37

I'm mainly interested in 3.5" floppies here, as they're what I have lying around, but other sizes could also be interesting to hear about if anyone has the data.

As I understand it, the head rubs against the disk and the disk rubs against the shell/sleeve when in operation. So unlike a CD, for example, surely the surface must wear out over time.

I can't find anything that gives any sort of ballpark figure, but obviously they last long enough for nobody to complain about it. Not referring to long-term data integrity while in storage here ─ that's a different matter.

I doubt anyone is going to leave a floppy to spin in their garage for however long it takes to answer this question, so I'm also cool with anecdotal evidence. I've never put more than a few hours into a single 3.5" disk before, so I don't have much personal evidence to go off.

spacer GIF
  • 471
  • 4
  • 5
  • 5
    On 5.25 and 3.5 inch floppies, there is a low-friction fabriclike sheet bonded to the inside of the shell/sleeve to minimize friction with the rotating diskette surface. I believe it may also contain some lubricant. Also as you suggest, the read/write heads unloaded when not reading/writing data, and the drives stopped rotating after a few seconds of idleness. – Armand Feb 15 '21 at 10:31
  • 3
    @Armand 3.5" disks never had a head-load mechanism (i.e the heads contact the floppy surface continuously). For 5.25" disks, only the early ones had actually retractable heads. – tofro Feb 15 '21 at 10:40
  • And for 8" floppies the heads are unloaded when not reading/writing because most (all?) 8" drive mechanisms keep the disk spun up continuously when the drive door is closed. – mnem Feb 15 '21 at 21:41
  • 2
    @mnem Even I am not old enough to have used 8" floppies more than once or twice - thanks. – Armand Feb 15 '21 at 23:44
  • @tofro Thanks, I guess the head-loading solenoid racket was burned into my brain by that initial PC. – Armand Feb 15 '21 at 23:52
  • 4
    This site seems to have a lot of technical info on floppy drives: http://www.retrotechnology.com/herbs_stuff/drive.html – Armand Feb 15 '21 at 23:53
  • 7
    Nobody complained about it because failed floppy disks were the normal state of things. – Mark Feb 16 '21 at 01:54
  • 1
    @Mark Were they? Looking back, my impression is that 5.25" DD disks were quite reliable -- my C-128 GEOS boot disk, which I used every day for about three years, never wore out. So were early (expensive) 3.5" DD disks -- think late 80s Amiga. Problems started with cheap HD disks disk in the early nineties, and became really bad in the late 90s, when it became almost impossible to buy quality disks. Probably because no-one relied on disks any more, using them only for data transfer. – Michael Graf Feb 16 '21 at 22:45
  • 1
    Working on 5.25 diskettes - as a transport medium for vital pgms/reports needed in the Jersey Branch of my Bank - cos we couldn't send 'em over the wire from London, otherwise they belonged to HMRC. Point of Jersey nullified right there. Anyway - came back to my desk after a ciggie break and someone had spilt an extremely sugary coffee all over my desk and diskette. I mopped it with tissues and chucked in the drive. No luck. The drive sounded like it didn't like me. – Malcolm May 08 '21 at 20:08
  • 1
    So I got a very sharp knife and cut off the outer casing - poured distilled water over it and cut off the top of a brand new diskette, removed it and put mine in. Chucked it in the drive - It worked! Never copied anything so fast in all my Life! – Malcolm May 08 '21 at 20:08
  • 1
    Related (and very important!) – 640KB May 19 '21 at 13:41
  • @640KB Very true! I forgot how much of a difference the cleanliness of any physical media or its drive that makes contact can make. – spacer GIF May 19 '21 at 14:35

9 Answers9

63

Anecdotal evidence (*1)

We had a bank as customer with branch offices all over the state. These were the 70s, the final years of batch only, although, already using remote batch with local data acquisition and storage. In this case all daily bookings were collected on 8 inch floppies (well, one was enough for a days work). During the late evening the main data center called each branch and requested all batched data entry. It got transferred into local spool and processed. All output, most notably all printouts from jobs and such requested for branch usage or customers, were spooled into transmission pools. Later, during early morning the branch office was called again and all collected data transferred, either for direkt output (print) or again locally spooled onto floppy disk for printout as soon as the clerks started their shift.

