Can I run my CPU at 100% usage for a long time?

Question

I have a Dell XPS 15 9570, 32 GB ram, 2.2 ghz and 6 cores. When i use an application (The parsim command on MATLAB) (for more details) then my fan starts running fast , during this my PC doesn't seem to be heating but it never got this noisy in the past so it got me worried. In my task manager , the CPU usage shows 100% when this application is running.

Is it a problem for the life of laptop, if I run the PC at CPU 100% for like 4 hours in a day?
Do I need to be worried about the noisy fan during the application?

Comments are not for extended discussion; this conversation has been [moved to chat](https://chat.stackexchange.com/rooms/105412/discussion-on-question-by-aadil095-can-i-run-my-cpu-at-100-usage-for-a-long-tim). — DavidPostill, Mar 10 '20 at 18:37
I have the same problem with my 2011 macbook pro 13 - Can't do heavy work, fans will starts...I think it is OK but with fan on and over heating, performance gets affected as well. — TheTechGuy, Mar 16 '20 at 14:19
Yes, it's a problem. This machine will overheat, and shut down. Mine does. Regularly. But I can't say whether ot not this shortens the life of the laptop. — Strawberry, Mar 17 '20 at 10:41
People make such a fuss over such a simple problem: it is not the CPU utilization (%) that could kill the CPU, it is the temperature. If your fan/heat radiator can keep the CPU temperature way under the critical threshold, then you have no reason to worry. It is that simple! Just find out your CPU critical temperature and current CPU temperature. Compare them and decide if they are too close. — Gabriel, Mar 18 '20 at 18:01

score 70 · Accepted Answer · edited Jun 12 '20 at 13:48

tl;dr– Issues like overheating and excessive wearing can be important considerations for computers that handle heavier loads.

A lot of folks do run their computers continuously at ~100%. This is common practice in computational work, e.g. with engineering simulations or data analysis.

There're two common things to watch out for:

Overheating:
Is your CPU, GPU, motherboard, RAM, disk, etc., getting too hot?
Excessive wearing:
Is a piece of hardware that can get worn out being used too much? For example, some consumer disk drives aren't made to be constantly reading/writing data.

1: Overheating issues.

CPU overheating:
In most cases, most heat will come from your CPU – which'll likely be the concern if you're observing 100% CPU usage (which is different from 100% GPU usage, 100% disk usage, 100% network usage, etc.). So, monitoring the CPU temperature is usually the initial focus.
Motherboard overheating:
A related issue is the motherboard's temperature. A lot of folks ignore it and assume that it's probably fine as long as the other hardware temperatures are in-control, which seems to be a reasonable guestimate in most cases so long as there's also good air circulation inside of the box. This can be more dubious in the case of compact devices, e.g. laptops.
GPU overheating:
GPU temperatures can be a big issue for folks who make heavy use of the GPU, including gamers, graphics professionals, and those who use GPU's as computational co-processors. I think I've seen even little Flash games in a browser go hard on a GPU, presumably due to inefficiency.
RAM (memory) overheating:
RAM (memory) temperatures can be an issue especially with highly overclocked RAM. Some vendors'll sell fans with their high-frequency RAM for this reason.
Disk overheating:
Some disk drives, e.g. 15,000-RPM hard disks, can have trouble with overheating. I haven't heard of overheating being a big problem with most lower-speed hard-disk drives (HDD's) and solid-state drives (SSD's).

There're other heats that one might watch out for, e.g. a heavily used server might watch out for heating on its network card while a computer with an embedded radio device might watch out there, but those seem to be less common sources of concern for typical users.

2. Excessive wearing issues.

Disk wearing:
Most^{(I think?)} disk drives aren't made to be continuously used at max capacity. So if you have an app that's constantly maxing out a disk drive, e.g. if it's constantly rewriting log files, then that might be something to check into. Servers and professional workstations often have disk drives designed to handle heavier loads.
Fan wearing:
Cooling fans continuously running at max might get worn out. A complicating issue with fans can be that if one wears out, but you don't notice it or monitor the related temperatures, then something might get excessively hot before the fan's failure is noticed.

