54

As far as I understand, computers don't always draw the same amount of power from the power supply all the time. There are times when hard drives are on standby and not using as much as when they're spinning and graphics cards save power when not fully utilized.

What happens when, for example, you have 100 hard drives installed in a desktop tower (or a server rack) with let's say 1000 watt PSU, and they're all on standby, and then suddenly some process accesses all of the hard drives and spins them up, drawing more power than PSU can give?

Is there some signal that hard drives send when they think they're not getting enough power? Or does each individual hardware piece ask the PSU if it can provide it X watts of power, and it may say "no, I don't have that available"? Does the motherboard decide if it can negotiate this power request and safely avoid sudden power loss and instant shutdown? Or is the standard protocol in this case to drop dead without trying to avoid this problem?

From my experience with my desktop and a few hard drives and a low power 350W PSU, it would instantly shut down if 5 hard drives were all trying to spin up at the same time. Nothing bad happened, fortunately, but I'd like to know if instant shutdown is an expected and planned reaction of hardware pieces, or just the motherboard (or PSU) freaking out and disabling everything unexpectedly.

To clarify my question: What I'm interested in is why the common result is system shutdown instead of a safe denial of power to the device which would overload the system? USB power management protects against such a scenario, so why doesn't SATA/Molex power cable management logic not have this (or if it does, why it so commonly fails)?


Update after seeing some answers: I'm really surprised there isn't some sort of power management logic built into PSUs like motherboards have for managing USB power distribution. That's what I got from the answers so far. If you know something that says otherwise, please share as an answer.

  • 9
    @Ramhound fortunatly we have circuit breakers and fuses so in proper design it should not fry source. – Maciej Piechotka Apr 21 '17 at 23:22
  • 10
    This is why in a typical build you want the power supply to be capable of delivering substantially more than the peak draw of all components. Not only does this protect from such events, but a power supply will also last longer when its normal load is a smaller percentage of its peak capacity. – music2myear Apr 21 '17 at 23:40
  • 7
    FWIW, this is why high-end RAID controllers spin up disks on boot in batches and not all at once. – Jonas Schäfer Apr 22 '17 at 10:51
  • 4
    @Ramhound That's not true. Many power supplies - in general, I'm not talking about desktop computer power supplies - will continue to provide a lower voltage, which may not be enough to operate the load, but won't necessarily damage anything. When this happens we say the load 'draws down' the supply. Other supplies have overcurrent sensing and will shut off, ideally with a beep or a visual indication. And of course, trying to use a device that's powered by rechargeable batteries when the batteries are low doesn't destroy them. – Jeanne Pindar Apr 22 '17 at 23:43
  • 3
    That's just how electricity works - as you draw more power, the voltage drops, and the electronics in your computer mostly have a lower limit on voltage; if you draw enough power, they will simply stop working. Smarter PSUs (pretty much standard these days) will completely disconnect to prevent damage (e.g. from one component failing before another in a dangerous way). What would be the point of a PSU that randomly disconnects devices? The computer would be just as broken, possibly more so. And it would be more complicated - less reliable and efficient, more expensive than a bigger PSU. – Luaan Apr 24 '17 at 7:44

11 Answers 11

71

From an electronics standpoint, once the current drawn from a supply exceeds the supply capacity then the output voltage will suddenly drop. The electronics that requires a particular voltage to work will simply turn off. This is effectively a power brownout.

In the best case, the power supply detects this overload condition and holds itself off for some period of time or tests to see if the load is still there in a safe fashion, holding the power output off until the load is gone.

In the worst case the power supply goes into constant turn-on and brownout cycles and potentially kills itself or one or more of the devices connected.

There is no provision for devices to "request" more power from a power supply except in intelligent devices like USB where power availability was a concern to begin with. A system power supply has absolutely no intelligent electronics.

