10

I bought an Intel CPU, but it seems it will take some time for the other components to arrive. My question is, if the CPU just stays in the box and is left unpowered, will it become unusable?

The CPU contains information such as instructions and microcode. Will the CPU's digital information become corrupted, rendering the CPU useless, if the CPU isn't powered for a long time?

In SSDs and HDDs, information becomes corrupted if the storage devices are left without power for an extended period of time. Are CPUs also like this?

onaquest64
  • 195
  • 1
    References: https://superuser.com/questions/1334494 https://superuser.com/questions/1647730 – Kromster Oct 22 '23 at 16:54
  • 13
    If CPUs went bad after being without power too long "retrocomputing" would not be a thing. Storage is much different than a Processing unit. – JoSSte Oct 22 '23 at 19:40
  • 1
    @JoSSte For Intel processors, Pentiums and older do not have updateable microcode, and so are likely not vulnerable to this concern. Also, older flash cells, while less dense than today, also aged slower. So retrocomputing with Pentiums and older is feasible even if CPUs went bad after being unpowered. – marcelm Oct 22 '23 at 20:05
  • 5
    @marcelm: Intel microcode isn't stored in flash inside the CPU; it's reloaded every boot (from the BIOS ROMs, and/or loaded by the OS). When things like the CPU model-name string (e.g. "Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz") and the built-in microcode version are programmed in at the factory, I assume it's more like a traditional PROM where the bits are literally tiny fuses that are programmed by burning them out or not. Similar to how they fuse off defective cores to sell a quad-core die as a dual-core SKU by hard disabling 2 cores. – Peter Cordes Oct 23 '23 at 00:33
  • Is the product turnover rate really that high in your region? Go to a random computer part store and they likely have months old CPU in a box they've never opened, maybe even a year or more, far longer than the delivery for your other parts will take. – Martheen Oct 23 '23 at 04:40
  • @PeterCordes Fair point. That doesn't change the fact that the comparison to old processors is flawed though... – marcelm Oct 23 '23 at 06:48
  • 2
    @marcelm: Yes. But my point was that a modern CPU will also be fine when it's 40 or 100 years old, not limited even by flash data retention times, regardless of being without power. So we actually don't need to distinguish between retro vs. current CPUs. (Their transistors are much smaller, so stuff like electromigration can hypothetically make them wear out if they're on for a really long time. But when off, I think they should be stable for a really long time.) – Peter Cordes Oct 23 '23 at 07:10
  • 3
    Also, the OP mentions HDDs. The magnetic storage is generally stable at room temperature for a pretty long time. And even SSDs are still usually ok for years, I think. After being off for years, maybe a good idea to read all the data or something to get the firmware to check for correctable ECC errors and maybe update any that are getting weak. Or do some writes to get wear-leveling to copy stuff. But that's after years. – Peter Cordes Oct 23 '23 at 07:14
  • 1
    Anecdotally, a man on YouTube buried 3 CPUs in his backyard, and after 1 and 2 months in the soil respectively, the first two CPUs booted up and worked fine. – Gallifreyan Oct 23 '23 at 10:02
  • "for a long time"... how long is long? – RonJohn Oct 25 '23 at 09:00
  • Why the question? What has prompted you to think this is an issue? Is this just a 'shower thought'? – Neil Oct 25 '23 at 18:41

6 Answers6

32

No, a CPU should be fine for many years if kept in appropriate conditions like the original packaging. Most of a CPU used an extremely advanced process analogous to printing - and which are then totally encased in a hard protective shell.

There may be some non volatile memory on the CPU - but this is pretty robust.