Power-supply unit (PSU) wearing can also be an issue.

Discussion

If your CPU is running at 100% in short bursts, I wouldn't typically be too concerned by that in most cases.

If it's a common thing for your computer, especially over long time periods, you may want to understand the associated heat and wear issues. Professional users, gamers, and other power users often design their own computers specific to their anticipated usage.

Another suggestion: sometimes cleaning the filters/air intake really helps. If the airflow is obstructed, fan will try to compensate by rotating faster. — Ivan Milyakov, Mar 08 '20 at 22:30
HDDs and NVMe SSDs are particularly picky about temperatures. SATA SSDs seem to be more forgiving about temperatures. — Ismael Miguel, Mar 09 '20 at 15:33
Can confirm from personal experience that SSDs are vulnerable to reaching a write limit if they are used as swap for extended periods.... https://unix.stackexchange.com/questions/560611/overloaded-the-ram-now-disk-ssd-is-behaving-as-read-only-fixable/560616#560616 — Clumsy cat, Mar 09 '20 at 17:13
Disc read/writes don't use the CPU. That's offloaded to... one of the bridges (north/south, I forgot. Probably south?) — Nelson, Mar 10 '20 at 10:27
@Nelson: Yeah, disk I/O doesn't tend to use too much CPU time (with some disk-encryption schemes being an exception). The concern'd moreso be that, if a program's maxing out CPU time, then it may be performing disk writes too. — Nat, Mar 10 '20 at 10:36
One more suggestion: it may also be worthwhile to undervolt the CPU using Intel XTU if your processor supports it. My old 7700HQ used to quickly reach into 95-102 range and thermal throttle running at 100%. Now I'm running with an 0.08V undervolt and it only rarely reaches 95 max on one core and 92 max on all other cores, almost never thermal throttling. As for stability it's processor dependent, but in my case I've found 0.08V to be stable and error-free for over a month of running 100% 24/7. I've tried 0.1V undervolt previously but it had stability issues after a week or so. — SamYonnou, Mar 10 '20 at 18:45
disk drives for datacentres are designed for higher utilisation — JCRM, Mar 10 '20 at 20:22
I will add "Thermal paste" dry-off in bold letter. It happened to me with an old laptop. I didn't know any such thing exists, so I damaged the laptop case before realizing the actual issue. — , Mar 10 '20 at 22:00
As to 1.5: IME, HD temps are likely critical to their lifespan. I've been using active (heatsink + fan) and passive (fan blowing on drive) cooling on my personal desktop drives for +15 years and haven't lost a single one, including a 80GB drive I still use as an OS drive. I don't use expensive drives with some being Maxtor ATA133 IDE drives. I have one mid-tower "server" running 5-6 of these drives. As a computer tech, I've seen most uncooled drives die in 3-5 years, usually right after their warranty ends. — computercarguy, Mar 10 '20 at 23:46
Older SSDs that tried to push the speed envelope definitely had heat issues, I've seen one or two that had a scorched controller chip, but they've benefited enormously from process improvement and general design refinement. Running hot is always just a question of which component will fail first. Even batteries fail faster from excessive heat! — SilverbackNet, Mar 16 '20 at 23:04
An extra concern; the plastic parts got hydrolysed faster by heat and sticky, you might feel them horrible. — Sadaharu Wakisaka, Mar 17 '20 at 02:23

score 57 · Answer 2 · edited Mar 08 '20 at 15:48

57

In theory, so long as a computer can keep itself within tolerance as to temperatures, then it should be able to do that 'forever'.

In practice, laptops tend to struggle with the keeping cool part and even some desktops have poor airflow characteristics which will result in long-term heat build-up.