  • 20
    "A system power supply has absolutely no intelligent electronics." – I don't think that's true. It wouldn't surprise me if a system power supply had more processing power than the Apollo 11 computer system(s). It does however have no way to communicate with the devices. – Jörg W Mittag Apr 22 '17 at 2:10
  • 30
    The reason why USB is even able to negotiate power is that it is, at its core, a communications protocol, not a power delivery protocol. It works because additional power may or may not be available. A switch-mode or linear power supply will have some safe limit above which it cannot provide more power, and this amount of power corresponds to some particular current at a particular voltage. Your power company is more similar to the computer power supply than to USB; your appliances don't negotitate for additional power, but the power company detects the additional draw and compensates. – a CVn Apr 22 '17 at 13:15
  • 20
    Actually, even a modern ATX psu has little to no digital parts inside, let alone a microprocessor. It would be of no use, so there's none. Fault conditions are so fast that they must be handled in the analog world. And the Apollo definitely had more computing power than your average PSU. – Vladimir Cravero Apr 24 '17 at 7:38
  • 5
    @whatsisname I am sorry, but what you say either isn't true or does not make sense. uC are cheap, but fast ADC and DACs are not. High efficiency is not achieved via digital control at all, power factor correction can be done with one capacitor. If you do not believe me, just open an ATX psu. And yeah, I just happen to work in the semiconductor industry, on... power chips. Please, do not spread inaccurate information. – Vladimir Cravero Apr 25 '17 at 4:57
  • 2
    @whatsisname microcontrollers might be cheap, but in a highly competitive market every penny counts and if you don't have to use it you don't. Newer SMPS units might have fancy FET based overcurrent protection, but there is no way a microcontroller is necessary when a much simpler (cheaper) and more dedicated piece of silicon can be used. Using dedicated silicon is also cheaper than using a controller that needs someone to write software for it, and SMPS unit simply don't need an intelligent controller. – Mokubai Apr 25 '17 at 6:13
47

The power supply detects an overload condition and shuts down. There is no provision to negotiate power requirements.

  • 3
    Does this happen with every power supply ever, or are there different scenarios for old/new, desktop/server supplies? – user1306322 Apr 21 '17 at 20:53
  • 11
    I can't speak to every power supply ever, but this is what happens on typical desktops and servers based on standard PC designs. – David Schwartz Apr 21 '17 at 21:01
  • 10
    Indeed, this is part of the ATX power supply spec. I believe it was in the AT power supply spec as well. To use the supply again you will need to remove AC power completely (unplug it...or if it has a hardware power switch on the back, flip that to off) for 10 seconds or so, to discharge a capacitor. Then restore power. – Jamie Hanrahan Apr 21 '17 at 21:47
  • 4
    @kasperd: There are other intermediate regulators between the power supply and the digital logic. – Dietrich Epp Apr 22 '17 at 5:01
  • 2
    @kasperd: The comment wasn't an attempt to "invalidate" anything you wrote. – Dietrich Epp Apr 22 '17 at 17:00
36

Nothing bad happened, fortunately, but I'd like to know if instant shutdown is an expected and planned reaction of hardware pieces, or just the motherboard (or PSU) freaking out and disabling everything unexpectedly.

It is both. It is the PSU freaking out, which is the expected and planned reaction of hardware pieces. Once in "safety shutdown" mode, you need to either disconnect the PSU for a few seconds or reset a trip switch which is usually on the back1. Very cheap PSUs might not trip, and just leave the equipment malfunctioning due to insufficient current.

Worse still, they might enter a "spin-up, brown-out, power-down, power-up, spin-up, brown-out..." cycle that occasionally even self-solves and goes on to a proper boot. What is happening here is that the PSU isn't tripping and the equipment is subjected to unplanned wear and tear. In such a case, I'd advise on replacing the PSU altogether. It makes little sense to run through hoops to accommodate what is a misbehaving power supply to begin with. And while you're at replacing the PSU, get a more powerful one, which solves the initial problem.

Hard disks

However, hard disks are a special case, since they are known to have much greater spin-up requirements. So some hard disks (and motherboards) have provisions to handle this by delaying the spin-up, using jumpers that delay the spin-up by a fixed amount of time, or supporting PUIS (also here) or staggered spin-up via backplane signaling. Jumperless solutions require a suitable motherboard, that is capable of sending the appropriate signal to the hard disk (pin 11 of the SATA interface, implemented by WD and others). The software is either left to the user or sometimes implemented in the BIOS.

Updates

To clarify my question: What I'm interested in is why the common result is system shutdown instead of a safe denial of power to the device which would overload the system? USB power management protects against such a scenario [...] I'm really surprised there isn't some sort of power management logic built into PSUs like motherboards have for managing USB power distribution.

