CPU was running at 85-96 C for several hours - is that OK or not?

I've gotten a couple BSODs on my HP ZBook 17 G2 (Windows 10) laptop recently, and I'm not sure if it's hardware or software related. Windows 10 did not generate any dump files or event logs for some reason.

So I've been running some benchmark programs just to make sure my system is stable. I ran Prime95 for a couple hours and the CPU temperature got up to the high 80s C. I also ran Aida64 for 4 hours and the CPU temperature ranged from 85 to 96 C.

Now I've been reading that you don't want the CPU to go over 80 C. Not sure if that's just for desktops or if it's also for laptop CPUs. Also, my CPU is a 4910MQ, which is a Haswell, and I've read that running Prime95 on a Haswell CPU might damage it? Just to be clear, I'm not overclocking anything, just using default settings and I never touched the CPU.

Do you think running my CPU at those high temperatures for several hours was OK, or did I damage or shorten its lifespan?

 

First, you don't need to run high stress benchmarks for hours on end

5-15 minutes should be enough, then switch to gaming or app's you use and test under real use.

You won't damage the laptop at the temps you are seeing, even for hours of use, I have run for weeks at full 100% CPU 100% GPU and laptops are running many years later.

If you get a flaky on the edge GPU / CPU or VRM / etc, then you can kick it over the edge into failure mode. It's rare, but it happens, RMA it.

The CPU / GPU's are rated independently by Intel / Nvidia / AMD, the CPU's are going to thermal throttle around 93c and shutdown at 100c.

If you are getting BSOD's, it's likely Windows 10 + Nvidia driver issues, Nvidia drivers and Windows 10 have been "challenged" lately being stable for many under gaming loads.

Read the Nvidia forums for each new drivers release, and here on NBR - search for the version in the title of threads to find the new ones - and see what others are experiencing.

And, don't forget to check the support forums at Game makers and on Steam for more relatable feedback from other game users on Windows 10 with Nvidia drivers at the same version - and hints for which versions do and don't work well with the game.

MS is pushing updates all the time, not with the best result - new drivers from many OEM's, not just Nvidia, and Windows updates - OEM's are chasing MS changes, and MS are chasing bugs - making changes to fix bugs - and creating new ones

If you can manage it, I'd switch to Windows 8.1 or Windows 7 for stability. Then you are only chasing Nvidia bugs.

 

People recommend running below 80C because it is speculated that exceeding 80 C over a long period of time (months, years) will adversely affect the lifespan of a CPU. In other words, you are totally fine as long as this is not a common occurrence.

 

It will decrease absolutely every lifespan in a laptop, such high temps on a laptop ruin everything (remember a laptop is essential a encapsulated heater blowing hot air like a bloody chinesium tornado).

Also @hmscott IIRC tjmax on laptops for CPU and GPU is 105 degrees, which is insane, but I guess reasons...

 

Tjunction Max = TJUNCTION
Junction Temperature is the maximum temperature allowed at the processor die. (Intel)

100c for 6700HQ / 7700HQ / 6820HK / 7820HK, etc etc
Under Package Specifications:
TJUNCTION 100°C 100°C 100°C 100°C
http://ark.intel.com/compare/97464,88969,97185,88967

 

Did not Intel a year or two ago say that the temp sensor in the cpu can be off by up to 10c

John.

 

You'd be surprised at how tough a laptop CPU is, sure it will lower the overall lifespan, but it is much more likely that something else will break before the CPU. I ran my M6700 hard regularly for over a year and did not experience any adverse behavior. Usually, on the BSOD itself, it will give you an idea of what the problem is.

 

I think it was that OEMs are able to adjust the tjmax lower (by up to 10 degrees?).

 

Knowing how they actually measure the temperature would be interesting too. Is is an actual thermocouple on the die which they calibrate, I would venture that such a device would be basically made along with the rest of the CPU and just part of the architecture of the CPU or is it a model that uses other variables as input?

 

Running at 80-90C permenantly isn't good.. It degrades the silicon and can cause failures..