If your machine can stay within tolerance, then all you are really doing is shortening the life of the fans.

edited Mar 08 '20 at 15:48

Community

1

answered Mar 07 '20 at 18:32

Tetsujin

47,296
8
108
135

12

The Dell XPS 15 is known to be doing well under heavy loads, but will get loud, so yeah, this is pretty normal. My sources, Linus Tech Tips and Dave 2D. – LPChip Mar 07 '20 at 18:35
8

Even good laptops at 100% will hit maximum temperature - we run Precision laptops at heavy CPU loads and have plenty that have done this day in and out for years. The chips all hit Tj(max) and effectively run at 100C most of the day, most days of the week. I've never seen a system fail because of it - at least not CPU/RAM. The chip will throttle to maintain 100C but, as far as I've seen, the chips are totally fine to run like that. – J... Mar 08 '20 at 17:44
5

Even when you keep eletronic parts within tolerances, you will still shorten the lifetime of most of them. For just one example you can check out Google's hard drive study to see how temperature effects lifetime in practice. Or see [this](https://www.xppower.com/resources/blog/electrolytic-capacitor-lifetime-in-power-supplies) for some nice graphs and formulas. Longeviety definitely suffers due to heat. It's definitely a long-term thing though. – Voo Mar 09 '20 at 08:30
2

Be aware that 3rd-party "laptop coolers" are available cheaply that can help mitigate temperature buildup and therefore noise. – xorsyst Mar 09 '20 at 10:35
12

Also note: if your computer *can't* keep itself cool at full speed, it will either slow down, or turn off - it shouldn't melt, or anything like that. – user253751 Mar 09 '20 at 11:43
1

@user253751 Yes, thanks to the [Catastrophic Shutdown Detector](https://manualsbrain.com/en/manuals/1246877/?page=606). At least, that's what Intel calls it. – Mast Mar 09 '20 at 13:39
2

@Voo CPUs are very different from hard drives, and especially modern Intel CPUs are almost bullet proof. I've seen a lot of computers live and die, I've seen a lot of components live and die... CPUs almost never, ever die. No matter how hard they've been used. – J... Mar 09 '20 at 14:06
1

@user253751 That's certainly true now. There was a time 15-20 years ago when you could easily toast an AMD processor by simply lifting the heatsink. Magic smoke + dead chip within seconds. These days it's very difficult to accidentally roast a CPU, though, that much is true. – J... Mar 09 '20 at 14:07
2

@J True, CPUs are rather resilient these days. But if CPUs and GPUs produce a lot of heat that also increases the environment temperature inside the case. This is particularly true for laptops. It's not as if the hard drives in Google's data centers are responsible for their temperature. – Voo Mar 09 '20 at 14:44
@user253751: nearly no laptops can keep themselves at max turbo, if they support turbo boost. The whole point of turbo is to take advantage of thermal inertia and run bursty workloads faster than the steady-state CPU frequency your lightweight cooling can sustain. [Why can't my CPU maintain peak performance in HPC](https://stackoverflow.com/q/36363613) shows an extreme case (Lenovo Yoga-3 with a Core M-5Y71 CPU; very low TDP that can still turbo high.) – Peter Cordes Mar 09 '20 at 17:13
1

@LPChip Maybe XPS's today are better but not my experience in the past. Source: Had 3 XPS 15's over ~7 years. – xiaomy Mar 09 '20 at 20:51
1

@xiaomy and you are 100% sure it was the CPU that died on each occasion, because you overclocked it and used it extensively? – LPChip Mar 09 '20 at 21:14
@LPChip I don't really understand your question. None of my XPS's died but they all suffered from frequent throttling as a result of bad design or material used. – xiaomy Mar 10 '20 at 19:50
@xiaomy well yes, that is exactly my point. They are designed to throttle, so no harm is done, which is what OP is asking about. – LPChip Mar 10 '20 at 21:25
1