USB is a communication standard between devices that are more "intelligent" than what is required from your average hard disk (granted, the computing power on a hard disk is nothing to be sneered at - some of them can run Linux).

But the problems here are many:

  • the PSU cannot be sure about who's draining current. One power line might connect to up to four Molex connectors and the 12V/5V lines aren't designed to carry information. It could be done but you'd need to essentially reengineer both the PSU and all hardware likely to need such a feature.
  • denying power to one device might defeat the purpose of booting the whole system. Or lead to potentially disastrous results. Think what would happen if a RAID unit booted one (or two!) disk short due to it having been "current denied".
  • if the extreme current requirement stemmed from a hardware fault, the whole system is faulting and therefore the current policy of shutting everything down is, to my eyes, the safest line of conduct. Keep in mind that large, too-important-to-fail systems will be built differently and with huge redundancies, so in those scenarios a limited shutdown is also the best response, and it might not even happen because the faulting unit isn't requiring more current but simply not starting at all (circuit protection and breaking directly on all powered parts. In old high-end IBM AS/400 systems you could short circuit a drive and the system would go on working while one drive bay was going up in flames and smoke - I saw it happen. The unit was logically and electrically disconnected from the backplane, but this didn't prevent it from going on burning, of course; but with enough money even that can be prevented).
  • on the gripping hand, it's economically unsound - such a smart PSU would cost much more than a dumber, sturdier, more powerful PSU which would be simpler to build and likely to last longer, and would solve the same problem equally well (actually, having more current at its disposal, and working farther from full capacity, it would solve that particular problem better).

(1) I do remember one Hewlett Packard mini desktop which had it on the inside, next to the cable strips. It also had a "power on" green light on the inside. I imagine these are specially built PSUs for some specific arrangement, that then are employed elsewhere. Disconnecting from the wall outlet should be enough to reset the PSU, but if it isn't, before giving it for dead, try checking the internal side. You never know.

  • 1
    That is one knowledgeable answer! Thank you. – Andre Terra Apr 21 '17 at 23:02
  • Thanks for updating to better answer my question. I imagine it wouldn't be too difficult to connect every SATA device to a USB<>SATA adapter (which are cheap and plenty these days) and make the USB portion of that handle power negotiations. Do you think that could work? – user1306322 Apr 23 '17 at 10:11
  • 3
    possibly in principle, but in practice there is little to negotiate : the needed extra current is not available. The best you can achieve with this approach is booting the system, yes, but one HD short. – LSerni Apr 23 '17 at 10:31
  • The last point: A smart PSU doing the job equally well as a dumb but more powerful one? I don't think so. The dumb PSU will probably do the job better than the "smart" PSU. – Oskar Skog Apr 23 '17 at 10:43
  • 2
    @OskarSkog , that is correct. I have also overlooked that a more powerful PSU would work, all else equal, farther from its full capacity, thereby being subject to lesser strain. This, in addition to simpler construction, would result in the "dumb PSU" being likely to enjoy a much longer operational life. – LSerni Apr 23 '17 at 13:36
7

In the specific case of something electromechanical such as a hard drive, the power draw from the device will typically be highest during initial spin-up, and it will then drop some after the device is in steady state. For this reason, good RAID cards (for example) will have a setting to stagger the spin-up for all of the attached drives, so that the start-up load is not placed on the power supply all at once.

Another scenario that can happen is if you are close to the power limits of the power supply, and not actually over, the power supply may not shut down. Instead, as other answers have said, the voltage will drop. The end result can be seemingly random system crashes (such as a BSOD in Windows). Some of this depends on the quality of the power supply. High-quality power supplies will handle being pushed closer to their limits than their cheaper, lesser-quality counterparts.

6

I've experienced this over a decade ago.

At that time my HDD was almost full so I had to plug another 80GB HDD in. After booting up everything seemed fine.

But then after a few days or so, the system hung up or the screen flashed sporadically. After each flash the system went back as normal but in explorer the C drive disappeared or some other strange thing happened. Error happens, message boxes appeared... But the strangest thing is that my main HDD now appears as more than 1TB in diskmgmt.msc and other disk partitioning tools.

