Safe long-term CPU temp?

15

11

I'm considering running a game server for some friends and I to play on with my laptop, but while the server is running my processor use tends to go to around 55-65% and my CPU temperature hovers around 70 C. I'm using a Dell Inspiron 17R SE. It has 8 GB of RAM, an Intel 3630 QM processor(2.4 GHz, Intel Turbo Boost goes up to 3.4 GHz), and an NVidia Geforce 650M graphics card. Is that a safe temperature for it to be running at for long periods of time(or even 24/7), or will that start to damage the computer? Thanks in advance for any help, and sorry if this isn't the right place to be asking this sort of question.

user320090

Posted 2014-05-03T00:33:18.093

Reputation: 159

Question was closed 2014-05-04T20:05:34.100

The safe long term temperature is any temperature within the specifications of the product. – Ramhound – 2014-05-03T00:50:29.080

So no damage can occur as long as it's below the maximum temperature listed? The specs listed for it say that's 105 C. I'm just concerned because I've seen wildly conflicting answers online. Some people have said anything above 60 C is going to start taking time off the machine's lifespan, but I've also seen people saying laptops are better at handling heat and anything below 90 C is okay. – user320090 – 2014-05-03T01:15:54.240

Theoretically, the lower the temperature the longer it lives. But processors almost never die and should not be of primary concern as long as it is in spec as @Ramhound alluded to. Chances are the PSU and hard drives will die long before the CPU suffers any real damage even at the upper threshold of stable temperature. – Austin T French – 2014-05-03T20:32:20.633

I can not agree with @AthomSfere conclusion about "processors almost never die". I have seen CPUs dying, and I have considered overheating as the most probable cause, due to symptons: several months record of high temperature readings. – Sopalajo de Arrierez – 2014-05-04T00:27:29.170

athomsfere is correct about lower temp also those # that intel gives out are a "risk management" calculation. intel expects the cpu to last say 5 years so they rate it for maximum temp with a % for saftey that they expect to allow it to live past 5 years. Just turning something on shortens its life, measurable % or not it will ware out. The question is will that affect be soon enough for you to care or notice. – Kendrick – 2014-05-04T00:32:51.170

I just noticed you mentioned a Higher nvidia card. I would be more worried about overall temp causing problems with the video than the cpu. The dell M series has a seperate nvidia card and it burns out all the time due to improper design/build and poor fan settings for bios. Yourse has onboard video which is about an inch away from the cpu or so, it also shares the heat pipe with the cpu. the more heat one puts out the less cooling the other gets. – Kendrick – 2014-05-04T00:41:17.417

@SopalajodeArrierez It is almost never die, compared to other components. I can say in the thousands of PCs I have fixed I have seen everything else fail various orders of magnitude more often than a CPU. Even with cooling issues, generally the motherboard bites it before the CPU. Cooling should always be a concern, but the CPU as the first component to fear failure in is unfounded. – Austin T French – 2014-05-04T00:53:26.797

Answers

29

As far as I know (if someone has precise data, please report), there are no serious statistics data about modern microchips lifetime estimations as a function of its running temperature.
I know of two reasons for this:

  1. When we could know data about lifetime estimation of a microchip technology, this is, after years from manufacturing, that technology is... obsolete.
  2. Only microchips corporations could be interested in researching to obtain such precise info about their products (or the competitors ones). And they are not willing to share it; even if they do, I wouldn't believe them very much.

So, I believe that end-users only have the (often intuitive-only) knowledge of experienced IT specialists.
This is the mine:

  • Microcircuitry engineering is something like cooking: it involves a lot of probabilistcs and will often have a rather random results. So, you don't know how good a microchip is until you have fabricated it. Even then, deterioration will have too a bit of probabilistic behavior.
  • 40ºC (104ºF) or below is heaven for every microchip.
  • 50ºC (122ºF) is a not bad temperature for any microchip.
  • Microchips starts getting damaged on its lifetime at 60ºC (140ºF).
  • A chip running at 70ºC (158ºF) during 24 hours and 7 days a week, will probably last 2-6 years.
  • A chip running at 80ºC (176ºF) during 24 hours and 7 days a week, will probably last 1-3 years.
  • A chip running at 90ºC (194ºF) during 24 hours and 7 days a week, will probably last 6-20 months.
  • In this matter there is no difference between main computer chips like GPU, CPU, Northbridge, Southbridge... etc.
  • Given a temperature, it is harder for the chip to maintain it at high processor usage than at low processor usage. For example: a CPU that achieves 70ºC (158ºF) during 10 hours on nearly-inactive Windows desktop suffers less than a(nother) CPU that achieves 70ºC (158ºF) during 10 hours of intensive CPU processing (i.e: SuperPI). Some hardware engineers report this could be due to that in the second case the CPU uses most of the microcircuitry, and in the first case only a small part of it.
  • The general rule: microcircuitry is like an ellectrical printed circuit board that has the tracks very close between them (there are often only 4-5 molecules between two tracks), so heating is slowly melting the tracks as time goes by. Keep things as cold as possible.
  • The general rule when reading the manufacturer's data: they want for you not to care about refrigerating anything, because then it will get broken just after the warranty period (sometimes only a few weeks after it; it is incredible, I know). "It is just bussiness", Alcapone dixit.
  • Preventing is important (better than waiting for failures to repair): when things start to fail, it could be due to tracks melting in the microcircuitry, or due to minor tracks dilatations. The second case is a temporal problem. The first one is probably a definitive one.

  • History: Changes in this document (read only if this is not the first time you visit it):
    • 03-05-2014: Added ºF temperature data conversions.

Sopalajo de Arrierez

Posted 2014-05-03T00:33:18.093

Reputation: 5 328

Thank you for the extremely detailed response! That's good to know. This laptop is starting to get a bit dated for my uses so I was planning on upgrading to a desktop fairly soon, so it still having a lifespan of 2-6 years is fine for me. I've seen some people online saying that laptops typically handle high temperatures better than desktops, though- can you comment on the veracity of that and if that would change your estimate of 2-6 years? – user320090 – 2014-05-03T02:57:15.007

you have to look at the cpu specs and verify what the maximum temp is for your chip. my 12core is 58C i believe. I can get it to 55C+ with about 10 threads of prime or another cpu burnin test. In my case currently i cant use over 75% of max cpu if i want to keep my processor alive. I would never run max temp either. another thing Sapalajo didnt mention is that any flaws in the chip will likely be the first to be affected so you may have stability or other quarky issues because that 1 part of the chip couldent handle the heat. – Kendrick – 2014-05-03T04:12:13.757

I have not found (based only in my experience as an IT specialist) a better behavior in laptops or embedded devices for the case of high temperatures. I would rather say that "Don't worry, laptops can handle heating." is some kind of a marketing trick. Underlaying microcircuitry technology keeps practically the same. And cooling a laptop is not as easy as a desktop; even dust cleaning (an important care if you worry about overheating) is harder to do. If I would have to make my bet, I would even say that an overheating desktop can endure a bit more than an overheating laptop. – Sopalajo de Arrierez – 2014-05-03T10:29:19.770

I agree with @Kendrick: sometimes the problem with overheating is not in computer/device not booting, but in awkward unstabilities that give you headaches to discover where does the problem comes from. Furthermore, if you have, say, a graphics card with thermal sensors on GPU, VRAM and VRM (3 sensors) you must worry about refrigerating all of them, because the overheating in just one of them will probably affect the entire behavior of the card. Same could be told if you have a CPU with, say, 5 thermal sensors. – Sopalajo de Arrierez – 2014-05-03T21:29:36.513

I would say that current cpu's can be intelligent enough to throttle down assuming the os and every thing works correctly. had a dell tech forget the heat sync on a D620 systemboard swap. the pc ran for about 6 months before the user complained to us. PC was running wicked hot [scalding on bare skin] and the os had throttled to 200mhz. It is possible for them to save them selves from cricical failures but I can't immagine how much service life that removed. on the flip side surviving is very different than preforming well. – Kendrick – 2014-05-04T00:13:14.883

1

I'm going to cover hardware first and then towards the end what you can do if you're stuck with that hardware as your only viable option right now.

My last CPU was an X3 720 (that unlocked to an "X4 20", it idled at 150 °F (66 °C), and I purchased it the very first day AMD 45 nm CPUs were released.

I'm now running an AM 8350 eight-core running at 4 GHz, and it always runs at room temperature.