Journeyman Geek
  • 129,178
davidgo
  • 70,654
  • 1
    What'd make it unfine with time? What's stopping storing it for 1000 years? – OverLordGoldDragon Oct 22 '23 at 15:27
  • 12
    @OverLordGoldDragon cosmic rays, oxidation, thermal cycling, probably mold if you give it long enough. – hobbs Oct 22 '23 at 16:26
  • 6
    There might not even be any non-volatile flash memory inside the CPU. With Intel CPUs at least, "microcode updates" have to be loaded each boot. The BIOS firmware and/or the OS are responsible for doing this; you're not actually updating anything inside the CPU that persists across reboots. When the vendor programs information into the CPU, I think that's done with write-once stuff like a traditional PROM, where writing is literally like blowing a fuse to break a connection. It's probably a permanent physical change, not just electron migration like flash. – Peter Cordes Oct 23 '23 at 00:19
  • 1
    There isn't any Flash in the CPU; it requires different processing steps. People have pulled "new old stock" parts out of storage from 30+ years ago and had them work. – pjc50 Oct 23 '23 at 08:41
  • 3
    I would argue that there is little to nothing preventing a CPU or any silicon chip to last several hundreds years, if stored at the appropriate T and RH. Most aging mechanisms are not significant unless you are powering the chip itself. This does not strictly apply to flash memory chips with stored data on them - the data would be lost, but the functionality not. Other kind of memory chips would keep the data instead. – Vladimir Cravero Oct 23 '23 at 13:53
  • 2
    @pjc50 Flash can be put on a CPU die, that's what most microcontrollers do (e.g. STM32). The main problem is the high voltages have limited the process to 45 nm. That was a significant factor behind the huge microcontroller/embedded CPU shortages around COVID. 20 nm embedded flash started coming online a year or two ago, and some vendors (e.g. Gigadevices) went multi-chip packages. – user71659 Oct 23 '23 at 19:55
  • @hobbs But would keeping the CPU powered all that time prevent cosmic rays, oxidation and thermal cycling? – Kaz Oct 25 '23 at 02:20
  • @Kaz no, but that wasn't OverlordGoldDragon's question. – hobbs Oct 25 '23 at 02:52
  • It is extremely rare for a CPU to die. That's not a typical component to fail. The capacitors on a MB would be the first ones, and then you will have problems finding compatible boards with the right mount. The CPU itself would probably work for decades if not more. – Nelson Oct 25 '23 at 02:57
15

The only "volatile" information in an unpowered modern CPU is any MicroCode software that is loaded in the CPU as an update to the hardcoded instruction-set of the CPU.

(This MicroCode usually contains security fixes and error-corrections for problems that were discovered after the hardcoded instruction-set was finalized. Changing the hardcoded instruction-set requires a very costly re-design of the CPU, which usually isn't done.)

This MicroCode is preloaded in the factory or, when the CPU is actually in use, can also be loaded/updated by the BIOS/UEFI of the motherboard or the operating system.

Regardless of how the MicroCode (if any) was put in the CPU, it is stored in a special on-chip memory that will hold its content, without power, for at least several years, possibly a decade or more.

If this stored MicroCode would somehow get corrupted the CPU will detect this on startup (as part of its boot-time internal diagnostics) and will simply not use it, falling back purely on the hardcoded instructions set. And then the BIOS/UEFI or OS can re-install that MicroCode after the PC is booted up.

So, if effect, your concern, while valid, doesn't really cause any problem in practice.

End Anti-Semitic Hate
  • 8,328
Tonny
  • 31,463
  • 3
    Do any Intel CPUs have a re-programmable memory for MicroCode updates? Microcode Update Guidance suggests MicroCode updates can be read from the BIOS SPI flash, which is external to the CPU, before the first BIOS instruction executes. I.e. is the MicroCode stored in the actual CPU mask-programmed? – Chester Gillon Oct 22 '23 at 15:08
  • 3
    @ChesterGillon Yes, the uCode is read from memory space (it doesn't matter what device backs up that memory region) and loaded into an appropriate, undocumented, buffer inside the CPU. The uCode Sequencer has appropriate registers for uCode patches, see this for more info. – Margaret Bloom Oct 22 '23 at 16:16
  • 3
    @ChesterGillon: Correct. With Intel CPUs at least, "microcode updates" have to be loaded each boot. The BIOS firmware and/or the OS are responsible for doing this; you're not actually updating anything inside the CPU that persists across reboots. When the vendor programs information into the CPU (like its model-name string and default frequencies, and initial microcode), I think that's done with write-once stuff like a traditional PROM, where writing is literally like blowing a fuse to break a connection. It's probably a permanent physical change, not just electron migration like flash. – Peter Cordes Oct 23 '23 at 00:21
  • 3
    Except its even better. The FIRST thing intel CPUs do when they are powered on is wait for a little bit to have microcode uploaded to them by the motherboard. This happens on EVERY BOOT. If they don't get compatible signed firmware they hang or reboot the machine. – Beige The Color Oct 23 '23 at 19:58
  • 1
    @BeigeTheColor: Oh, that makes sense. People talk about things like Skylake-server and Kaby Lake shipping with the loop buffer (LSD) already disabled in the initial microcode versions they shipped with. But I guess that's really "in the earliest microcode versions available to mobo vendors that support them", if there is no microcode version baked / programmed into the CPU itself. (Not an important distinction most of the time, so it's a useful simplification in phrasing. But CPUs do still need non-volatile storage for their model name string like Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz) – Peter Cordes Oct 24 '23 at 00:27
  • @BeigeTheColor I thought Intel chips shipped with a permanent, read-only baseline microcode that would be used if the supplied microcode was corrupt or just wasn't provided, rather than hanging, and has a volatile buffer used to store and use any provided microcode, which it switches to if verified valid – simpleuser Oct 26 '23 at 18:48
  • @simpleuser They do come with a baseline rom microcode. But it is my understanding its minimal and they absolutely would refuse to boot if they are not fed a properly signed microcode. Maybe this is highly-specific to hardware versions. My understanding is all recent cpus would rather hang and reboot the machine if no microcode or "rejected" microcode is uploaded. – Beige The Color Oct 28 '23 at 19:48
10

The question doesn't mention the Intel CPU type, but the Intel datasheet should provide some guidance.