I couldn't find out the reason but I decided to replace the power supply when I saw an ad run by the famous PC store near my place for exchanging old keyboards/mice/power supplies for new ones. I took the new power supply home and cried when noticing that the 24-pin connector couldn't fit my 20-pin mainboard. After an hour I observed that the 4 extra pins can be pulled out to make it 20-pin compatible. Since that time no more strange thing happens and the PC ran happily ever after.

It's just that the old power supply is marginally enough for the old things plus the new HDD in normal cases. But in some situations the power requirement increases significantly and overload the source, making the voltage drop and brownout happens. Undefined behaviors will happen, like data loss, HDD disconnected or not recognized...

It cost me tens of GBs of data and taught me a new lesson.

End of the story


Now about the phenomenon:

Normally a power supply will provide a (nearly) constant voltage within its working power range. If a device draws some more power, voltage will drop a bit and it'll try to increase power to balance the load and increase voltage back to the normal value.

However once the power increases over its ability then the situation can't be recovered, the voltage will drop forever and never went back. If the dropped voltage is in the allowed range of the devices (like 12V to 11.5V) then it will still work. If it drops too low, obviously the whole system will be down because the chips don't work anymore with that voltage.

Maybe a smart power supply can just shut a device that causes the overload situation, but that's very complex and requires separate outputs for different devices and constant measuring of their power usage. How about multiple devices increase power at the same time? Which one will you decide to shut down? If that's the CPU or RAM will you shut them off?

There's no way to prevent it, except don't power that device, or require the device to manage its own power. This can be seen in USB standard. USB devices always start with the minimum power requirement (1 unit load). Once connected it'll negotiate with the host to give it more power. If the request is approved then it'll power the other necessary parts (like the HDD in the enclosure). You can also see that in old USB harddisk enclosures which require 2 USB ports, if you plug only the main cable it'll refuse to start, because it sees that there's not enough power.

  • Interesting story. Although I'm more interested in mechanisms that exist to prevent any potential failures caused by hardware wanting to use more power than the system can provide, not the actual consequences, which I know can't be good. – user1306322 Apr 22 '17 at 12:17
  • Yes I've added some more information – phuclv Apr 22 '17 at 13:03
5

When your system begins to draw more current that the PSU is rated for, one or more of the following will happen in order of likeliness:

  1. The system firmware will detect a power fault and will halt the processor and/or attempt to shut down the power supply. A power fault condition can be detected in many ways. None of those ways involves any kind of digital communication with the power supply. Some motherboards have sophisticated monitoring chips, others have basic circuits for doing so.

  2. Almost all (even cheaper) PSUs have an overcurrent protection circuit. Once the current limit is exceeded for a certain period of time (usually less than a millisecond) the PSU will simply shut off completely. It will require being disconnected from the mains (pulling plug or flipping switch), and reconnected before it will operate again.

  3. The current draw will overload whichever power rail it's drawing from, and the voltage will begin to drop in order to increase the current supplied. The motherboard's power regulators will no longer supply proper voltages to the CPU and/or other components. Either the motherboard, CPU, or memory will be unable to function and the system will halt or shutdown completely.

  4. The power supply will draw too much current causing components to heat up and burn out. This would only happen in the absence or failure of the things listed above, along with other protections that are in place to prevent such a situation.

If you want to know about the electrical details of the various things listen, you should ask EE.

  • 2
    The system has no way to monitor how much load is on the PSU, and preventing #4 or other bad things from happening is exactly why the PSU simply shuts down when it can not maintain the proper voltage. – psusi Apr 23 '17 at 2:02
4

In the case of the PC there's one more factor at work: The power supply is sending out a power-is-good signal. During startup it has a certain period of time in which to establish this (as of course at the very instant of startup the power won't be good.)

This acts as a deadman switch for the computer, if the signal drops the machine immediately shuts down (like what would happen if you held the power switch) as this is deemed less destructive than potentially flaky operation of the electronics causing uncommanded write operations.

Long, long ago computers didn't have that sort of protection which is why the old advice was to remove your floppies from the machine before turning it off.

