11

I have to admit that I didn't ever really think about how a hard drive worked until recently. I only had a vague understanding that there was (at least) one hard disc and some kind of "needle" that did "something", most likely not etching onto the disc since that would be rather bad for rewriteability. Turns out it adds/removes electromechanical charges.

Now I think I understand everything except one pretty basic thing. Are, or were in the past, the discs constantly spinning?

Because when you turn on an old computer, you could hear the HDD's motor start and seemingly was spinning the discs as fast as possible, instantly and perpetually. Why, though? Why cause that kind of wear and tear (and extra noise) at all times?

Maybe I'm misremembering. Maybe it only read the first data needed to boot the OS, and that's what I remember, and after that it settled down when reading from the HDD was less frequent?

Surely the motor that spins the disc must only be moving when it needs to move to a different section and it cannot be accessed by the needle moving as much as it can?

Maybe this is seen as a silly question with an obvious answer, but I'm not sure at all. Maybe it's no longer the case, but was it at some point a fact that hard disks spun their discs all the time while powered on?

Jyrese
  • 111
  • 1
  • 3
  • 2
    Aside: Apple floppy drives had variable operational speed depending on the track. The outer tracks could be run faster and store more data. – Weather Vane Sep 29 '22 at 18:59
  • 5
    Have you looked at Hard disk drive? – HABO Sep 30 '22 at 02:54
  • @WeatherVane so did Commodore and Sirius. There was a program on the Commodore 8032 floating about which could play tunes using the variable speeds. – cup Sep 30 '22 at 05:23
  • 2
    What? You can still buy laptops and desktops with spinning hard drives. – Kaz Sep 30 '22 at 05:49
  • Now I feel old. Yes, one can still buy a computer with "spinning rust" inside. It is neither trivial nor of particular usefulness. – fraxinus Sep 30 '22 at 06:46
  • 10
    Do hard disk drives really count as retrocomputing already? – gerrit Sep 30 '22 at 08:16
  • 1
    @gerrit, No they do not, this question is off-topic – Omar and Lorraine Sep 30 '22 at 09:18
  • 3
    The question has a key misunderstanding - disk rotation is not just to "move to a different section", it's so that the disk works at all as a storage device. Turning the magnetic patterns on the disk into electrical signals requires moving the disk. (This is said in one of the several answers, but I thought it worth a comment on the question itself). – dave Sep 30 '22 at 12:38
  • 2
    @gerrit Maybe not yet. In the future everything from the other computing stacks will have to get migrated here. – Theodore Sep 30 '22 at 13:49
  • 1
    @fraxinus Commercial storage arrays or high-capacity servers with spinning rust are quite common still. Storing petabytes of data on SSD is expensive so using HDDs is particularly useful. – doneal24 Sep 30 '22 at 14:12
  • As an aside on @cup's comment, the Commodore 1540/1541 does variable density zones in the smarter way: the disk spins at a constant velocity, and the electronics just read/write data faster or slower. No need for anything fancy on the mechanical side. – Tommy Sep 30 '22 at 16:36
  • 1
    @fraxinus, Re "...spinning rust...," Now that IS retro. AFAIK, it's been more than a couple of decades since anybody coated hard drive platters with red iron oxide. Modern platter coatings are way more high tech. – Solomon Slow Sep 30 '22 at 20:06

5 Answers5

21

Yes, mechanical disk drives are constantly spinning, for two reasons:

First, the wear and tear on the motor and drive is much bigger during spin-up and spin-down than when running at a constant speed. Simply spoken, hard disks die when switching on and off, not while running. The drive bearings and the motor are designed for constant operation.

The second point is that the (minimal) distance between head and disk is actually maintained not by a mechanical, but aerodynamic effect - The heads are in fact flying on a cushion of air rotated by the spinning disk. This cushion can only be properly maintained when the disk is turning at a very constant speed, that is why during spin-up and spin-down the heads need to be moved out of the way into special landing zones.

There are (were, now obsoleted by flash media), in fact, mechanical hard disks that do not stay running all the time: The IBM Microdrives and other CF-Card-sized storage media typically spin down after about 5-10 seconds (and need about the same time for spin-up and spin-down) - But this was designed-in more because of their intended usage in cameras and other mobile (i.e battery-powered) devices that are not very frequently accessed than to avoid mechanical wear. Such disks are, however, relatively useless in a computer - there's a video on YouTube by Phil's Computer Lab that clearly shows this is not a good idea.