So far much like early batch without moving punch cards and print outs in return around the state or later remote batch with punch card reader and printers at each branch office. Except that late stage offered input on keyboard and CRT with nice forms offering support for correct entry and prevalidation against whatever rules there were for account numbers and alike.

At one point a branch office reported transmission errors. Batch data was incorrect and rejected. Service man went to site, but couldn't find anything, all was working properly, including the drives. In the following weeks an increasing error rate was reported. Each and every time the system was checked and found working flawless.

The situation escalated rather quick. At some point not only parts, like floppy drives, but the whole remote setup was replaced by factory new components. By that time customer - for sure not a small one - already threatened cancelling all our contracts and remove our hardware from all of their offices including the data center - mind you, a rather big customers with multiple CPUs alone in the data center for that state.

Amids all of that trouble, one day the field engineer was once again at site. By chance he had to remove a floppy from the machine. Something usually never happened, as by customer process, all data had to be removed before non-bank-staff could do anything with any of their machines.

Guess what, one could see thru the floppy.

Not just badly worn, but the magnetic coating was outright polished away, all the way to the transparent carrier film.

Of course the story was now revealed fast. Each of the clerks had his own floppy, and this guy had used this very same disk since the day he was trained on the new system about a year ago. No need to say that some education happened. In turn the bank ofc had to pay for all the effort - and as well kept a good customer for years to come. The floppy was kept as kind of a trophy and shown to new apprentice and non believers:))

So, I give you several hundred hours of continuous use before a late 1970s 8" floppy quits.

The most amazing factor here was that the drive was still able to read some data, althogh almost no magnetic surface was visible anymore.

*1 - I have a strong feeling I told the story already on RC.SE, but couldn't find it. Maybe it was in comments. Success: Schezuk found it.

Raffzahn
  • 222,541
  • 22
  • 631
  • 918
36

I did some assembly development on a dual-floppy (5 1/4" double-sided) IBM PC 5150 in 1982/83 (PC DOS 1.1 and onward). Boy, did that exercise the floppy drives! We wore out floppies all the time and soon settled on using only Verbatim brand as those seemed to last the longest. Even with Verbatim, heavily used diskettes lasted maybe a month before they started showing errors. In that context, "heavily used" probably meant an hour or two per day. Edit: As J-F Fabre and supercat suggested, the assembly work was constant reading/writing of source, temp, object, linker, etc, so it wasn't just disks spinning. This usage was indeed heavy on directory changes with all the temporary files and renaming.

Armand
  • 579
  • 3
  • 6
  • 7
    it may be because of the constant writing on the disk, not because of the spinning. – Jean-François Fabre Feb 15 '21 at 20:02
  • 10
    Writing and reading should cause the same amount of wear. A disk will primary wear in the spot under the head, however, so usage patterns which need to keep rereading or rewriting the directory would cause those areas to wear out faster than those which spend most of their time reading or writing data areas of a disk. – supercat Feb 15 '21 at 22:35
  • @supercat physical wear, yes, but a large number of writes (with usual variations in head alignment and timing) can "muddy up" the magnetic transitions and slightly impinge on neighboring tracks, to a point which might be detrimental to data integrity. Think low-tech rowhammer. – hobbs Feb 16 '21 at 02:35
  • 2
    @hobbs: When using a variety of drives of varying alignment, that could be an issue, but I think ~180K/side track spacing should be sufficiently loose to minimize such issues with properly-calibrated drives. My Apple //c can, with suitable software, write tracks at 75% of normal spacing and still read them back provided that tracks are written sequentially. I would not expect that to work reliably with repeated writes, but the fact that writing things that way works at all would suggest that track spacing is 33% wider than needed to actually store data. – supercat Feb 16 '21 at 03:12
  • 1
    @hobbs: That's not to say that I don't believe that problems could and did occur when using the drives that people had installed. If there had been a readily available and widely used drive calibration disk which could confirm that a drive's read tolerance was symmetrical, that could likely have helped things a lot. A disk where sector headers are precisely where they should be could be reliably read, or written and read back, with a drive that was about a sixth of a track high or low, but a sector which is written with a drive which is a 1/6 track high and then a 1/6 track low... – supercat Feb 17 '21 at 15:37
  • 1
    ...may not be reliably readable with a drive that was 1/6 track high, even though it might have been readable by the 1/6-track-high drive if it had only ever been written by the drive which was 1/6 of a track low (the signal would be weak, but automatic gain control might be able to compensate in the absence of magnetic residue). – supercat Feb 17 '21 at 15:40
32

