I owned a PC in the mid-to-late 90s, and while everyone said when you turn it off and back on again, count to 30. But it also had an instant reset button down below. Was there ever really any risk by not counting to thirty?
I am an EE system designer. I have designed many PC systems (and non PC ones). I do lots of power supply work and have dealt with many HDDs. As always, there is bound to be an exception somewhere. If someone told me to design an exception I certainly could, but here is my take:
30 seconds was always overkill in terms of any digital state. Even very large capacitors on a motherboard (not true inside the AC/DC supply or CRT) are completely drained in fractions of a second. I have taken many scope measurements on many systems and sometimes find a 3-second hold-up on badly designed circuits where a voltage that matters has very high impedance when the system is off (usually something in a stand-by power circuit with a load switch) and they'll prevent a clean reset. But 3 seconds is about the limit of what I've seen.
30 seconds was - and is - true for in-rush current limiters in your AC power supply. These come in two varieties negative and positive temperature co-efficient (search PTC and NTC ICL if you want to learn more) and they need some time to get back to the cold state to do their job. If you are removing AC power and your power supply is hot then you may damage a PTC with a quick power cycle. May. I have seen it a few times (you'll find a blackened component near your fuse). A well designed circuit should be able to handle a couple of hot-power cycles, but if you remove AC power and re-install it once within 30 seconds you should wait before doing it a second time. I guarantee you (but not really, right?) that in the testing of your supply the engineers hot powered it a few times... but they didn't do it 5 times or if they did and it failed they didn't consider it a design failure. But if it failed on a single hot cycle they would fix that.
About HDDs. I have never seen one damaged in an in-system power cycle (I have seen them damage by hot-plugging the HDD power cable: don't do that - they aren't tested for it). But I did not extensively test any '90s PCs for this, and parking heads is real (and was more real then, the HDDs were much less smart). However, starting in 2001 I did a ton of HDD power-cycle testing including very late 1990s units. I never saw what we took to be a "drive still spinning when power was re-applied" failure. The only failures were from voltage surges and mechanical shock. But each model of HDD would have its own personal weaknesses and I am sure there is some unit out there that can't handle certain power cycles.
So yeah, don't wait 30 seconds if you held the front panel button of your PC until you forced a power-off. 3 seconds is good. Don't wait 30 seconds if you removed AC power once (though if you think your supply is smoking hot then DO!). Do not remove and re-insert AC power over and over. You will surely burn it. I'd bet many supplies could be burned in less than 30 seconds of intentional add/remove AC power.
The reset button does not affect the power supply at all. It sends a reset signal to the CPU (and probably the bus).
Some power supply designs indeed will wear quicker if quickly switched on and off, but this is not typically about still-charged capacitors. Depending on the circuitry, the following things among others could be a concern:
Cumulative thermal load on the rectifier (which has no time to cool down) due to inrush currents, if the power supply did have enough time to discharge the capacitors during the 2 seconds. Usually, it has; otherwise, the computer would keep running for several seconds before even reacting to the switch.
The same principle, applied to startup circuits: these can be designed in a way that they can take the short term stress of operation, if given time to cool down afterwards.
In some cases, the power rails downstream of the power supply might not yet be at 0V if you quickly switch on again. This is not very likely (but possible in some edge cases, think CMOS latchup effects...) to do permanent damage, but can start up the system into a confused state (e.g., if some power on reset circuitry is kept from triggering). A confused system, while it is unlikely to suffer electrical damage, still has potential for data loss.
My best guess is that it was (and is) all about getting the computer into a well-defined state.
To make the CPU start at the correct address, it needs to get a RESET signal or something equivalent. And in the old days this was often generated by slowly charging a capacitor at power-up, so that RESET stayed low for some time after applying power. And typically this capacitor (and supporting cirtuitry) also took some time to reliably discharge. So, having power off for too short, the RESET signal wouldn't be there reliably. Pressing the REST button that the old machines still had, didn't use this capacitor, so it worked, no matter how short you pressed the button.