Most components that can fail due to temps have been replaced with those that won't, such as capacitors. Now the problem is (still) heat expansion, where when the parts heat up & expand or cool & contract at different rates, causing stress fractures. Keeping an even temp, you can potentially let your computer run "forever", but doing that isn't going to be easy. If you shut if off or even let it sit without doing constant calculations, it'll change temps causing potential stress, and so will heating it back up. Smaller temp ranges are good, but hard to maintain. – computercarguy Mar 10 '20 at 23:52
@LPChip Yes probably no harm done. But I wouldn't call a computer that throttles frequently under heavy load and thus impairs its performance greatly "doing well under heavy load". Again, don't know about the current XPS. Just talking about what I've had. – xiaomy Mar 11 '20 at 14:01
@computercarguy Do you mean capacitors don't suffer usable life reduction due to heat? Because that's absolutely not the case. See [here](https://www.xppower.com/resources/blog/electrolytic-capacitor-lifetime-in-power-supplies) for some graphs and formulas for electrolytic capacitors to name just one example. Or you can simply look at Google's data for HDDs that shows life reduction for high temperatures (with no changing temperatures). – Voo Mar 12 '20 at 13:23
@Voo, I was saying that most of the electrolytic caps have been replaced with solid state caps that aren't as subject to heat problems. And HD's keep getting better, even at heat management, as with everything else. I've had active or passive cooling on my HD's for the past +15 years and haven't lost one, even the 80GB Maxtor 133 IDE I use as an OS drive for my mid-tower "server", which is 1 of around 5 sub-TB drives in that machine. – computercarguy Mar 12 '20 at 16:07
@computer It's mostly an academic issue, particularly for the given scenario since we're talking about 4 hours of high load for a few days. But I'd be pretty confident that if you took two laptops and let one run at 100% CPU and GPU in a small enclosure 24/7 and had another one working modestly at reasonable room temps, that the first one would die much sooner on average than the other. After all even solid state caps are pretty strongly affected by high temps, it just doesn't matter too much due to their long lifetime. Also I've seen cheap MBs with electrolytic caps not too long ago. – Voo Mar 12 '20 at 21:03
@Voo, it depends on how old and how abused the components are, too. A gaming rig that's used heavily might die quicker, or it might be hardened against this usage. An older business desktop might have really cheap components that won't stand up to the abuse for any length of time, especially if the electrolytic caps have dried out. Looking again at the specs for the OP, it's a bit of a tossup. Decent specs, but low star reviews, and I'm not fond of Dell because of their lack of quality. If this was a personal machine, I'd be worried about remaining life, but if it's business, blow it up. ;-) – computercarguy Mar 12 '20 at 21:24

score 13 · Answer 3 · answered Mar 08 '20 at 23:20

There's nothing to worry about, but download Intel Extreme Tuning Utility and see if you're getting throttled.

It shouldn't normally be a problem on that specific laptop, but there are power limits as well as hard thermal limits. On my laptop, for example, I can run 100% CPU all day long at 2.0 GHz. But, if I let it spike up to 2.2 GHz or higher, within a couple minutes I start getting throttled due to power limiting, and then my laptop drops itself down to 400 MHz. From what I've read, this is common across several manufacturers on the 14" class laptops.

The point being... it's counter-intuitive but sometimes you can get that workload done faster by lowering your limits. Use XTU to find out for your specific use case.

score 8 · Answer 4 · answered Mar 09 '20 at 11:04

Not entirely without worries. Though I guess you can probably do it, if you don't plan to do it for months.

There are different aspects to consider, the most important one for this particular notebook being that its cooling is rather insufficient for that purpose, so it will get hot (generally an issue with notebooks, they're not designed for that purpose).

If you google for e.g. "stress test", you will find that under high load the CPU reaches around 90-92°C (which is acceptable) and throttles down to 1.5GHz. The notebook itself reaches peak temperatures on the outside of 50-55°C. Which is, well, not that awesome.
If it has that temperature on the outside, then that necessarily means it has that temperature (and presumably another 3-5°C more) on the inside, too.

For some components (the CPU, notably) that's a total non-issue, but for other components, it's not so awesome if a long life is desired. So... yes, this may be a problem.

Other issues with very high loads over long time that one might in general have to consider is the cooler fan's lifetime (running at max speed all the time), and generally the CPU overheating (used to be an issue especially with cheapish cooling, but modern CPUs make this problem almost non-existent by throttling down accordingly).

Another issue for very high loads on laptops in general is the (usually cheap) power supply unit running somewhat on the upper end and croaking after short time. That shouldn't be an issue with this particular notebook since its PSU is designed to accomodate not just the CPU but also the GTX-1050Ti (which is idle during Matlab use), and if the specs are truthful should be at around 40% load only.