2

It depends upon the SMPS, its quality and which EE standard it is following. I had similar experience few year back when my SMPS blew up, I was in college and had less money hence I bought a chinese SMPS. It used to work, but as soon as the CPU temperature used to go up and CPU fans used to kick in, the system use to start hanging and sometime I also used to get Blue Screen. At first I was not able to figure out that this is happening due to SMPS, but after I temporarily swapped the SMPS with my friend's, My system got OK, but the new SMPS got burned in my friends System. The local vendor has given me 1 month warrenty but was reluctant to honor it, but finally he gave me a used SMPS, I took it but the system used to reboot endlessly, the problem this time was that the smps was not able to provide enough power to even start the system. Later I bought a crosair SMPS and every thing went fine after that. But when my motherboard blew up, I reused the smps in my college project for making a refrigenerator using peltier device and there I noticed that the crosair SMPS used to shutdown if I short circuited the output or use to put heavy load on it, but it never blew up while the chinese ones never used to shutdown but used to burn up under load.

  • It's simple to look for power supplies that advertise overcurrent or short-circuit protection. Silent PC Review sometimes uses "the more the merrier" to summarize the protection circuits category in their reviews; see for example their review of the Seasonic X-400 fanless PSU. – a CVn Apr 22 '17 at 13:38
1

To answer your "Update" question, there is no power negotiation protocol because there's no use case for it. Imagine smart computer components which could negotiate power. What would you expect them to do if there's not enough power for them? Halt?

The problem is, biggest power consumers in a typical system are all essential for its functioning. If you have a CPU, HDD, DRAM or video chip that halts, the apparent result for the end user is the same as a brown-out: the system doesn't work at all.

On the other hand, such smart power management system would create a bunch of problems of its own. Incompatible protocol versions, devices and PSUs providing inaccurate power values and similar issues would result in systems which refuse to boot that otherwise could have worked just fine.

Actually, since you mentioned USB power management, here's a fun fact: virtually no device implements USB power management spec to the letter. Few devices that do (Sony PSP comes to mind) are known to only work reliably with original chargers and leave a much worse impression with end users compared to similar devices which ignore this part of USB spec.

  • "What would you expect them to do if there's not enough power for them" − warn the user with a notification on screen which BIOS interprets as "device with ID N at interface M in slot X, Y and Z are reporting that they can't start up coz there's not enough power for them" instead of just not booting or booting but instantly shutting down, potentially harming something or ruining the data on the hard drives. I understand that in cases where data security is important, accidentally installing too weak a power supply is not a common case, but accidents happen and by 2017 precautions could exist. – user1306322 Apr 25 '17 at 20:34
0

Running a PSU over capacity for short and long term can have all kinds of effects. It mainly depends on the components involved. Power supply can be cut off (fuse, temperature cut off), parts can melt (or age quicker) or the power becomes noisy, voltage drops (or raises). The effect this has on the system ranges from shut downs to (more nasty) bit errors and failed calculations (and as a result of that corrupted data or blue screens).

  • Sorry, but I don't think you understood the question. I am interested in learning more about the exact mechanisms of computer hardware acquiring power, and maybe some protocols that prevent abrupt shutdowns from happening. Not what happens when such protocols (if they exist) fail to prevent power overdraw. – user1306322 Apr 22 '17 at 17:13
  • Well, besides random failures there is only one protocol, turn off the PSU. Most computer parts will understand that :) – eckes Apr 23 '17 at 17:13
0

electricity 101 : the hardware will pull the power it needs from the supply ( regardless of where it comes from) if the power supply is not capable of supplying with the demand 3 things can happen a) it will burn out. b) it will continue to supply under immense strain till A or C happens c) power supply shuts off the output due to internal overload protections in place.

A and B are very likely in consumer applications and C not likely in those applications. C is likely in commercial grade equipments (1000W+ power-supplies) but A and B still do happen very often overall.

you can picture the power supply like a faucet. there is a certain amount of pressure and volume available. in proportion the faucet is connected to a dam with a pressure regulator. The size of the pipe is voltage this is constant at the outlet , the amps is what can be troublesome , if not enough amps that is when things fail. of course if there is not enough pressure at the tap you will get "brown outs" of the equipment but like i said behind a "dam". The hard drives are not too likely to have that happen but still possible as most of the power is consumed by graphics card and CPU ( usually) but if you have big disk array it can become a problem.

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.