To understand how CPUs and heat correlate you need to understand two things, architectural design and manufacturing design. AMD and Intel are both architectural CPU designers though AMD is an underdog and split off their manufacturing into an independent company called Global Foundries in order to stay competitive with Intel who is ahead with manufacturing technology by a full node.

What is a full node and what is a half node? A full node is a mainstream technology where the sizes of the silicon walls are shrunk. The smaller the CPU/GPU die the shorter the distance and the more is done by electrons flowing through the silicon corridors inside the CPU which also increases efficiency.

  • 65 nm Circa 2007.
  • 45 nm Intel late 2007 / everyone else late 2008.
  • 32 nm Intel January 2010 / everyone else 2011.
  • 22 nm Intel 2012/everyone else 2014.
  • 14 nm Intel late 2014/early 2015 / everyone else about 2015 (if we're lucky)

Now you also have to think about CPUs being made a lot like Gramma baking cookies... they're not all made the same. Some just don't get built as well as others. My socket AM3 quad core couldn't clock from 2.8 GHz to anything beyond 3.1 GHz and change while my already blazing fast 4 GHz eight-core beast easily goes to 4.444 GHz before needing voltage to continue overclocking (haven't messed with that yet). Keep in mind that my 32 nm 4 GHz oct-core CPU was on a mature/refined 32 nm process, not when it first came out.


CPU Cores / Watts

My friend has the same CPU and his house was 48 °F (9 °C) one "winter" day in Florida...his CPU was also 48 °F (9 °C). Now that won't matter when you have a high/full load on your CPU. Our 8350s have a 125 watt TDP (total power draw). The more they work the hotter they'll get and that depends on how you configure your OS and applications.

You also can't forget that with Intel; unless you drop $600 you're stuck at four cores tops while $200 got me eight cores. If you're talking load balance for example, Firefox will occasionally freeze and use 25% of a quad core's CPU cycles, on my eight-core CPU it uses only 12.5% load. So take a CPU's TDP and divide by its core count. 125 watts / 8 = 15.6 watts per CPU core. Keep in mind it's not all CPU cores though it should give you an understanding. An Intel 3630 QM's TDP is 45 watts, 45 watts / 4 = 11 watts, but keep in mind it's a mobile CPU. Most desktop Intel CPUs run at 77 watts now, 77 watts / 4 = 19 watts per core.

If you want to reduce the heat my hardware suggestions are as follows...

The more cores you have and the lower the TDP per core the better.


Software Configuration

Kill the Superfetch program and kill off anything that sucks CPU cycles is your best bet. You're likely stuck with this as your only option (unless you can buy some new hardware).

Run msconfig, go to Services, hide all Microsoft services and disable things except your anti-virus. Go to the startup tab and do the same (there is no checkbox there for Microsoft items though).

Also keep in mind things like Windows Updates (e.g. the MS .NET language compiler is a HUGE CPU hog) will spend an hour or two at near 100% CPU usage.

Make sure you have more RAM than you need, kill off the pagefile (keep the hard drive from grinding to death and increases system performance simultaneously), and you'll get the most out of your system.

John

Posted 2014-05-03T00:33:18.093

Reputation: 1 490

An interesting info, @John, but I have a question: do you mean Windows Update spend an hour at 100% CPU when installing new updates? I have rather found that updates tend to be small, and get installed in just a few minutes. – Sopalajo de Arrierez – 2014-05-03T16:07:24.087

John the page file is incorrect. MS states it may be possible with lots of ram, but I don't know of many apps to this date that behave properly. Firefox was running @ 1.5gb the other day and I only had 20-40 tabs open. I have 8gb of ram and even when running Linux I have been stuck having to have a swap file due to bad apps. adobe flash and pdf are the first that come to mind. Also some apps are written in such a way that they don't run right if there's no swap – Kendrick – 2014-05-04T00:52:09.340

@Kendrick Not incorrect, if you want performance then you should be aware of how much RAM you're using and thus how much RAM you should have installed. Windows does have erroneous behavior where it actively leaves items in the RAM, you can view these items using RAM Map: http://technet.microsoft.com/en-us/sysinternals/ff700229.aspx

– John – 2014-05-04T04:14:26.077