_{Also, even when operating within thermal limits, there is degradation which is proportional with use, although whether your CPU lives 35 years or 17 years probably makes no practical difference (since a capacitator on your motherboard will die long before that time anyway, or something else).}

score 6 · Answer 5 · answered Mar 09 '20 at 20:37

If I planned to do intensive CPU-bound processing on any computer, I would use a "server-style" computer to do it instead of a laptop. In my opinion, laptops really are not designed to "run hot." I think that you need to use a computer with a larger case that is truly designed, from the ground up, to be "a server." One that is designed to be able to siphon-away all of the heat that might be generated by the electronics when everything is running full-bore for hours or days on end.

I am looking into this option thanks ! MATLAB has this thing called MATLAB parallel server , so I can use their servers instead of my PC — aadil095, Mar 09 '20 at 21:31

score 5 · Answer 6 · answered Mar 16 '20 at 15:03

Short answer - extended periods of your CPU hitting 100% will reduce performance and shorten the life of your CPU. Why? Because 'Electromigration'

Basically you have billions of transistors in a modern CPU and billions of electrical pathways which degrade over time. Some paths will short and others will cause corruption. In either way, your CPU will carry on regardless just at reduced capacity. Consider a road network where a crash happens at a major junction. Does all traffic cease to move? No people just take different routes - the same is true of your CPU. If it wasn't then the merest imperfection or path failure would cause a CPU to fail and the lifespan would be measured in days rather than years.

Running a CPU under high load causes additional heat and also 'electrical wear' on the pathways which lead to further degredation and the further need to run at excessively high percentages to counternance the failed pathways etc.

The question is how much of an issue is it? Probably pretty low. Anecdotally I've not seen CPU's diminish in performance enough to be noticable during their active lifetime (i.e. within a few Operating System releases). Of course a Pentium 2 is going to perform much worse than a brand new i7 - but would it be chronically slower if it ran at 85-90% of its speed when new? Well, probably not.

+1 because the observation of the CPU degrading is both technically true and conceptually interesting. That said, it probably isn't a big issue in most cases -- extreme overclocking being a possible exception. — Nat, Mar 17 '20 at 19:16
Seconded: Also, see https://en.wikipedia.org/wiki/Electromigration. It used to be something that people worried about 40 years ago, mostly when upgrade cycles were slow (so the thinking was that a CPU had to last for a long time). Nowadays we chuck stuff out after a year or so. I suspect the design margins built into gates make them more susceptible these days than in the past. Still something to consider for specialised, long lifed, impossible to service, mustn't ever stop systems. — bazza, May 12 '20 at 18:20

score 4 · Answer 7 · answered Mar 09 '20 at 18:28

Beware of dust!

A laptop (or a desktop) that rarely starts its fan and runs quiet most of the time can tolerate a rather dusty environment quite well.

A few hours with the fan maxed out in a room full of, ex. wool, is a recipe for blocked cooling system and a thermal problems ranging from throttling all the way to starting a fire.

score 2 · Answer 8 · answered Mar 17 '20 at 01:13

Welcome to the exciting and dangerous world of high-performance computing. :-) Basically, there are two aspects to consider: a) whether your hardware components are designed to run at high load continuously in the first place, and b) what happens to your computation and data if/when a component fails anyway.

The first aspect has to do with the fact that cheaper components mean a better margin for the hardware vendor. If the average laptop under the average load needs to spin its fans for only 1% of the time, why use expensive heavy-duty fans that would outlast the laptop as a whole? The same goes for the hard disk or SSD. A hard disk is a mechanical device and subject to physical wear. An SSD suffers from electric wear. In both markets there are models for heavy-duty use (often marketed as "datacenter-grade") and they are expensive of course. Lastly, motherboard and power supply unit components such as capacitors can also vary in quality. There used to be times when cheap motherboards were notorious for ending up with blown capacitors when the CPU had to draw maximum power for days on end. So normally those involved in high-performance computing would just build custom desktop PCs from reputable components. When I used to do that, there were proven brands and models to use for things prone to physical or electric failure when under load: the motherboard, the CPU fan, the power supply unit.