In such a case, a repaste of CPU+ GPU is due.. And while your at it, replace the thermal pads on the GPU with some good ones like Phobya XT

Sent from my LG-H850 using Tapatalk

 

i doubt silicon is that un-durable. When Intel states that max temp is 100c, I expect they tested the cpu by running it for thousands of hours at full load, without issues.

The question is whether the motherboard can take the radiant heat.

 

Yup, in the really old stuff, with constant use, the components approach mummification and brittle, and so does the PCB.

It's best to cool things down, but you can still get long life out of hot running components, running them till they expire.

 

It's often the heating and cooling cycles that age components and connections, much like an incandescent bulb left on will tend to outlive one that's turned on and off frequently.


Sent from my iPhone using Tapatalk

 

Yup, the constant expansion and contraction of metals, ceramics, plastics, board materials over time, riddling them with cracks and voids.

Nothing like working on stuff from ancient times, and having a component go to "powder" at the touch

 

The answer to that question is yes. The CPU may be at that temperature, but the heatsink temperature will be lower. Also, radiant heat can be summed up at the simplest q=σ(T^4-Tsur^4) where σ=5.67x10^(-8). Basically, the sigma constant is so small that it takes a very large temperature difference for radiant heat to be significant. I'd be more worried about conductive and convective heat transfer affecting the lifespan of components than radiant heat transfer. As a rule of thumb, in chemical engineering, unless the temperature difference is in the 200°C-250°C range (edit: or higher), radiant heat transfer is often ignored in calculation because it is order of magnitudes lower than any other heat transfer modes. It is not 0, but it is still very small.

Also, went and dug a bit into and this is likely the method used to measure CPU temperature: https://en.wikipedia.org/wiki/Silicon_bandgap_temperature_sensor. So yeah, the sensor is in the die design.

 

A bit of personal background first;
I used to work in electronics manufacturing and had the QA organization as one of my primary responsibilities. Part of what happened there was to test new products to determine their environmental specifications (to set and verify the limits of operation and storage) and for product testing and "burn-in" of newly manufactured products before they were sent to the end users.

If a product is rated at 100 C it means that it can sustain that temperature throughout its entire life-cycle at an failure rate lower than the rated specification. There is an entire world of statistical thermodynamics devoted to determining stress temperatures, duration, accelerated aging and life expectancy.

If you want to dig in to it, look up "Arrhenius equation".

Roughly this means that for every 10 C of temperature increase there is a doubling of reaction rate; this applies to chemistry and materials science.

We would "burn in" products at the rated maximum temperature for 196 hours. There was a certain percentage of devices that would fail in that interval with the largest number failing early on. In large numbers (thousands of units) the failure rate would fall off to near 0 at the conclusion near the 196 hour mark.

--------------------------
Depending upon the complexity of the assembled product the system failure rate could be as high as 10% during burn-in. But what ended up leaving the 196 hour burn-in might only have a return rate of 0.1%.

---------------------------
In electronics; like a laptop, the most sensitive component to premature aging will be the power supply capacitors. Most caps are rated at -40 to 85 C for 1000 hours, some will be rated up to 105 C.

What happens to a cap is that its capacitance value decreases (it becomes less of a capacitor), electrolyte dries out and its ESR (internal leakage) goes up. For all of the things in a laptop I would be most worried about those components dying off first.

Using the rough guideline of the Arrhenius equation you can see how an 85 C cap might be rated at 1000 hours, but if you limited it to 75 C that rating increases to 2000 hours, (65 C, 4000 hours, 55 C 8000 hours....).

This does not mean at 1000.1 hours the components all die, it is a statistical number and the rate of failure increases beyond a certain threshold of what is acceptable. (how that is different, the components in a satellite will have a much higher reliability number than something in a toaster).
-----------------------

If you are a real nerd here is an applications note on capacitor selection. This article is about surface mount capacitors (similar to what is used in a computer). Look at the wide array of temperature ratings and read about how this impacts performance;

https://www.maximintegrated.com/en/app-notes/index.mvp/id/5527

There is a big price difference between a capacitor rated at +85 C and the same value rated at +125 C. Motherboard manufacturers make their profit margins by putting just the minimally functional component on the board, not the most expensive or highest rated.