Another factor might have been the DRAM contents. 30 seconds should be enough to discharge all DRAM capacitors making up the individual bits of the RAM chips, so that nothing remains of the previous RAM contents. But to me that's not very convincing, as I don't think the typical DOS/Windows bootstrap process took care about anything found in RAM. Instead it filled the significant parts of the RAM freshly from boot media.
All the other answers talk about delaying switching back on to prevent possible damage/wear of the power-supply (which I fully agree with). The other reason for a delay that was "common knowledge" way back when was to prevent damage to hard disks.
In normal use, the heads of a hard-disk are incredibly close to the platters, and it is partly the spinning of the platters (thanks to the Bernoulli effect) that keep the heads from touching the surface (with catastrophic consequences if it happened at speed).
Early hard-drives were fragile things: you were advised to
park the heads before powering down (moving the heads to a "safe" part of the disk to "land"). On power-on, the platters would be allowed to get up to full speed (and the heads safely "floating") before the heads were moved back to "active" areas of the disk.
Allowing a "reasonable time" between removing the power and restoring it would allow the disks to "spin down" and the heads to land gently; rapid power-cycling ran the (probably small, but non-zero) risk of "something going wrong" and heads landing where they shouldn't and potentially losing data (as well as the mechanical strain on motors that were designed to run more-or-less constantly).
Even as drive-technology improved (with auto-park mechanisms; perhaps "locking" the heads so they could not touch the surface), it feels "one less thing to go wrong" if you let the drives power-down fully before restarting.
You are still recommended to wait about 30 seconds when turning a modern computer off and on again.
The risk of any damage was very small but not zero. The risk nowadays is probably even smaller due to protection circuitry being present or better than in the '90s.
The idea is that you wait enough time for any charge held in capacitors to dissipate. The risk is that, on powering up, any residual charge may cause a current surge that would damage the chips, processor, memory or interfaces. As I said, the risk was very small, I never saw any machine damaged by blipping the power switch. But I still wait about 5-10 seconds before turning a machine back on.
Count to 30 is an ill advice. You simply shouldn't switch off-on a computer within the same 2 seconds.
The reason is old computers were switched "hard-off", on the mains side of the power supply. To avoid your computer failing everytime there is a short break in the mains (happens once a week or so, depending where you live), the power supply has a one-second reserve.
If you switched off-on the computer within 2 seconds, you enter the grey area in which that one-second reserve may be drained - or not, depending on the configuration of the computer. There may be also some depowered circuits, while others are not.
If you had some RAM depowered but the CPU still running, software may do nasty things.
Two factors may require leaving a computer off for a significant period of time before switching it back on:
Some computers distinguish power-on resets from other kinds of reset by looking for a certain pattern in memory. If that pattern exists in memory, the computer may assume that certain other things in memory will be valid and should be kept rather than forced to known values. If the pattern exists but those other things are not valid, attempting to reset the computer without erasing the pattern from memory may be ineffective. Note that in some memory devices, some bits will have a slight bias toward "1" and some will have a slight bias toward "0", but memory designers often try to minimize the bias. While random manufacturing variations are likely to cause most bits to have a significant bias one way or the other, some bits may end up almost perfectly "balanced" and thus take an unusually long amount of time to lose their contents when power is lost.
While many devices consume enough current during operation that removing power will cause VDD to quickly fall low enough to force a clean reset, many chips support a variety of low-power modes. Even if a design never deliberately puts a chip into a super-low-power mode, a chip might somehow end up in such a mode anyway. If that occurs, the bypass and storage caps may be able to maintain VDD for many times longer than usual.
If removing power for a second and then reapplying it causes a system to perform its normal power-on boot sequence, odds are extremely good that one second was long enough. If it doesn't, it's possible that leaving the system off for a longer period of time might work; one shouldn't send the system in for repair until one has tried at least that.
Indeed, a discharged capacitor has a higher inrush current than a charged one.
