39

Simply put, what's the expiration date of a computer? The reason I ask is for the sake of a society recovering from some sort of cataclysm and trying to reclaim lost technology. However, I could also see this being useful if there's a cold-case mystery/treasure-hunt that depends on the hard-drive of a computer that once belonged to a key witness.

Here's the setup:

  1. The computer is disconnected from any electrical grid. (No worries about electrical surges.)
  2. The computer is protected from the worst of the weather. (No water or sunlight damage.)
  3. The computer is subject to yearly temperature fluctuations. (All climate control is offline.)
  4. Before being put into storage, the computer had a fresh operating system of your choice installed, with no unusual backup systems or special care given to it.
  5. The computer must be some common variety of desktop computer that is commercially available as of 2014 or earlier. (Sorry Mr. Babbage, no one-offs, plans, or prototypes)
  6. The computer must have about a 50% chance of still being able to turn on and retrieve files or run programs, given the proper power supply, at the expiration date.

Bonus: If temperature was constant, how much longer would the computer last? For example, if it was in a cave system, an air-conditioned warehouse, or a deep underground vault.

Emmett R.
  • 1,391
  • 12
  • 17
  • 45
    I can hear the cries of the archeologists now, as they boot the only known pre-apocalypse computer in existence... "Why did it have to be Windows ME???" – GrandmasterB Dec 07 '14 at 03:52
  • 1
    Is this going to be a computer designed to last a long time, or just commodity consumer hardware? That would make a big difference. – Adam Miller Dec 08 '14 at 03:47
  • 1
    @AdamMiller Good question. That's the reasoning behind rule 5. It must be something that's mass-produced and sold commercially in market quantities. I'm also alright with the idea of military hardware, because that's also not going to be a small run of a few experimental machines, but something that's plentiful and meant to be used everyday, and might eventually enter the market as military surplus, if it lasts. Nothing experimental, nothing custom, nothing boutique or bespoke. – Emmett R. Dec 08 '14 at 03:57
  • 2
    @closevoter: I think that this question is fine, especially since the OP included his motivation for this question, as well as for future readers: "The reason I ask is for the sake of a society recovering from some sort of cataclysm and trying to reclaim lost technology. However, I could also see this being useful if there's a cold-case mystery/treasure-hunt that depends on the hard-drive of a computer that once belonged to a key witness." ....you may also wish to join the discussion here – Shokhet Dec 08 '14 at 05:26
  • 3
  • @Sisir Actually, that may be as close as we've gotten to an answer so far, since I would hope for an intact operating system, which requires storage integrity... – Emmett R. Dec 08 '14 at 05:40

7 Answers7

23

Are you allowed to perform maintenance before turning it on?

For a hard-drive-based computer in storage in a reasonably dry climate, there are two main sources of degradation: loss of electrolyte in the power-supply capacitors, and breakdown of lubricant of various moving parts. If you can re-lubricate the fans and the hard-drive motors, and replace the capacitors, you should be able to get a computer working again several decades after it was last shut down. Over time, the contents of the hard drive will be lost to randomization of the magnetic fields, but this is a very slow process.

Solid-state drives introduce a new failure mode: charge loss in the SSD's memory cells. Unlike the above problems, this is irreversible, and depending on the storage technology, may render the computer un-bootable in as little as a year or two.

Mark
  • 17,827
  • 1
  • 44
  • 85
  • 1
    I would think that you are "allowed to perform maintenance before turning it on," given the wording of item 4 in the question ("Before being put into storage, the computer had a fresh operating system of your choice installed, with no unusual backup systems or special care given to it."), the only limit is maintenance done *before* putting the computer away. Nice answer, +1. – Shokhet Dec 09 '14 at 02:05
  • 1
    @Mark: I highly doubt that SSD loses data that fast when not in use. Mainly because from personal experience, I have TF cards and SD cards over 10 years old that I can still access. Flash memory typically have data retention times of over 100 years in storage (compared to 10 years for magnetic drives). Here's a spec sheet from Freescale that states that their flash cells can last up to 1150 years when stored at 25 degrees celcius: http://www.freescale.com/files/microcontrollers/doc/eng_bulletin/EB618.pdf – slebetman Dec 10 '14 at 02:58
  • 6
    @slebetman, your TF and SD cards use low-density single-layer memory cells, as does the Freescale spec sheet (from 2005) you link. Higher densities and multi-layer cells reduce the persistence time greatly, and triple-layer cells reduces it even more. – Mark Dec 10 '14 at 03:09
  • 1
    @Mark do you have a citation that it is reduced by such a large amount? – March Ho Jan 31 '15 at 19:41
  • 1
    @MarchHo, no. My numbers come from an overview paper listing techniques for raising the data-retention time of TLC flash to ten years, plus anecdotal evidence. – Mark Jan 31 '15 at 20:14
  • @MarchHo: The situation would be analogous to degradation of paper documents which have different sizes of print on them or, in extreme cases, have three different texts printed in cyan, magenta, and yellow inks for the purpose of being viewed independently using red, green, and blue filters. – supercat May 20 '16 at 00:33
  • 4
    Lubricating a fan is very viable in non-laboratory conditions... If you take apart a modern hard drive enough to be able to relubricate motors, you chance damaging it permanently. – rackandboneman Sep 09 '16 at 12:15
11