My 1980 edition of the Apple II DOS manual mentions that, for a 5.25" disk:

With reasonable care a diskette will give you an average life of 40 hours——which is a lot, when you consider the few seconds it takes to LOAD most programs.

There is no source for where they got that figure from.

Simon27
  • 421
  • 3
  • 3
  • That's a great find. Can you imagine a writeable storage medium being advertised today with only a 40 hour average life? If even a budget flash drive was that unstable it would be laughed out of the room – Kevin Feb 15 '21 at 22:58
  • 8
    @KevinWells I would guess there are still plenty of cheap flash drives that may not reach 40 hours of cumulative write time. :) – Armand Feb 16 '21 at 00:04
  • 5
    This is consistent with the other answer, which puts the figure at 25-50h (1-2h/workday for a month). – Polygnome Feb 16 '21 at 10:36
  • 2
    For 360K of data, at 31.25K/s [https://docs.rs-online.com/41b6/0900766b8001b0a3.pdf], fourty hours is 12500 write cycles. This compares favourably with with SSDs at 3000 write cycles

    The same link gives, under a "Media Life" heading, 3.5 million passes per track and 30,000 media clamps, but how to tie those to hours of use is another question...

     – Dragon Feb 16 '21 at 21:32
  • @Polygnome - but you wouldn't spend 1-2h per workday accessing the disk, would you? You'd work in memory, remembering to save whatever you're working on once or twice an hour if you were careful ;-) – Michael Graf Feb 16 '21 at 22:49
  • 3
    @Michael Graf The 1-2 h/wkday in my case was indeed accessing the disk. Not enough RAM to hold all the files, so constant reading in/writing out for maybe 20 minutes at a time for a full assembly and link run. – Armand Feb 17 '21 at 00:34
  • 2
    @Armand: I wonder how much faster devtools could have been, running on old hardware if they'd used an OS and disk formats designed to improve build speed, and if they were optimized for that? A two-stage build process where phase 1 copies all the data to a scratch disk in Drive 2 that was written using Prince-of-Persia-style track-at-once I/O, the user replaces the source disk in Drive 1 with a second scratch disk, and the build then moves data between the two scratch disks could probably achieve performance an order of magnitude better than approaches adapted from minicomputers. – supercat Feb 17 '21 at 17:07
  • @supercat Hard drives for the PC were available within a year of that time, and remember that floppy drives were a big improvement over cassette tape :) (I had an Exidy Sorcerer from some years earlier) Occupying oneself for 15 or 20 minutes with something else wasn't too bad as long as you didn't have to actively monitor the process. Also, RAM availability rapidly increased to 256k, then 576k, which made the biggest difference. – Armand Feb 17 '21 at 21:33
  • 1
    @Armand: PC hard drives were quickly offered for commercial sale, to those who could afford them. That doesn't mean that they would have been available to most programmers at that time. Tools which were designed for quick hosted development on floppy-based (or even tape-based) microcomputers blew the socks off those which were evolved from tools designed for minicomputers. – supercat Feb 17 '21 at 21:51
  • 1
    @supercat We were a small-time operation (1 owner and 2 part-time student programmers) so we pretty much went with what was available in stores like Computerland. We were not very sophisticated in terms of tools or software engineering (using BASIC with MASM), but did get an original Bernoulli Box, which solved our diskette speed and capacity issues. – Armand Feb 18 '21 at 04:42
10

As I understand it, the head rubs against the disk and the disk rubs against the shell/sleeve when in operation.

