Some fascinating stories in this discussion thread.

It starts with discussion about computers overheating, but about halfway through the thread, it switches to discussion of mainframe installations in which it was apparently typical to have an emergency power off button that was big, conspicuous, placed in a very visible and easily reachable location, and basically had everything short of a 'PUSH ME' sign over it. Needless to say, these buttons often got pushed, with results that make amusing stories after the fact.

Why exactly were these buttons there? The guess that comes to mind is that the installations were designed by engineers used to other kinds of industrial machinery, such as gears that could catch and mangle an unwary limb, or pumps transporting toxic or flammable fluids, where an emergency power off button could have safety benefits, and they kept the habit. But is there any reasonably probable scenario where there would be a safety benefit from powering off a mainframe a few seconds faster?

  • 19
    The history of the emergency power off switch dates back to 1959, when a fire in the Air Force's statistical division in the Pentagon caused $6.9 million in property damage and destroyed three IBM mainframe computers. ref: datacenterknowledge.com/archives/2007/05/07/… another ref: totaluptime.com/epo-emergency-power-off-button
    – Brian
    Commented Nov 25, 2020 at 21:53
  • 16
    Given how much power these machines consumed when working normally, power supplies would be able to supply a lot of current, and fused accordingly. A short in the wrong place could probably cause a fire. (That's easier than you might think -- I once got an Amiga floppy power cable glowing red, just by shorting the 12V line and using a (cheap, aftermarket) ATX supply...) Commented Nov 25, 2020 at 21:55
  • 9
    An orderly shutdown on a mainframe could easily take an hour, not "a few seconds". Before disks replaced the majority of "working" tapes, there was almost always someone in the machine room who could be overcome by smoke and fumes (not to mention the discharge of Halon which followed).
    – grahamj42
    Commented Nov 26, 2020 at 14:33
  • 4
    "basically had everything short of a 'PUSH ME' sign": for the simple reason that most of them were intended to be /pulled/ :-) Commented Nov 26, 2020 at 17:28
  • 4
    I never saw one that didn't have a plastic coffee cup taped over it, for safety. (That's the red button, not Carol). Commented Nov 26, 2020 at 22:19

5 Answers 5


Yes, huge safety concerns as I remember engineers sitting inside the cabinets of large mainframes while it was running, fully powered, large currents in each cabinet powering fans. Cooling water being pumped through the frames. Huge wiring looms hanging across the floor to great logic analysers on wheeled trolleys; trip hazards.

One person regularly smoked their pipe whilst sitting inside the open cabinet using naked flames to re-light it. Different times.

Once the water cooling developed a slight leak and it was months before the 4 foot deep water was discovered in the dark false floor void.

An emergency button was certainly re-assuring.


It starts with discussion about computers overheating,

That discussion seams to include some quite vague memory, so I wouldn't put to much into here. Still, such buttons were available and even installed after market, depending on company or state regulations. A CRAY Y-MP EL used at TU München is a great example, with its big power off:

enter image description here

(Picture taken from this report)

The massive plastic ring is an after-market add-on by university staff to avoid unintended use.

Needless to say, these buttons often got pushed, with results that make amusing stories after the fact.

Not really. Never had any such in several decades of working in computing centers.

Why exactly were these buttons there?

Regulations that require machinery above certain power level and/or in certain environment to have such an instant off facility.

But is there any reasonably probable scenario where there would be a safety benefit from powering off a mainframe a few seconds faster?

Emergency off switches are exactly for the time when there is no educated personnel to operate a regular shutdown. It's like emergency measures in contrast to security. Security is to allow secure operation by (hopefully) trained operators, while emergency measures must be made in a way that everyone can operate.

Imagine a mainframe with smoke is emerging - what is the better solution, having some random guy, noticing this jumping ahead and pressing the button, or walking to the next phone and calling some help desk to get hold of an operator?