Nayuki
  • 113
  • 1
  • 3
tofro
  • 34,832
  • 4
  • 89
  • 170
  • 5
    I don't think motor wear and tear when starting and stopping is significant, and some drives moved the head away from the disk rather than moving it to a landing zone. I think a bigger issue is that drives would take awhile to get up to speed, and adding five seconds to any disk operation that was preceded by a few minutes of inactivity would have been annoying. – supercat Sep 29 '22 at 21:07
  • 3
    The 2.5" hard drive in my Amiga 1200 spins down after ~15 minutes, and takes less than a second to spin up again. It's over 20 years old and gets used for several hours per day - no problems so far! – Bruce Abbott Sep 30 '22 at 06:06
  • The annoyance of having to wait until a spun-down disk starts up again is of course an issue (today), but wasn't: Eary hard disks never spun down as long as they had power. – tofro Sep 30 '22 at 06:20
  • 7
    @tofro Disk power management (spin down after some idle time) has been standard for decades - for desktops and laptops.. – Zac67 Sep 30 '22 at 07:35
  • 1
    The spin-up time is not much of an issue for non-OS drives. – Michael Sep 30 '22 at 11:08
  • The original HDD's were rated for 10k start-stops and 10kh operation. Head crashes were common if there were contamination leaks in production or the surface burnish was inferior or the ceramic sliders too harsh or the oxide coating too soft or the aluminum substrate had pits.. – Tony Stewart EE75 Oct 05 '22 at 04:26
  • @supercat I have an eco drive in my PC that spins down after a period of inactivity, and it's annoying as heck. There are lots of operations that require having the drive be operational, even if you're not using the drive - and your system sits frozen while it spins up and initializes. – Mark Ransom Oct 06 '22 at 18:33
15

Generally speaking, yes. Traditionally, the spindle comes up to speed and stays at speed while the computer is on.

Why, though?

Stopping and starting a motor, especially going from zero to really high RPMs very quickly, can impose higher wear-and-tear on the mechanisms than simply leaving it running at a constant speed.

Also along those lines, older hard disks had physically larger platters than modern drives. It could take many seconds, even a few minutes, for a really big cylinder to get up to many thousands of RPM. Do you want to wait 25 seconds every time you go to access the hard disk?

It may have had some inertia by tradition/norms, too. The first hard drives were attached to very large mainframe computers, they were always on (or being repaired) and doing work, usually accessing the drives. So why would you want to turn them off? You don't turn computers like that off.

Eventually, drives got slower (in terms of max RPM) and smaller (in terms of platter size) such that it is feasible to spin them up and down in just a few seconds. Most modern (roughly from the mid-90s) drives can spin up and down in a few seconds, and the operating system may place the drive into power saving mode after just a few minutes, particularly in laptops. For server applications they're usually spinning 24/7.

Surely the motor that spins the disc must only be moving when it needs to move to a different section and it cannot be accessed by the needle moving as much as it can?

Most drives have a head that can only move along a fixed track across the radius of the disc. Basically, one thin line across the disc is visible to the head. To view another part of the disc (such as the next sector not under the head) the disc must rotate.

Indeed, even the action of reading what's directly under the drive head requires the disc to spin. Hard drive heads float on a small cushion of air created by the spinning disc, and would immediately crash if the arm was not retracted before the drive spun down (handled automatically in all modern drives). Similarly, the head must be moving relative to the platter to read data, just like a tape recorder. No motion, no changing magnetic fields, no voltage induced in the read head, no data.