Correct. But the head is only rubbing against the disk when data are being transferred (either reading or writing).

On an 8" or early 5¼" drive the read/write head (or heads on a dual-sided drive) is only pushed against the disk by a head load solenoid when data are being transferred. When data aren't being transferred the heads rest away from the disk.

On a late 5¼" or any 3½" drive the heads are permanently loaded, saving the cost and space of a head load solenoid, but the drive motor only operates when data are being transferred. When data aren't being transferred the heads rest on the stationary disk.

I used floppies for data storage with a CP/M system for several years and don't recall having any problems with floppy wear. (It did have a hard disk for OS/program storage.)

Graham Nye
  • 961
  • 5
  • 8
  • By any chance, was your hard disk CP/M system an Altos? And, if so, which model? – Craig Estey Feb 17 '21 at 22:24
  • @CraigEstey I'm afraid I don't recall the brand now. – Graham Nye Feb 17 '21 at 22:27
  • The reason I ask is that I was a S/W engineer at Altos. Do you remember the year you got the system? Altos was the first to manufacture a hard disk CP/M system [and for several years they were the only company that made one--it was what "made them" as a company]. There's a picture of the system here: https://www.old-computers.com/museum/computer.asp?st=1&c=456 Does it look familar? – Craig Estey Feb 17 '21 at 23:13
  • 2
    We used 5.25" floppies with a CP/M system in the 80's. Being kids, we bought the cheapest possible floppys - and got exactly what we paid for. While they specifically may not have "worn out", they stopped working irregularly often. With quality floppys (e.g.: Verbatim), I can concur. – Kingsley Feb 18 '21 at 00:11
  • @CraigEstey I used it around 1983/4. I don't know when it was purchased. It wasn't the pictured system. It was in the UK, not the US. – Graham Nye Feb 18 '21 at 00:15
  • Most of the early Altos looked like the previous picture. My last shot ... By 1983/4, there was a 8086 based system running [among others] MP/M-86, the ACS 586. It looked like: https://www.old-computers.com/museum/computer.asp?c=515 – Craig Estey Feb 18 '21 at 00:37
8

In the 80's, 360K 5.25" disks for PC were highly reliable. I did not have a hard drive, and compiling C and COBOL would make them work hard. I used them mostly without problems for months or longer. Verbatim were the most reliable. Some people in this site have posted a lifespan of 40 hours. It's hard to tell, but I would say that it was longer that. I used them all semester long, and I rarely had any problems.

In the 90s, 3.5" disks, with a capacity of 1.44 MB, were a very different story. No matter the brand, they seemed to go bad even when used lightly. That was surprising, considering the hard shell protection, and the cool sliding cover. My theory was that the higher density made them more sensitive. It's possible that small misalignments in one machine could made the disk unreadable for other computers.

I made sure to keep duplicates of 3.5" drives, or save files to two disks, just in case one failed.

luiscolorado
  • 181
  • 2
  • 3
    Calibration issues definitely seem to be relevant. I have two brand new USB floppy drives and one can read 90%+ of the old floppies I have without issue. The other one has trouble with many of the ones the first effortlessly read,but can easily read almost every disk the first had trouble with. (The remaining ones neither could do have bad sectors according to every drive I own.) Also, I've seen people tear open 1.44MiB disks and it looks like some have a proper protective felt sleeve, while others just have a little patch of felt on one side of the access window as a cost-cutting measure. – ssokolow Feb 17 '21 at 03:44
  • 1
    @ssokolow This is interesting. I have a cheap USB 3.5" floppy drive (which looks identical to the more expensive ones anyway - I figure they just slap their brand on the cheap ones) and it's largely okay with new floppies only it has written to (ejecting them slightly always gets them to read) but it fails to do anything with a stack of older disks. Can't read any data from them, even to dump them as an image, be it on Windows, Linux or to boot from the BIOS. Not that there's anything important on them anyway. Should probably find a different drive to compare with. – spacer GIF Feb 20 '21 at 19:10
7