Taking the example of an Intel(R) Xeon(R) CPU E5-2620 v3, the Intel® Xeon® Processor E5-1600, E5-2600, and E5-4600 v3 Product Families, Volume 1 of 2, Electrical Datasheet contains the following about Storage Condition Ratings: enter image description here

Where the Timesustained storage parameter has a maximum of 30 months with the CPU in an unopened bag.

Chester Gillon
  • 447
  • 13
    Reading the notes, I suspect the 30 months (2½ years) is the longest Intel have left this part and checked it is still working. – TripeHound Oct 22 '23 at 14:55
  • 1
    I had an IBM (not Lenovo) DOS machine with a hard drive that I left in my basement for several years (at least 2, more like 3) and I wanted some old data off the drive. The machine start, but the drive did not. So the CPU was good all that time. – John Oct 22 '23 at 18:58
  • 4
    Good answer, but should explain what time_{sustained storage} actually is, it's certainly not the time after which the CPU will be useless. That's obvious, but not to the kind of person who would ask such a question as op. – Nobody Oct 22 '23 at 20:03
  • 2
    @Nobody Good point. I hadn't previously seen the Time_{sustained storage} used by a device manufacturer before. The 12th Generation Intel® Core™ Processors Package Storage Specifications suggests is related to moisture sensitivity. Not sure if the balls on the CPU are also subject to oxidation if exposed to free air when the CPU is yet to be installed in a socket. – Chester Gillon Oct 22 '23 at 20:13
  • 2
    I suspect the time limit has as much to do with the release of newer products as it does any actual physical deterioration. – barbecue Oct 22 '23 at 21:13
  • @barbecue: Or perhaps the amount of silica gel they include in the bags, in case of being near the moisture limits? The pin contacts themselves should be gold or gold-plated. I think the CPU package itself should be pretty well sealed, so water shouldn't be able to get in to oxidize the tiny copper wires that connect parts of the chip. – Peter Cordes Oct 23 '23 at 00:27
  • @John that's because the HDD also stores parameters about its internal workings AS data on the HDD itself. – Beige The Color Oct 23 '23 at 20:21
6

I had a Pentium 3 CPU disconnected and in a closet between 2006-2020, and it still booted fine :) that's 14 years disconnected.

hanshenrik
  • 1,695
  • 3
  • 21
  • 37
5

Will the CPU become unusable if left without power for an extended period of time?

No. I have kept Integrated Circuits and at least one CPU with dense Anti Static foam on the Pins and left them for a very long time with no damage.

I think it is important to have the Anti Static material on the pins to prevent any inadvertent static electricity damage.

You can keep them like this with no power for very long periods ( at least 3 or 4 years).

John
  • 49,923
  • Notwithstanding a solar superstorm of course. But that's unrelated to the question. – Clockwork Oct 22 '23 at 18:45
  • “for a very long time” mentioning a number of years would be interesting (even if it's an approximation). – A.L Oct 23 '23 at 15:25
  • I suggested at least 3 or 4 years. I believe that to be reasonable. Depending on the actual chip and saving conditions, it could be longer. (I have working electronics in my basement 65 years old and fine - not computers of course) – John Oct 23 '23 at 18:58
  • IMHO, giving a single numeric value, such as 3 or 4, is pretty meaningless. Is that supposed to be a mean lifetime before it breaks? Median? What's the distribution like? Narrow, broad? Or is the number supposed to mean something like "95% of all CPUs are still working after 3 or 4 years"? Again: What's the distribution? – Andreas Rejbrand Oct 24 '23 at 17:50
1

Answer Possibly yes, but you'd have to expose it to dangerous conditions.

  1. Be outside the atmosphere, where the silicon is exposed to solar radiation and radiation from the Van Allen belts
  2. Be near a nuclear explosion, releasing radiation directly or an Electromagnetic Pulse

The likelyhood of these situations is low, so not worth worrying about. You're more likely to suffer a flood or fire or theft.

Criggie
  • 2,469