The older the better. Modern computers are hard to keep functioning because they are closer to the fundamental limits of the materials.

Modern computers tend to have about a 3-5 year lifespan. It could go longer if the computer was not used.

There is an active community of owners of PDP-11s, some of which are from 1970!

If the comparisons of the lifespans of actively used computers is a good metric for how long computers would last when unplugged, the older the better!

Cort Ammon
  • 132,031
  • 20
  • 258
  • 452
  • 4
    Also, cool, stable temperature, dry environment, and being underground all help a lot. Modern computers are vulnerable to temperature variations (solder/caps), high temperatures (electronics/caps), moisture (mold), and cosmic rays (ICs/storage). Basically, any of the above kills a computer. Although cosmic rays can usually be ignored as the damage is highly random. – Ville Niemi Dec 07 '14 at 07:02
  • @VilleNiemi All valid failure modes. I just wonder which would be likely to fail first and when. Also, I would think that cosmic rays would be at least partially mitigated in an underground situation. – Emmett R. Dec 07 '14 at 17:32
  • @EmmettR. Yes, that is why the first sentence of my comment has "underground" in it. In stable conditions the hardware could stay functional for a long time, but modern mass storage is vulnerable to "bit rot". Eventually a file needed for booting would become corrupted. That would be my "first-to-fail" guess for a computer without faulty components or environmental stress factors. – Ville Niemi Dec 07 '14 at 18:08
  • 2
    "3-5 year lifespan" = wouldn't that be more of a 3-5 year practical use lifespan due to software demands more than electro/mechanical failures? Very few of my computers have died at year 5. – DA. Dec 08 '14 at 07:19
  • @DA: That may be a YMMV caveat. I have found my hardware to last 3-5 years. In particular graphics cards and power supplies seem to love to give out on me. My hardware lifespan seems to fit well with the practical use lifespan, so I used those numbers. I did have a P2 which lasted 15 years, but that's an older computer. Consider your hardware blessed ;) – Cort Ammon Dec 08 '14 at 15:47
  • I think the 3-5 years is an underestimation. I'm writing this on my Lenovo T61, which I bought in '07 0r '08, and have used as my main machine ever since. (I did replace the drive with an SSD a couple of years ago.) And I booted my old Dell laptop a couple of months ago, after it had been sitting for years. Worked fine after the battery charged up. Oh, and I have an HP-12C calculator from about 1982 that I still use. – jamesqf Jan 29 '15 at 04:33
  • What do you have against the PDP-11? It can run UNIX, and with some care, even TCP/IP. :-) FWIW, while I have replaced parts and extended it somewhat over time, my current PC just passed the five year mark, and is humming along nicely under my desk. No special precautions, really (but I do have it on a UPS, though the PSU is newer than the UPS). The one I had before then got to become about eight years old, and I probably could have breathed life back into it if I hadn't taken that as a good a time as any to upgrade. – user May 23 '17 at 17:02
9

The single biggest issue for the long term dormancy of a computer would be the humidity/moisture. Moisture will allow corrosion and some of the parts don't need much to make them useless. Moisture also allows for the growing of mold etc. which will also damage the machine. Barring that, however, a cool dry place can keep a computer viable for quite a long time.