However, some power supplies have a "slow start" logic which limits the inrush current into the discharged capacitors when the power supply gets connected to the main. In some cases, quickly power cycling a power supply does not reset the "slow start" logic and the nearly fully discharged capacitors get charged without any special current limiting. This is especially true when the inrush current limited is an NTC. The NTC needs to cool down before it can act as a current limiter again.
TL;DR version: an NTC thermistor is a very cheap way to limit the inrush current into a power supply (or a table saw or any other machine with a high power-on current). The NTC has a high resistance when it's at room temperature, so it will initially significantly limit the current going though it. It will then heat up quickly. At that higher temperature, its resistance will drop significantly and it no longer limit the current in a significant way. The temperature reaches an equilibrium where the NTC's resistance remains quite low, but still high enough to generate enough heat to keep it at the higher temperature required for low resistance. When the device is turned off, the NTC will cool down again - but not instantaneously.
I am not aware how widespread the use of NTC as inrush current limiters was in PC power supplies. My 2017 NAS uses one...but I guess that's non-standard.
Yes there was but it doesn't seem to be where most of these posts are coming from.
I build and repaired computers from this era in this era. While caps certainly could hold a charge, I can't think of any ever that could hold one for 30 seconds, and was used directly in the PC.
The reason for the 30 second rule was two fold.
First, and formost it gave time for heat to dissipate. These older computers didn't often have fans, and a poorly built, or poorly cleaned one could over heat. Counting to 30 would let even the highest (non-board frying) temperature cool down quite a bit. Now it's not like a two second cycle would be "hotter" then a 30 second cycle. But if your PC was overheating, it was better to wait a bit.
Second was connected peripherals, specially the monitors. In this era some form of vga monitor was typically in use. They were generally CRTs, and generally speaking your turned off the tower, but you left the monitor on. Power cycling the tower would me going from a, lets say 800x600x60 graphical mode to a text mode. In some monitors this would cause, well it wouldn't be good. You could, and I have seen, plenty of, fried monitors because the PC was power cycled too quickly. Other peripherals had other common issues. Modems needed time to reset, printers needed time to time out and realize the next chunk of data wasn't coming. Even some keyboards and mice could get "stuck" if you power cycled too quickly.
These things combined, made it a fast rule to just tell people count to 30, then to try to explain how to, or why to check all the little bits and pieces. In the end, it's a rule of thumb that was just easier to tell someone then to try to explain. Even today computers are little magical black boxes where you shove electricity in and get fancy moving pictures out, to a lot of people. Most people don't care to know the reason why, or how it works.
So in short, yes there was a benefit, and a need, but it was mostly social engineering.
Background: I'm a EE specializing in embedded systems.
You want a 10+ second delay between turning items off and back on. This still applies to modern computers (if actually removing power, not just soft shutdown / restart) and equipment with embedded processors including e.g. routers.
The 10+ second delay is primarily to allow power supplies (stored charge in capacitors) to fully decay to near zero. What this does is:
- Puts reset capacitors (which have a pullup resistor to power, usually with a parallel diode to help discharge them quickly at power-down and make this less of an issue) into a known state.
- Some devices (including e.g. SD cards and USB sticks) have built-in reset circuits. I've recently seen one make of SD card not function (100% repeatably) if the 3.3V supply wasn't taken down to under (IIRC) about 0.6V before re-applying power.
- Many years ago, I worked on a (low-power handicapped aid) system that we found would not reset - ever - if a serial cable was connected to a device with its own power supply, when we were using certain types of serial interface chips. The serial input power fed back through the protection diodes and kept the power supplies from discharging fully, leaving devices in bad states.
- If a device has gotten into a latched-up state (parasitic SCR), it's not going to un-latch until power is fully removed.
Issues with power-supply inrush protection also apply, per other answers (I'm focused primarily on digital, not power-supply design).
I have no idea about 1990s PCs, but the same advice was given for Amigas - and the reason given was because a soft reset or a short power cycle didn't allow for enough time for the RAM to drain of residual power and reset, and only a portion of the RAM was deliberately reset during the power on cycle, which meant that nasty apps could potentially survive a reboot by hiding in the upper memory space.