Can't really put a figure on it. I've had the same disk in a portable USB 3.5" drive for 5 years which I used for installing licence keys on numerous machines and uploading to a database.

Doesn't this depend on whether the disk is spinning continuously or whether it just starts up, reads/writes the data and then stops. I seem to remember the 8" ones spin continuously. Some of the 5.25" spin continuously. Some have the start stop sequence. All the 3.5" seem to use the start/stop sequence.

On the Commodore and Sirius machines, the disks didn't always sit properly. Sometimes the centres got a bit mangled. They were still usable but many people class them as "worn out".

Normally the disks had to be discarded because there was something wrong with the casing. For the 5.25 and 8", the top side would split and sometimes prevent the disk from ejecting.

With the 3.5", the spring mechanism would stop working or get stuck. It always took ages to remove the little metal bit from the disk drive.

The other common problems with 5.25 and 8" were leaving them near a radiator, in the sun, or leaving them to collect dust instead of remembering to put them back in their sleeves. There was also the odd magnetic screwdriver that caused the disk to behave badly.

We've had to discard disks when one of the engineers inserted two 8" disks in the drive and had difficulty getting them out.

Very few were discarded because they actually "wore out".

cup
  • 2,525
  • 1
  • 9
  • 21
  • Were add-on centre hole reinforcers available for 5.25" disks, or did you have to buy disks with them already in place? – Andrew Morton Feb 16 '21 at 16:02
  • We had to buy them already in place. The problem is we couldn't always find ones with reinforced centres. – cup Feb 16 '21 at 17:09
  • I remember add-on center hole reinforcers being available for 5.25" diskettes in the early 1980s at least. I'm not sure how well they worked, though. – Armand Feb 16 '21 at 17:16
6

I never trusted one for longer than about half a semester. I worked in a university computer lab back when 3.5 inch floppies were how students saved everything. Each student had 10 MB of "private cloud" storage, but no one except computer majors trusted it.

Around the end of each semester we did a lot of helping people recover files, and got a lot of stories about how, "I've been using this disk all semester and it has always been very reliable." Of course, that's with the disk not being very cared for, always thrown in a backpack and often exposed to the Arizona heat, and saving after every sentence "just to be safe." I've had them last several years with moderate use and careful storage.

Karl Bielefeldt
  • 371
  • 1
  • 3
1

I think actual wear depends on the finish of the head as well as the material of the floppies. In addition the actual "dirt state" of the head may be important as well as the cleanness of the floppies.

So my guess is that the number actually will vary a lot. Also I think "loss of magnetism" will limit the life of the floppies in reality: I never saw a worn-out floppy, but I had a lot that couldn't be read after some time.

Eight inch floppies typically had the motor on all the time and used a "head load/unload" (plonk!), while the smaller ones (5.25, 3.5") had no head load any more (AFAIK), but turned off the motor when not in use.

I think the greatest problem would be a dirty or scratched head or some dirt on the floppy.

Rumors say a drop of coffee with sugar is very bad when spilled on floppy disks. A dusty drive probably is bad, too.

U. Windl
  • 263
  • 1
  • 4
0

According to this the floppy disk can last 3-5 years and up to 20 years. I have lots of floppy diskettes including the ones from an old Macintosh computer but some are already non-working and probably reached their lifespan.

Update: I want to mention that this is for the shelf life of a floppy diskette.

hugseirvak
  • 199
  • 8