that this idea will work, at least technically. Economically is another beast. That is the reason for my questions above. Please comment if you have time.