I think someone should figure out what the half life for computers are. Because that appears to me to be a good model to use.

bowlturner
  • 49,161
  • 5
  • 104
  • 238
  • So, perhaps some sort of desert climate would be best for the computer, however it got stored? Or perhaps a warehouse computer at a silica gel factory? – Emmett R. Dec 07 '14 at 17:37
  • Yes, definitely a dry climate is best. However, your question was how long it would last, not where to store it. The purpose of my answer was to point out that the answer could vary by an order of magnitude or more depending on the age of computer you need to keep running. The older they are, the longer they last. – Cort Ammon Dec 07 '14 at 18:05
  • @EmmettR. A dessert wouldn't be bad as long as the computer was constantly subjected to the temperature differences between night and day. a bank vault could possibly work too. – bowlturner Dec 07 '14 at 18:34
  • I don't know if mold would necessarily be a concern...it doesn't grow on metals, plastic and other inorganic surfaces very well (though if said surfaces are dirty, then all bets are off) – DA. Dec 08 '14 at 07:15
  • @EmmettR. if you're looking for a plausible storage location for long term viability, how about a salt mine? http://www.needcoffee.com/2010/09/07/underground-vaults-interview/ – DA. Dec 08 '14 at 07:18
6

Right now I have a 1984 year Elektronika_MK-61 microcomputer in front of me, and it works perfectly if I turn it on. All these years it was stored in my grandfathers' garage. So, probably the computers can survive few decades in a less than perfect environment. Especially the ones manufactured between 1980 - 1992 years.

vodolaz095
  • 5,815
  • 2
  • 20
  • 35
5

The first thing to go is always your memory...

Cold storage

Cool, dry conditions that preserve a computer at least well enough to prevent corrosion and physical damage are a must. In damp or otherwise corrosive conditions, a computer may not last a year; delicate wiring can very quickly decay, especially if there is no moving air to limit humidity inside the case. Temperature can also play a key role; too wide of a range or too rapid of change can stress physical components. If the computer is in a moderately stable environment, above 50 and below 110 degrees Fahrenheit, it has the best chance of lasting a long time. Sunlight - or even low ambient light - can destroy almost every part of a computer, so it should be stored somewhere out of the light. If a computer isn't stored somewhere cool, dry, dark, and free of vibrations, expect it to last a few years; the worse the conditions, the lower the lifespan.

Memory, Sweet Memory

Regardless of storage medium, duplicate data will mean a much higher chance of being able to boot up successfully. A standard mirrored RAID can extend the lifespan of a drive (assuming it works) by many years. The more duplicates, the higher the chance the data will survive intact, possibly even doubling or tripling the lifespan.

  • Modern (magnetic) hard drives use tiny magnetic fields on platters. Given the right circumstances, an unpowered magnetic hard drive will last between 2-20 years. At the low end of that range, the drive itself would still function, but the data may be corrupt; by the the end of that range, the lubrication on the drive will have failed, and the motors won't work. Theoretically, given perfect circumstances, the information stored on a hard disk will last for decades, but the drive itself will fail long before that. If the future user is willing to completely dismantle the drive and lubricate the motors, I could see a drive lasting for an additional decade or two. Interestingly, it's very difficult to find any websites describing how long a hard drive would last, unplugged.
  • Solid state drives would probably fail within a decade. A subsection of SSDs (RAM disks) use RAM with a battery to keep the data fresh; those would fail very quickly. Regardless, the data doesn't seem to last as long as magnetic hard drives, and because of drive fragmentation, it would be much harder to rebuild as well. On the bright side, the drive itself would remain usable much longer, even if the data stored on it were gone; I can't find any good sources, but it looks like an SSD could last over 30 years in the right conditions.
  • Punch cards last a surprisingly long time. They may not hold much data, but a stack of cardstock punch cards kept in ideal circumstances could last thousands of years!
  • Magnetic tape has been shown to be perfectly readable in excess of 30 years; as long as the tape itself is undamaged, there is a high chance that data will last many decades. Duplicate data could extend that two or three times.
  • Floppy disks rarely last 10 years, though I have personal experience with a floppy that worked after almost 15 years - however, other disks in the same box didn't last longer than two or three years. The chances of any given floppy disk working after a decade is quite low.
  • CDs were initially promised to last hundreds of years, but in reality many disks failed within 5 years; most websites I found list the lifespan of a 'cold storage' CD as about 10 years, though there is a chance that a CD may last longer, as long as the foil backing doesn't degrade. Given the option, I would stick with DVDs.
  • DVDs, especially high-end, commercially manufactured DVDs, may last 100 years or more; the data is stored differently than CDs, which means that as long as the disk itself doesn't degrade, the data should last a very long time - some sources claim over a thousand years in ideal circumstances. I would expect average media to last about 100 years.
  • BluRay and other very modern optical storage have similar manufacturer claims of 100-150 years; there isn't as much data on those, but I would guess it is on par with DVDs.
  • Flash drives, SD cards, and other small, cheap media will likely lose data within 10 years. The better drives may have longer lifespans, but I wouldn't count on it.
  • Older EEPROM chips should last up to 30 years before the bits decay; some chips are very susceptible to UV light, however, so any sunlight would drastically reduce the life of the chip.
  • Data stored directly in transistors and diodes (or even wires/resistors) should last nigh-on forever, barring physical decay; instead of using magnetic or electrical charge, they use a physical connection to power or ground to give the 1s and 0s. Obviously, this data would be read only.