  • 3
    After the magic smoke is gone, the rest of the machine doesn't matter much. Commented Nov 25, 2020 at 23:57
  • 17
    @manassehkatz-Moving2Codidact Beside that disks and alike are way more important than the CPU, What about people and building? They as well may be more expensive than just the CPU.
    – Raffzahn
    Commented Nov 26, 2020 at 0:23
  • 10
    Anecdote from the 1970s : We were renting time on an IBM site and a colleague who was mounting a tape had to evacuate the machine room because of smoke pouring from a controller. He hit the emergency stop button on the way out, and was roundly criticised by the IBM operator who wanted to fight the fire without taking down the system (IBM equipment had a "Unit Emergency" switch for this situation).
    – grahamj42
    Commented Nov 26, 2020 at 14:25
  • 3
    @Raffzahn: The people may be more expensive than the CPU? ;-) Commented Nov 27, 2020 at 9:01
  • @MichaelGraf the value of a person is single-digit millions of dollars, IIRC. Might be 10 million now with inflation Commented Mar 23, 2022 at 15:24

At CMU in 1968-69, we had an IBM 360/67, with the big Emergency Power Off switch at the bottom right corner of the blinkenlights panel.

It also had an 8" wide "THIMK!" sign atop the panel.

One midnight shift, the paper ball struck the "THIMK!" sign, which fell on the EPO switch.

Instant OFF!

In the morning, after the IBM Field Engineer had come and reset the system, the Director visited the machine room to find out what happened. In a hand-waving explanation, somebody replaced the sign, and showed how easily it could be knocked off to plummet down to the EPO switch, again!

Instant OFF! Again!

The "THIMK!" sign found a new home.

  • 3
    THIMK! ? what's that supposed to mean?
    – Raffzahn
    Commented Nov 27, 2020 at 0:44
  • 2
    According to Google; A spoof by Mad Magazine on the famous IBM "THINK" signs. Which is kind of what I thought (except that I didn't know it was from Mad, though not surprised) and fits with "on top of the IBM mainframe console at a University". Commented Nov 27, 2020 at 1:47
  • 8
    This is a nice anecdote, but alas, not an answer. Commented Nov 27, 2020 at 2:22
  • 3
    I am more concerned how it was " the paper ball" instead of "a random stray paper ball" that hit the sign. XD
    – stux
    Commented Nov 28, 2020 at 10:57
  • 1
    > THIMK! ? what's that supposed to mean? I remember reading a book as a kid around 1970 called "The Day Of The Drag Race" in which the protagonist had a "THIMK" sign in his garage, and the narrative explained that replacing thn N with M somehow "removed the sting of the injunction". Commented Dec 22, 2020 at 5:38

Stop A Big "Oops"

While I suspect that, short of national security situations (like say shutting off WOPR before it launches a first strike), mainframes were not deliberately shut off to stop a running program, I have heard (first-hand accounts, but not my own actions) of situations where, for example, a failed RAID mirrored drive was being replaced and the rebuild started going in the wrong direction - i.e., "new empty" copying to "old good" instead of "old good" to "new empty".

In such a situation, the main power switch may be the fastest way to stop the process and hope for successful restoration of the partially-erased "old good" drive. A "master" switch in that case makes more sense than an individual switch on each component because you want to get "everything" off absolutely ASAP. A few more seconds, and especially a few seconds of "complete current batch automatically including writing all disk directory updates" would actually do quite a bit of harm in that particular scenario.

  • 4
    In such a case I'd hit the 'offline' or 'power down' button on the disc drive.
    – dave
    Commented Nov 28, 2020 at 14:14
  • 1
    @another-dave In many cases that would indeed be a much better choice. But it depends on the specific situation - type of hardware, ease of access to separate switches, etc. Commented Nov 28, 2020 at 22:59

Not an answer: just a silly story on trips and on/off switches.

This happened in the early 80s. We used to have a VAX, and MV8000 and a few Data General Eclipses. One of the eclipses had a power problem and used to trip every so often.

The problem was the position of the reset switch - it was on the back wall, about 10ft up. In front of it was the 11 platter disks, the VAX and the MV8000. You had to get up on a ladder and reach over to reset the trip. The problem was that the ladders were kept in the janitor's office which was locked when the janitor wasn't around and it wasn't safe to stand on the disk drives - they'd sometimes shake like washing machines. Then someone got a bright idea - just use a long stick to flick the switch over.

This worked quite well until one day when one of the shorter engineers couldn't control the momentum of the stick. He's switched the Eclipse on but in the follow through, switched off the trip for the VAX, which happened to be facing in the opposite direction. Then, to switch on the VAX, he swing the stick in the opposite direction, switching off the Eclipse.

The head of dept promptly raised an order to change the position of the switch so we didn't have to improvise using sticks.

  • Not an answer, but I like anecdotes.
    – ghellquist
    Commented Nov 28, 2023 at 7:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .