24

I'm using my laptop to copy a couple of harddrives to other harddrives. My laptop will probably be crunching through 20 TB of data...

I'm wondering if this will affect my CPU long term in any way. And what if I did 1 TB at a time, and let the computer idle a bit instead of constantly copying for a day non-stop?

I have an Intel Core i7 10750H (averaging 4.2 GHz)

(I also wonder if long term effects depend on the CPU.)

FYI, all harddrives are HDD. Would the same answer be for SSD copying?

Summary:

  1. Does copying a lot affect the CPU in any negative way (if you are copying HDD or SSD)
  2. Could the answer to one depend on the CPU?
  3. Would idling the computer between the job at intervals help if there are in fact bad effects?
  • 18
    To the best of my knowledge, you will not affect your CPU in any considerable, non-negligible way. The "wear and tear" done to the HDDs is more considerable as HDDs performance is wittled down by each of its read/writes, but that's the point of the HDDs anyway, so avoiding using an HDD for read/write is like avoiding driving a car as to not put "wear and tear" on the engine - what's the point of the car? – Tyler N Dec 4 '20 at 17:24
  • 32
    20TB is a tiny amount of data. – Eugen Rieck Dec 4 '20 at 17:53
  • 11
    As others have noted copying data is not the most CPU intensive operation. It is, obviously, HDD intensive, and HDDs do generate heat. If one of them is internal the heat will add to the heat that needs to be dissipated by the cooling system, something like the weak spot of laptops, due to their small body and fan size, if any. So do the proper things: Make sure the laptop is on a hard surface with air below its body (not in bed on a mattress or blanket); keep the fan openings dust free; keep the laptop out of the sun; the cooler the room, the better (but beware condensation). – Peter - Reinstate Monica Dec 5 '20 at 8:51
  • 5
    A computer is a machine. This is one of the things that machine is designed to do. – Thorbjørn Ravn Andersen Dec 7 '20 at 9:20
  • 4
    @EugenRieck, 20 TB is not a tiny amount of data. Very few people have that much storage in a single place, and if it's being copied from and/or to spinning rust, you're probably looking at at least a day (24 hours) to copy that much, especially if you're doing any data integrity validation. I moved 2 TB (10% of 20 TB) from an old machine to newer storage about a year ago, and it took literal days (admittedly, also over gigabit LAN by necessity, and I was doing it in chunks, so wall time exceeded actual operational time, but still...). – Matthew Dec 7 '20 at 17:58
40

Damage to CPU – Minimal, if any

1a. CPUs are not heavily involved in file copies. The CPU's job is to tell the other hardware to do the transfer, and the actual transfer of 1s and 0s is left to other chipsets/IOs to do.

1b. Even if a file transfer did use a lot of CPU, CPUs have incredibly good longevity and usually processors don't fail or go bad. They do get replaced regularly due to being outdated, which speaks to the fact of its longevity – a CPU will outlive its efficiency lifespan.

Damage to HDD – Still minimal, but worth considering

  1. Using an HDD wears an HDD. It has moving parts, so the more they're used, the more potential for damage from wear – but if you can't use it, it's useless anyway. The best practice for HDDs is to know that at some point, it will fail and to have backups. This is the only real protection you have against HDD failure: backups.

  2. You can reduce the wear and tear on the HDD by ensuring that you run a single copy command at any given time (which is what I suspect is the typical method), so whether you do a drag & drop or run a command like robocopy, make sure you only do one of these actions and let it complete before starting another one.

Your specific questions

  1. No, you will not affect your CPU in a negative way. In fact, you can open task manager during this and probably witness your CPU not go over 20% usage, depending on what else your computer is doing in the background.

  2. No, all CPUs are built to be robust. A 3rd gen i7 vs an 8th gen i9, or a Ryzen7, what have you, is engineered to last and endure utilization far above what a file transfer will do to it.

  3. No, this would not be a helpful method to reduce the damage that we've already stated is negligible. Even if your CPU isn't properly cooled, this low utilization should not make it overheat – and if it does, its performance will only be affected while it's too hot. Its lifetime performance/longevity will not be affected by overheating.

  • 1
    How does doing single copies strictly sequentially reduce wear and tear on the hard drive? The amount of data doesn't change. The meaning of that advice is also unclear: Do you mean explicitly file by file? That's probably impossible due to the amount. If you drop and drag a folder or use a third party program the degree of parallelization depends on the OS or program anyway. – Peter - Reinstate Monica Dec 5 '20 at 8:40
  • 8
    @Peter-ReinstateMonica: I think the idea is to reduce movement of the HDD’s head. Not sure how much difference this makes with a modern operating system (and in any case the wear is minimal). – Michael Dec 5 '20 at 16:22
  • 2
    If files aren't sequential on the source drive, the head will be all over the place anyway. If the new drive is empty, they'll be written effectively sequentially. Also, if one/both drives are external, those enclosures tend to not have good ventilation so the drives will tend to get hot, not the CPU. Having a small desk fan blowing on the enclosure to force air through would be good for the drive's lifespan - you don't want them to be 40°C+ for the hours that a 20TB copy will take. – FreeMan Dec 7 '20 at 15:44
  • 1
    I made a paper "funnel" for a small desktop fan to the vents on the back of a drive enclosure for a drive clear/test process that ran about 36 hours. The temps spiked to about 42°C until I added the fan/funnel, then dropped into the 30s for a comfortable time while it completed the full disk read/write/read process. – FreeMan Dec 7 '20 at 15:45
  • 1
    @jamesqf: Look at the total throughput with multiple copies happening at once, vs. sequential. Mainstream OSes (and drive firmware) are normally tuned to favour latency somewhat, and will switch between read streams frequently, leading to lots of seeks and significantly lower throughput, even though you have lots of RAM and the drive has an internal 256MiB of RAM as a cache/buffer. With a drive throughput of 180MiB/s for example, 0.01 sec (10 ms) seek time costs 1.8MiB of data not transfered. So sequential reads / writes need to be much larger than 2M to amortize seek costs to minimal. – Peter Cordes Dec 7 '20 at 20:58
23

No, for two main reasons.

  • Intel and AMD CPUs in desktop and laptop computers, if properly installed, cooled, not overclocked, and not subject to electrical problems like a bad power supply, will last a very long time even if all cores are constantly 100% utilized--such that you don't usually have to worry about "wearing out" your CPU.

  • Furthermore, even from the first PC's of the early 80's, disk drives used something called Direct Memory Access (DMA). Floppy drives have used DMA since the first PC, and hard drives started to have DMA modes probably sometime in the late 80's.

    DMA basically allows drives to transfer data to and from RAM without the CPU in the middle. So hard drive transfers have even less of a CPU impact than you might think.

    Where the CPU might still be in the middle is if you have disk encryption running on your system, such as Bitlocker, TrueCrypt/VeraCrypt, or other. But it's still not going to be anything to worry about.

You are causing an HDD hard drive (not an SSD) to work harder by copying from it constantly. You are adding wear to an SDD by copying to it. Spinning-rust HDDs and even SSDs do have a finite lifetime that is likely less than your CPU.

TL;DR - you should be much more worried about the lifetime of your drives (and have backups/multiple copies) than the CPU.

  • It's worth pointing out in addition to the longevity of a CPU at 100% at all times, a file transfer is a very minimal usage of CPU, so it's like having a shield for a snowball fight - no damage should be expected for multiple reasons. – Tyler N Dec 4 '20 at 17:26
  • 1
    @sitaram yes the USB controller supports DMA. – Rodney Dec 5 '20 at 15:53
  • 2
    While agreed the drives are more of an issue, a one-time copy of 20 TB is nothing for any drive not from the stone age (and thus likely to fail any moment anyway^^) – Frank Hopkins Dec 5 '20 at 17:53
  • 1
    I think the most wear incurred to a drive is on power up/power down. If the drives are old try to do the copy in one shot with no power interruptions. – LawrenceC Dec 5 '20 at 21:32
  • 7
    @Prometheus 20+ years, even for hardware from 20+ years ago. Back in 2017, I had the pleasure of decommissioning a SPARCclassic (from the first production run even, so dated 1992). The CPU, RAM, and most of the other hardware worked just fine despite being 25 years old, the only real issue was the dead battery for the NVRAM module and the insane power consumption and noise of the 12 5.25" SCSI-2 HDDs connected to the system (the noise in the server room went down by almost 30dB that day, and power consumption went down enough to save a few dollars a month on the electricity bill). – Austin Hemmelgarn Dec 6 '20 at 15:56
16

Most hard disks are rated for a few terabytes read/write per day for several years. Copying 20 TB once will use up a few days of the disk's lifetime, which is counted in years, usually 5-10 years.

You may assure yourself that this is true for your disks by consulting the manufacturer's specifications.

As regarding the CPU, such an operation will not use it at all. The CPU will most of the time be waiting upon the disks, so will not even warm up noticeably.

So don't worry and go ahead. If something fails during the operation, this only means that the component was already in bad shape and would have failed anyway in a few days time.

4

The answers stating that the risk is basically 0% are completely correct.

However I would add one thing to

Would the same answer be for SSD copying?

As said in the top answer, both HDDs and SSDs do have a very much more relevant wear-induced failure risk than CPUs do. The failure modes are quite different:

  • HDDs wear out from changing position. This happens on both read and write operations, but is minimised by changing large amounts of data in one uninterrupted batch. So, copying the entire contents is about the most wear-friendly way you can access a HDD.
  • SSDs (flash-based ones, that is) don't care about reads (whether sequential or random-access), but wear out from write operations, and that at a somewhat constant rate. You'll find very different figures about how many write cycles a flash cell can endure, ranging from 100 to 1000000. Also the actual failures are stochastic, meaning some bits can fail quite quickly while most of the rest are still good. As a result, in particular cheap SD cards have a tendency to fail quickly aften being written only a couple of times.
    Decent SSDs will have error-detection and block-remapping algorithms that can avoid any catastrophic failure for a very long time, but it is still limited, in particular if you're going to write over and over and over again.

The conclusion is that what you should not do is to copy data from a HDD to another by first copying part of it onto a flash storage, then to the target HDD, then next part to the flash and so on. Like, for example, if you have only one USB port and connect HDD A, copy to laptop, connect&copy to HDD B, connect again A...

Practically speaking, even this would only matter if you either have a truely huge amount of data (petabytes), or use an underpowered flash medium as the intermediate storage.

If you would need to copy with intermediate storage, a HDD would be the better choice for that, because batch copying is what it's best at.

  • 1
    I've never seen a credible claim of 1E6 write cycles to a flash cell. Only the best EEPROM have that, and they outperform flash on this metrics by 1-3 orders of magnitude. More typically, the claim is a few thousand, tens of thousands at best, which is orders of magnitude worse than HDD, but this is mitigated by wear levelling - unless the SSD is nearly full, which is yet another consideration... – Zeus Dec 7 '20 at 2:03
  • Right. That figure seemed fishy to me too, but I wanted to avoid either having to dig deeper to get a reliable source for more concrete range or else getting comments about my numbers being out of date, that's why I basically just stated “flash will wear out eventually”. – leftaroundabout Dec 7 '20 at 10:56
3

I'm wondering if this will affect my cpu long term in ANY way.

Everything will affect your CPU life in SOME way.

It is just that many things (like your copying data from one HDD to another) do so in such a minuscule way, it is not worth mentioning.

So yes, your use of CPU for few hours will shorten it's life by few hours. You would also shorten it's life for few hours if it was showing screensaver for few hours, instead of copying your data. And even if computer was turned off, the CPU (and other components) would age (albeit at a usually much slower rate).

So to answer your question: It doesn't matter as the effect is incredibly small. Just copy the things you need to copy.

Anyway, the rest of your computer system (like motherboard and aforementioned HDDs) will very likely die much sooner than your CPU, leaving your computer useless. And probably, even before that motherboard and disks die, your computer will be so obsolete that you will not be using it even if it was still functional (like, find a few decades years old computer in museum and try to make it do such a simple task as browsing this web site)

  • 3
    Saying that the CPU will age more slowly by turning it off may not be true - most circuits suffer greater stress from going cold and then back to full temperature than from many hours of normal operation. – Mike Brockington Dec 7 '20 at 10:51
  • @MikeBrockington agreed, the thermal stretching is dangerous. I was thinking more along the lines of "if you unplug a CPU and put it in a locker for a year, it would age less then if you were running CPU benchmarks on it for a year -- but it would still age". – Matija Nalis Dec 8 '20 at 5:25
1

As a partial counterargument to some of the answers... if you're copying that much data, I would seriously consider using a copy tool that includes some sort of integrity checking in the process (e.g. rsync). Contrary to most of the answers, this will involve your CPU doing some work, since it has to read back the newly written data and verify it against what was read from the original.

That said, the other answers are also correct that CPU's are designed to do work. Spinning rust is sufficiently slower than recent CPU's that even this isn't likely to be the same order of workload as the sorts of number-crunching loads that represent "stress tests" for CPU's. As noted, in any sort of sane workload, including this, it is unlikely you will wear out a CPU before the point you would want to replace it due to it simply being outdated.

If you're really worried, just keep an eye on CPU temperature for a while. If it's getting pegged at over 90°C, you might want to figure out a way to improve your cooling. If it rarely cracks 70°C, it's probably no more harmful than the machine idling.

  • I wasn't asking about integrity checking, it was about normal copying and CPU wear and tear. Obviously CPU's are designed to work, but they do deteriorate eventually. – Dave Dec 8 '20 at 1:43
0
  1. Does copying a lot affect the CPU in any negative way (if you are copying HDD or SSD)

It depends on the age of your CPU.

  1. Could the answer to one depend on the CPU?

Yes. Many years ago 10,20,30? - I don't how, the CPU moved byte by byte itself. This work would heat the CPU (don't know to which extent!). At a certain time a technique called DMA (direct memory access) was invented to circumvent the CPU for such stupid transfer tasks.

  1. Would idling the computer between the job at intervals help if there are in fact bad effects?

Yes it would if you had a heat issue with the CPU, but that won't help with a modern CPU which prevents itself from overheating.

Apart from your question, you should rather focus on heat in your drives. If you are copying completely assembled drives in external plastic housing (bad!) you should not wonder about a temperature increase in your drives. You could prevent this by copying a drive sectorwise using linux with ddrescue, monitor the temperatur with smartmontools and interrupt the copying process once the temperature exceeds your desired limit.

0

If you are going to be doing this a lot, I advise copying as much in a contiguous operation as possible and do not artifically break-up large datasets into smaller ones and waiting periods of time in between copying the smaller ones. Here is why: As long as all operations are performed well within the OPERATING specifications of the mechanical hard drives (Max and Min Temps, Humidity, Thermal cycling, shock, etc) then there is not any MEANINGFUL statistical difference between copying files a few at a time or all in one go for the INFREQUENT single given dataset like your 20TB example. HOWEVER...keep in mind that when you DO copy files a few at a time, and have a waiting period that causes the HDD to shift into power-saving mode in between copy segments, the R/W heads perform numerous landings and the discs perform numerous spin-ups and spin-downs due to power-reduction sequences they would otherwise NOT experience when the copying is done in a single sustained operation. These operations, mechanical head loading, disc spin-up, and power cycling are the more frequent points of catastrophic failure in mechanical HDDs.

0

copy- data is read from storage and placed into RAM. paste- data is read from RAM and written to new storage location. your main storage is the most likely bottleneck in this case. not having enough RAM to cache the data would slow the entire operation down too

  • That's not quite true, and depends on a lot of different variables. Additionally, this question has already been answered. – music2myear Jan 4 at 3:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.