Hardware

  • Ok, I lied. The first thing to fail is probably the fan on the processor, rather than the storage. For systems with a fan, anyway. The fan is designed to spin all day every day, and without that motion to distribute the lubrication, it will quickly stick. Luckily, most computers will work without a fan... for a little while, at least.
  • An optical punch card reader could, theoretically, last for thousands of years. It has so few components, and is designed to work with such high tolerances, it may well outlast the computer itself.
  • Floppy drives need to be thoroughly cleaned after less than a decade. Dust and mold will cover the reading head and reduce even the best floppy drive to a floppy-destroyer. Once clean, however, the drive should last for at least 50 years.
  • Similarly, magnetic tape readers will need to be carefully cleaned. On the bright side, the tape itself would clean the heads eventually, at the cost of whatever was on that first few feet of tape. Many commercial readers/writers are kept in hermetically sealed cases, which would keep the drive clean. I would expect a tape reader to last for 50-100 years, as long as the components hold out. The motor would need to be lubricated.
  • Optical drives, especially drives that are normally sealed, will fare much better; solid caps will keep the drive working for 100 years, at least, and probably longer. Unlike sensitive hard drives, the motor on a CD, DVD, or BluRay drive will grind away without lubrication, for a little while at least. If cleaned, I expect an optical reader to last roughly as long as its media - between 50 and 100 years.
  • Capacitors in the power supply or on the motherboard will likely be the first point of failure there, at around 15 years. The hotter it gets, the lower the lifespan; however, capacitors are fairly easy to replace; as long as the user turning on the computer knows how to fix them, they won't have any effect on the lifespan of the computer. Solid-core capacitors would fix that problem; there are a number of commercially available motherboards with solid caps, however, there are usually not available for power supplies.
  • The BIOS, as well as various other chips on the motherboard, may become corrupt over time; I would expect modern chips to survive for several decades, and older chips to last even longer. Much older motherboards may use ROMs that are much less prone to corruption, and still older devices would use wires and AND/OR gates as memory, reducing the chance of corruption to almost nothing.

Conclusion

If a modern desktop computer were kept out of direct contact with weather (sunlight, rain, etc.) in an average climate (say, in a garage or shed), I would expect it to last for between 5 and 10 years. If the same computer were buried in a storage container in the desert (controlled climate, no sunlight, no damp), I would guess that it would last up to 20 years. The first thing to fail would probably be the data storage.

However, a computer with an SSD could be reinstalled with handy DVD installation disks; some computer could even boot and run from the DVD itself. Without a hard drive to worry about, the same computer would live another 10-20 years.

An old, single-board computer with solid-core capacitors, bulky wiring, fewer components in general, and much higher tolerance for noise (due to much slower processor speeds) may last centuries; as long as the chips and wires themselves don't decay, the device should keep working for hundreds of years.

Even then, there are still simpler computers that would last even longer. The NES, for example, uses a very basic processor with easy-to-replace components, and boots from a cartridge using a die-cast circuit for data. As long as they clean the contacts on the cartridges, someone could be playing Duck Hunt on an NES a thousand years from now!

ArmanX
  • 12,345
  • 3
  • 31
  • 49
  • Magnetic drives use magnetic fields, not electrical charges. 2) There are no battery-backed SSDs; they use trapped electrical charges. There are battery-backed ramdisks, but those will only last a few days without power. 3) Pressed CDs can last hundreds of years, though you may need to re-coat the reflective layer. Recordable CDs store the data in the reflective layer, and can't be recovered this way. 4) Transistors and diodes cannot store data without power.
  • – Mark Jan 29 '15 at 01:34
  • You're correct, I should have said magnetic fields. Replacing the backing on a CD is possible, of course, but difficult. However, diodes and transistors can store data without power, indefinitely; tying them to power or ground sets them permanently to 1 or 0. Depending how old school you want to go, resistors also work, but take up more space. – ArmanX Jan 29 '15 at 02:05