When I was in high school, I heard from various sources that if you had to manually turn off the computer, you were supposed to wait 10 seconds before turning it back on. Otherwise, you could wind up with corrupted data.
So w/ present day computers (computer produced after 2004+) still need to follow that rule? Or has that problem been pretty much fixed?
I think that might have been a good idea in 1984, but by 2004 that advice was already outdated.
IIRC, early hard drives did not park their read-write heads as a matter of course when being powered down. I can imagine that re-applying power to a drive that was spinning down could cause a problem. I think that got straightened out by the late 80s, though.
The reasoning I heard was that some capacitors will keep residual power that they might keep some memory non-zeroed and therefore cause unspecified glitch upon restarting - nothing about hard drives. I don’t know if it’s true and I’am also interested in straightdope on that.
It was never a good idea to power down one of those machines without parking the drive heads. If they didn’t do it automatically, you were supposed to run a program to park them before shutting down. It’s not really much different now–you have to ask the OS to turn off the machine for you, otherwise you could lose any data that it hasn’t gotten around to writing out.
The reason I always paused before turning a CPU back on was to make sure there wasn’t any residual data left in the RAM. Some power supplies in those days were massive, over-filtered linear monsters capable of keeping the RAM “alive” for as much as a second or two after you flipped off the power switch, and software wasn’t always rigorous about initializing variables, especially if it didn’t notice there was a power cycle.
The pause was probably never an important precaution, and 10 seconds would almost certainly be overkill. (I just waited until the CRT stopped crackling.) But you never power cycle a computer when things work right. The pause gave you one less thing to worry about when things went wrong.
Anyway, ritual can be comforting if you’re dealing with something that you can’t fully understand. One drawback of LCD displays for me is that I can’t rap my knuckle on the screen. I know it serves no purpose, but I’m ritually trying to wake up the machine when it seems distracted. Sometimes these things are more for our comfort than any technical need.
The Princeton University study (PDF warning) on cold boot attacks revealed a latency period exists when shutting down a computer. While the scope of the study was an attempt to reduce the opportunities for gaining access to encryption keys and protected data, a byproduct of the study says one should allow time for a computer’s memory and residual power to fully discharge and ensure data security. It makes sense to wait a sufficient time during a cold reboot that residual data does not remain. We were told to wait a full 60 seconds between cold reboots.
You won’t break the computer by turning it on quickly.
A small number of glitches don’t go away on a reboot. You have to cut off power (possibly in the back) for a little bit (often longer than 10s, in fact) to get the comp’s head straight.
I’ve experienced these a couple times (on modern machines): they’re definately real, and a little scary because you first think you’ve fried your motherboard for good. It’s not about RAM. RAM actually lasts minutes with no power (there’s this incredibly scary attack that got invented a few months ago that uses this fact to hack into almost any encryption system… well, scary if you’re the NSA. normal people can be hacked much more easily.) Also, there’s no software that doesn’t initialize its memory before using it. It has to do with controller chips’ internal states not being reset right because they’re still getting power from charged capacitors.
Question: was there any electronics that really could break from power cycling too quickly? Perhaps if you played with the switch for a fraction of a second? Or was it always that you wanted to wait to completely clear the old state?
Btw, most chips have a special reset pin that’s used to clear the chip’s state (as well as special instructions that a driver uses to do the same). However, some designers get lazy/careless and don’t use or create such a feature, thinking that the loss of power will do the work for them.
The idea is to let the power supply output voltages decay to low levels before you turn it on again. Some circuits may not initialize or reset properly if the voltages are not allowed to decay to low levels. This applies to many electronic devices. With battery operated devices, sometimes you need to remove the batteries, let it sit for a while, and replace the batteries.
My understanding was that it was a benefit to the power supply, rather than RAM contents and so on. When the operating system boots, it loads from disk or ROM and starts copying itself into RAM, loading files from disks etc. There is (generally) no persistence of state between power cycles, so it doesn’t matter what the RAM contains. The bootstrapped machine just see it all as new memory to use.
In most modern systems RAM that is allocated through the operating system is often zeroed by the OS before being passed to the program that requested it (a potential security issue - otherwise you just alloc as much RAM as possible then read it all in case anything interesting was left behind).
The old linear power supplies were generally temperamental beasts and the new switched mode PSUs (aka SMPS or SMPSU) aren’t too much better. Flick them on and off quickly and you can make them pop!
Given that most PSUs have a pair of very large capacity filter capacitors it can take 5-10 seconds before they fully drain. I measured the output of some SMPSU on the old arcade games I work with and they can deliver +5 for 5-10 seconds after powerdown.
tim
What do you mean by this?
Nobody uses powerless to initialize state in digital logic. Because it does not work. There is nothing inherent in the design of registers that guarantee that they will power up in one state or another after prolonged removal of power. There are specialized circuits to generate reset by detecting power loss. These can be simple or they can be complicated. Simple ones need the power to be off for a while in order to discharge the capacitor that is used to detect power comping up. I have not seen a simple reset circuit in real use for a long time. The more complex ones will detect short power transients and apply reset to the chip.
I wait 10 seconds, and that seems to work well. But I can’t cite or quantify.
These days you push and hold the power button for 4-10 seconds to force the machine to power off.
Unless there is no power transient. The “complicated” reset circuits are limited by the same mechanism as the “simple” ones.
The reason the RAM is not the issue is because of what trmatthe said. It would be a really bad bug if software made any use of uninitialized memory.
There was a time when really bad bugs were not uncommon. They are really, really, really easy to create even nowadays, and the power-off pause originated when many programmers were still working “down on the bare metal”, as they used to say. There was no memory manager to hold your hand for you. Forget to initialize one variable, and you ship out a time bomb.
Even if the motherboard won’t allow RAM corruption, there are still peripherals and whatnot with their own registers or RAM space that can be corrupted by a transient. But it usually doesn’t come up because computers don’t have a real power switch anymore.
It comes up a lot when you call your ISP and they tell you to sit there with the router unplugged for half a minute.
If there is no transient then the power is still on.
Oh. I always figured they just wanted to finish their coffee or something.
(Seriously, when I worked for Apple tech support we always got a bunch of PMU reset calls after a large area had a power outage.)
It’s a good idea to wait 10s or longer for any device with capacitors. You always want the caps to have a little drain time before you resupply power.
This is especially apparent in older CRT televisions, which have huge caps.
Also, as has been stated about hard drives in days of yore, the hard drive head would need to park. IIRC, the ST-225 (20 MB Seagate MFM drive) and the ST-251 (40 MB Seagate MFM drive) did not have the magnet which would pull the hard drive head on power down, but the ST-251-1 (also 40MB) did. This would be the start of the era where your hard drive head is parked after shutting it down without having to park it manually.
Here’s the trick - the magnet is only strong enough to pull the head when the drive is spun down all the way, so you would need to wait several seconds for it to spin down before you move the computer around, or you would bounce the head around on the hard drive platter and corrupt your data.