I have observed a faulty IC (actually a discrete PLL) where the internal counters did not reset unless the board had been powered off for hours - or the power/ground pins were shorted. This is clearly an extreme case (the vendor claimed they could replicate but not identify a design fault), but suggests that odd corner case behaviour is quite possible - particularly in the absence of good brown-out reset and even rigorous reset synchronisation.
15 answers and no one mentioned RAM. The point of shutting down a computer device (any computer device. PC, cell phone, printer, anything with RAM), waiting X seconds, and then rebooting, is to make sure the RAM completely clears. RAM can only store information as long as power is supplied. Any fault in the RAM causing random glitches, will be cleared by power-cycling with X seconds of wait time (to guarantee that power is completely emptied from the circuits). Doesn't necessarily have to be 30 seconds. 5-10 will do.
None of the other answers reflect what it was like to be a PC user in the mid 1990s. About 1994 or thereabouts, power supplies on Microsoft PC's changed from a regular switch to something less intuitive.
With a regular switch, it was obvious when the power was off. I could switch off my Vic-20 in 1982, count to 3, and then turn it back on. I knew for certain that it was completely powered down. Off meant no power. Want to turn off the monitor, PC, printer, and a lava lamp all at once when you are done playing Gorf? Just flip the on/off switch on the power strip. Similar for PC's around that time - there was an on/off switch.
Around 1995 or so, they introduced modern power supplies, where you hit the power button, and what happens isn't as obvious as the old switches. Maybe it will sleep? Hibernate? Do a shut down cycle that takes a while ("It is now safe to shut down your computer")? Maybe it will restart? Maybe it will do nothing at all, and you'll need to hit it again? Maybe it will wait 10 s then start to do something?
Because we lost easy control of the power supply, the only way to practically know if the computer was actually off was to wait until the lights were off, then wait a bit more to be sure it actually stopped working. It always seemed like if just might start up on its own after a short wait. As this stuff is more reliable now, 30 seconds is overkill. But ~1995 many computers were still really crappy; and power supplies were buggy.
30 seconds was a good way to know the computer was actually off; especially if you were trouble shooting your grandma's computer over the phone. They'd expect to be able to switch it on and off like a blender.
Over the last 35 years I can remember two instances of too fast a power cycle causing a major failure.
The first was a full height winchester hard drive which crashed it's heads when it was powered up before it had finished parking. After restoring from backups to a new drive I took it to pieces to find nastily deep scratches in the platters. By the mid 90's most hard drives would have been auto-parking however.
The second was a monitor whose power supply popped when the computer was quickly flicked off then on. From the other side of the room, I heard the cluck, clunk as the switch was thrown off then on, followed by the WOM of the monitor powering up, soon followed by the POP. This was in the days of the IBM Big Red Switch, where the monitor was powered from the switched mains side of the BRS. Again, by the mid 90's, few systems would have had the monitor powered directly from a PC power switch.
I have also had multiple instances of fans stalling after power was re-applied while they were still spinning down.
Conversely, I've also seen PCs which will happily keep running for a few seconds after mains power is removed, so unless you wait a little longer, you didn't actually end up power cycling them at all.
As a result of these experiences, I've always waited until a machine was quiet before turning it back on. That's not too long these days for SSD based machines with slow, quiet, PWM controlled fans, but old servers with arrays of 10 platter 10k rpm hard drives and multiple heavy duty high speed fans could be much longer.
On a side note, even if you power cycle your system for 30 seconds, you might be surprised at how in tact the 'dynamic' memory in your system is.
If you want to be certain that in-memory cryptographic keys have completely decayed, then you might have to wait significantly longer to make sure that Data remanence is not an issue. It's even worse if someone has access to perform a Cold boot attack.
Don't know about the specific unit you mention, but I personally witnessed when an IBM 5100 "portable computer" was bricked by turning the power off and back on rapidly (probably under 1 second). This would have been the late 1975. The problem with this unit was that the internal "power-on reset" signal would not be generated, burning out the power supply itself and a couple of other components.
Note that this unit used a "switching" power supply, not much different from more modern units, at least into the 1990s.
My take is that one needs to allow enough time for the main power supply voltages to be drained down -- 10 seconds or so.
It's still advisable to wait a few seconds, if your HDD is mechanical, i.e. doesn't apply to SSD disks.
People are talking about capacitor discharge. It's not the reason for waiting.
The reason was and still is for HDD to stop spinning, so that when the power turns back on, HDD's heads wouldn't land on spinning disk and scratch the surface, leading to disk damage and data loss.
I've only heard it as 10-15 sec, and as I remember is was to make sure everything in RAM was cleared. I had a Commodore Amiga back in the 1980's. You could mount a virtual drive in memory which would survive a restart (soft reset) and often even a crash that resulted in a reboot. I experimented a bit for how long the virtual disk could actually survive a power off, and I think I landed about 4-5 sec. In general it can be tough for hardware to be turned on and of quickly as others talk about here, but I mostly waited to make sure memory was cleared. This was before having a HDD and I used the virtual drive for extremely fast reboot as I then didn't need a slow floppy disk.
A PC is called a PC because IBM called their "brand new" microcomputer a PC back in 1981 or thereabouts. I started programming microcomputers in 1976 and I can tell you that even some electronic and electrical engineers were afraid to recommend any design they had not personally vetted.
The power supply, the RAM, the CPU and HDD were all susceptible to the problems mentioned in other replies. I would say THAT is the reason for the 30 second rule. The engineers used that as their rule of thumb because people were having power supply on/off problems right and left. As engineers got smarter they protected their designs from crazy user behaviors.
If you are working on vintage equipment you may want to be extra cautious. If in doubt, leave that puppy unplugged for AT LEAST 30 seconds. Make sure every indication of remaining power is OFF. "just saying" it, be careful.
The risk is you will end up with a brick for a PC.
Yes, a 37-second shutdown was sometimes needed on some 1990s computers.
I worked with computers in the mid- to late-1990s that were networked as part of a corporate network. The computers I worked with had Windows operating systems. Some computers on the network ran on Windows NT, but none of the computers I worked with did.
I was responsible for writing custom software for some of these computers. The simplest way of making sure that each computer was using the most up-to-date version of the software was to place the up-to-date copy in a particular place on one "fileserver" computer, and have a desktop shortcut on each "client" computer that pointed to that location on the "fileserver" computer.
Sometimes a computer (I forget whether it was the "client" or the "fileserver", or maybe it was sometimes either one) would get confused about how to find the up-to-date version of the software. (I forget if the problem was a network connectivity issue, or a mis-caching of the software.)
We found that the best way to clear the problem was to shut down a computer (whichever one seemed to not be finding the correct file) for at least 37 seconds, and turn it back on again. 35 seconds was not sufficient. I suspect that this fix involved two timeouts -- one of 30 seconds, and one of 5 seconds. I do not know what aspect of the computers or the network needed to be cleared by each timeout.
"Back in the day", the more expensive hard disks on "mainframe" computers had provision to prevent the heads from ever touching the media surface. To make them affordable (simple) enough for the personal computer market, the newer smaller HDDs would actually allow the heads to land on the media surface, but in a special "landing zone" where no information was recorded. This meant that on removal of power, the drive had to automatically seek to the landing zone before the disk slowed to the point the heads could no longer float (handled by the drive's controller).
Brief power interruptions (say a few seconds) ran the risk that, although the heads would position for landing, upon restoration of power with the disk still spinning, the controller could initiate a seek even though the drive was not yet spinning fast enough for proper float. This could lead to damage of the recording surface.
Waiting 30 seconds allowed the drive to come to a complete stop so that on restart, the controller would recognize the disk as starting up and not permit a seek until the drive had attained normal operating speed.