RETRAC
  • 13,656
  • 3
  • 42
  • 65
  • 3
    Did drive RPMs really decrease? AFAIK early fixed disks couldn’t spin very fast, partly because of their platter size, and really high speeds (15kRPM) were only made possible with smaller platters. – Stephen Kitt Sep 29 '22 at 18:49
  • 1
    @StephenKitt Fair point. Low-power drives in the 3.5" form tend to run at lower RPM (less power and faster start/stop), typically 4500 - 5400 RPM. But yes, 10-15,000 RPM has been standard for performance drives for a long time. I was thinking about the increasing prevalence of the 5400 RPM eco models. – RETRAC Sep 29 '22 at 18:59
  • 4
    Ah yes; performance slowly increased up to 7200 RPM in mainstream drives, and then crept back down. But by retro standards the trend back down is recent ;-). – Stephen Kitt Sep 29 '22 at 19:05
  • To your point though, spinning more slowly probably helps, but fast drives also have low-power modes and can spin down and back up quite quickly; so it’s not because drives started getting slower again that it became feasible to have them routinely stop in normal operation. – Stephen Kitt Sep 29 '22 at 19:12
  • 4
    The increasing prevalence of the 5400 RPM is because they're cheap to manufacture and most people don't even know what it means. Anyone who does, isn't buying a mechanical HDD these days, and going backwards from 7200 (mainstream for enthusiasts for at least the last 20y) is not an option. Xboxes have a 5400 in them. Once they learned people would wait for 40 gigs of textures to load off of those, they were like : hey we should flood the market. – Mazura Sep 30 '22 at 03:22
  • Here's the second answer stating that starting/stopping a motor would increase mechanical wear (Of what? Bearings? Why?). Maybe any explanations? Currently I don't see how that's could be true. – lvd Sep 30 '22 at 06:19
  • Also in my experience I had 1 or 2 HDDs failed at startup (not spinning after a power cycle), however every time that was a pure electrical problem (a killed driver IC). – lvd Sep 30 '22 at 06:21
  • @Mazura I beg to differ. I buy mechanical drives for bulk storage, since double-digit terabyte flash devies are still not cost-effective and I even buy 5400 RPM ones... for the high-availability backup drives that only get an rsync or rdiff-backup every night rather than active access. – ssokolow Sep 30 '22 at 08:01
6

Hard drive heads were designed to "float" immediately above the surface of a disk, without touching it, any time the disk was spinning and ready for use. Various mechanisms were used to avoid having the head contact areas of the disk that held data when it was powered down. A quality drive could spin for years on end continuously, without ever powering down, since the only part of the drive that would be subject to any wear from such spinning would be the bearings.

supercat
  • 35,993
  • 3
  • 63
  • 159
  • 4
    And when the heads did not "float" but actually touched the disk, Bad Things ™ happened and you generally would want a recent backup of everything on that disk. – doneal24 Sep 29 '22 at 17:50
  • 1
    Every self-respecting programming office had its souvenir disk platter, with the gouge marks, displayed on the office wall. – dave Sep 29 '22 at 23:11
  • 2
    @doneal24 This is the origin of the word "crash" in computing, in fact--it originally referred to the heads crashing into the disc surface, which destroys it in short order and is generally a problem that takes a lot of effort to recover from. The use of the term has since spread to other problems that take some effort to recover from, though usually not as much effort as getting a whole new hard drive and restoring a backup. – Hearth Sep 30 '22 at 01:42
  • @another-dave Have that along with some 8" floppies. My core memory disappeared during an office move several years ago and I never had the opportunity to take a 14" platter home. – doneal24 Sep 30 '22 at 12:17
3

One might ask whether there were such things as idle disks.

From the 'large computers in an air-conditioned room' era, a computer with a disk-based operating system would be unlikely to have its disks idle for very long. That would mean the system wasn't doing any work - and computers were too expensive to have them sitting around doing nothing. The goal of an operating system was often understood to be 'keep the CPU busy'.

Then, of course, there were removable disk packs to consider. Likely, if a pack was loaded, then it was probably needed by the current workload. If not needed, the disk could be spun down and removed; so in a sense this answers your question in the negative. The disk would not only be not spinning, it would be not in the drive.

dave
  • 35,301
  • 3
  • 80
  • 160
2

Are, or were in the past, the discs constantly spinning?

Hard disk were initially designed to spin constantly while powered on. Beginning with mobile computers/laptops, power management was added, so that idle disks could be spun down after a few minutes (became common in the early 1990s AFAIK). Subsequently, this power-saving feature was added to desktop computers as well (mid/late 1990s).

The increase of spin-up/spin-down cycles required an improved landing-zone or ramp-landing system than with constantly spinning disks, but these became standard accordingly.

Zac67
  • 3,914
  • 1
  • 17
  • 23
  • 2
    SCSI drives could be spun down already in the 80s. It's part of the SCSI standard command set. – Patrick Schlüter Sep 30 '22 at 10:11
  • @PatrickSchlüter I was about to mention that but wasn't sure when exactly that feature was added or if it was there from the start. – Zac67 Sep 30 '22 at 10:19