Today there are laptops marketed for high-performance use, but honestly I would take that with a grain of salt. Mechanical components are just too tiny in laptops and so likely to wear down and fail much sooner than the bigger desktop components. Also, the laptop internals are really cramped, so the air ducts are more likely to get clogged with dust etc.

In any case, one cannot rule out a hardware failure completely even if using top components. So it's a good idea to plan for a failure. Would you be able to resume your computation that has run for days or weeks when the CPU overheats and freezes? And what about the disk storing your computation results? So normally one would want to a) use computation software that keeps state and is able to resume from it, and b) use some level of computationally cheap redundancy in the storage subsystem such as RAID1.

score 1 · Answer 9 · edited Mar 17 '20 at 19:21

My personal experience with laptops is that most hardware problems are battery-related. New laptops can run X hours on battery alone, but after a few months the battery starts to wear out and only run half of the original time.

In any case, no laptop can run for days relying only its batteries, so you'll have to attache it to the power supply for long run, which is a problem in terms of heating because charging the battery while the machine is running causes the battery to get very hot — which will overheat neighboring components. Due to this, I've burned quite a few laptops, and usually the problem started to manifest with the hard drive interface wearing out (I'd replace the internal drive only to burn it again).

You mentioned running a laptop "at CPU 100% for like 4 hours in a day", which probably can't be done using batteries only with many (if not most) laptops, unless their battery is brand new.

Some laptops can run without battery, using only the power supply from the charger, others won't start if the battery is not inserted. The former are better, because they won't overheat.

Other factors contributing to long-run heating are hardware dependent — e.g. the type of mass storage employed (SSD vs classic drives), as well as how the laptop's cooling system is designed.

For the above reasons, I think it's difficult to provide a universal answer for all laptops, without considering all their hardware, and whether they can be powered without the battery inserted or not.

Gabriel · Answer 10 · 2020-03-18T18:15:41.443

GRRRRRrrrr.... People make such a fuss over such a simple problem: it is not the CPU utilization (100%) that could kill your CPU, it is the temperature. If your fan/heat radiator can keep the CPU temperature way under the critical threshold, then you have no reason to worry. It is that simple!

What to do?
Just find out your CPU critical temperature and current CPU temperature. Compare them and decide if they are too close.
Consider that you will probably throw your CPU in your garbage bin (don't do that! dispose of it properly!) because of moral degradation BEFORE any physical degradation will start to show up.

Real-world
All those number-crunching computers and servers that are running for years most of them at 100%. They are discharged because of old age not because the CPU failed. Most computers that I have seen discharged were discharged because of old age.

Generally, I would rather worry about HDD failure than CPU failure. Plus, if the CPU fails, you just buy a new one. Not a big deal. If the HDD fails, wellllll... you can also buy a new one, but your data.....

Now, that being said... should I tell my odyssey of finding a proper fan/heat radiator for my 8 core CPU? ;)

score 0 · Answer 11 · answered Mar 19 '20 at 00:40

I would worry more about the fan itself that the CPU.

Moving parts are often the weakest link, CPU should shut down before it does damage but this can be inconvenient.

I'm speaking from experience, RabbitMQ ran 5% CPU, made the fan run and eventually killed my laptop. All that needed replacing was the fan but it took a while to get the part. When I got the part I replaced RabbitMQ, asap. If you can chill the machine sufficiently with external fans you'll find they are easier to repair.

If your fan was not making much noise before and now is I'd worry, order a new fan now, they are cheap.

score 0 · Answer 12 · edited Mar 20 '20 at 10:20

Not too familiar with Matlab or how you're using it but it may be worth checking resources dedicated to it. Sometimes a hardware issue is a symptom of a software problem. If you aren’t using best practices designing your jobs it could overstress your laptop.

Case in point: a few years ago I was trying to boost performance on a modest gateway running Vista. CPU was always pegged at 100. I bought a faster CPU and the issue continued. The room would actually get hot running basic legacy accounting software. Turns out this software hated Vista. Reinstalled with XP and I had a nice PC again.

Maybe you have some kinda loop somewhere in your code.

Can I run my CPU at 100% usage for a long time?

12 Answers12

1: Overheating issues.

2. Excessive wearing issues.

Discussion

Linked

Related