About 18 months ago I bought a shiny new i7 8-core PC and immediately installed BOINC to run 8 CPDN tasks. All was well until a few months ago, when on reboot I started getting BIOS "CPU over temperature" errors. I ignored these, but I did notice that my average credit for the PC was steadily dropping. Last week, for the first time, the PC powered off without warning. It rebooted OK (plus over temperature error), but this has now happened 3 more times. Some Googling suggested that if the CPU temperature exceeded some value the PC would automatically power off. I've just installed RealTemp, and it shows all four cores running at 100C. If I snooze BOINC the temperature drops to around 97C after a minute or so.

Could somebody please advise whether the problem is running BOINC, or whether it might just be something more mundane such as the PC getting clogged up with dust, reducing the efficiency of the fan. And does anybody know what temperature these cores should run at when BOINC is running (i.e. 100% CPU all the time)? Many thanks.

---

You might want to clean the dust bunnies out of your computer, i clean mine every month or so, and disassemble the heatsink on the GPU every 6 months,

Claggy

I found this link where the author concludes 100c as the max and recommends running under 80c.

http://www.pugetsystems.com/blog/2009/02/26/intel-core-i7-temperatures/

I suggest you read the whole article, lots of info there.

Personally, I would never ignore any error messages about CPU temp. I hear the CPDN projects are CPU intensive. You can change settings in the BOINC project manager to reduce the amount of CPU time the projects get. You can also change the % Processors used setting. Don't run all of the processors, cut it back for a while and see what happens.

Dust is definetly an issue too. I don't know what your environment is like but it doesn't take long for fans to get clogged. That's a lot of heat you need to dissipate so you gotta keep that air flowing.

Also, make sure that your computer is placed in a spot where it can get fresh air. If you have it jammed into a cabinet it can't breathe.

Just some ideas to start with...

---

And does anybody know what temperature these cores should run at when BOINC is running (i.e. 100% CPU all the time)? Many thanks.

In a well maintained system, one with several fans, filters and a clean environment, out of direct sunlight and not next to a central heating system, you may expect an older i7 to run at 40-60C under load. 40s when clean, towards 60s when dustier/dirtier. Idle temperature is always way lower, in the 20s, Celsius then. Max temperature is 105C. Go over it a lot and you can damage the CPU.

So yes, I'd take the thing apart and get all the dust out. Also probably check the amount of thermal goo between the CPU and heatsink. Did you apply it yourself, or did the whole thing come pre-packed?

---

My dual core laptop runs just shy of 70c... but I guess laptops run at higher temps from what I read online.

---

Thanks for the replies. It seems that 100C is excessive then. I'll try reducing the amount of CPU it's allowed to use and see if that helps. If it cuts out again I'll take the cover off and give it a good clean out. (It's a tower and was purchased from a well regarded supplier BTW.)

---

Thanks for the replies. It seems that 100C is excessive then. I'll try reducing the amount of CPU it's allowed to use and see if that helps. If it cuts out again I'll take the cover off and give it a good clean out. (It's a tower and was purchased from a well regarded supplier BTW.)

I'd recommend doing the cleaning anyway. At 18 months you're probably due and if it's a nice system you wanna take care of it, right?

Do you change the oil in your car regularly or do you wait until the engine explodes?

---

Well after it tripped again I finally put the vacuum cleaner to it. It was a bit of a mess inside, and I cleaned out all the fans and grilles as well.

Before I cleaned it out, all cores were running at 100 with CPDN running. Snoozing CPDN didn't make much difference to the temperature - down to 95 at the lowest. (I've since read that once the temperature hits 100, circuitry slows down the CPU, which could well explain why my average credit had been dropping.)

After a good cleanout, it's now varying between 94 and 100 with CPDN running, but the temperature drops to 60 within 30 seconds if I snooze it.

So... Yes it was a mess, and I should have cleaned it out regularly. And CPDN does indeed seem to hammer the CPUs.

I'll monitor the situation to see if the RAC increases, and whether I still get any over temperature errors on reboot.

---

Still sounds a bit hot to me, if you can get it down another 20 to 30°C it would be better, (to around 70°C)

Claggy

Again many thanks for the replies. I will try some of the more advanced suggestions for cleaning it out at some point, starting with the brush technique. However I can now report that it is reporting trickles about four times faster than it did before I cleaned it out - and I no longer get over temperature errors when I reboot. So at least I'm part of the way there.

I hadn't realised that the circuitry throttles down the speed when it reaches the maximum permitted operating temperature, but clearly it does so rather dramatically! As they say a bit older and a bit wiser.

---

I use canned air to dislodge the dust bunnies while holding the vacuum cleaner nozzle nearby to suck them outta the air. (So they don't just re-circulate and go right back in...)

Some sites caution against letting your fans spin when blowing canned air into them and suggest only doing this while the computer is off. (good advice)

You will be suprised as how much dust gets in and around the fan.

I guess it's possible that the folks that built your computer didn't put any heat sink paste (not sure what it's really called) between the CPU and heat sink. Was it built on a Monday? ;)

You might also want to re-consider running CPDN. I thought that was geared mainly for systems equipped with GPU's.

Also, look at the top of the forums under "Automatic Temperature Regulation." Fred has a neat temperature regulating program that interfaces with the BOINC project manager. http://efmer.eu/boinc/index.html

Instead of waiting until your CPU thermal protection kicks in, you can set the software to shut down BOINC, suspend, or even shut down your computer. It's a nice piece of software, unfortunately I can't use it 'cause I run Linux now, but it should work for you.

Good luck!

---

You might also want to re-consider running CPDN. I thought that was geared mainly for systems equipped with GPU's.

No, CPDN is very firmly a CPU-only project, and will remain so for the foreseeable future.

@Richard, You're right, my bad. As CPU intensive as their projects are it seems that they would benefit from GPU crunching.

---

Intel mobile (laptop) CPUs are die-level cherry picked to (1) run unthrottled up to 100C, while (2) exhibiting low leakage current at idle to increase battery life (low supply and high demand explains why they are expensive). I'm pretty sure desktop cores run up to 70C before throttling. Going back to at least Coppermine, all Intel cores have thermal diodes which read core temperature at the hottest spot on the core, and throttle them when over this temp. Throttling is quite dynamic and will change quickly as the load changes. All Intel CPUs will run (slower) with clogged, incorrectly installed, poorly engineered, or missing thermal solutions. Not sure about current AMD cores, but at one time they became hand grenades under these conditions. Several years ago, Tom's Hardware created videos of both vendors' cores booting Windows with no thermal solution. The AMD core test required the tester to wear eye protection to prevent the exploding die from blinding him.

This is not meant as an advertisement of Intel products or a ding of AMD products. Rather, it is my reality based on several years as a hardware engineer running power viruses to measure max power of Intel mobile cores prior to launch, and write white papers with results so our OEM customers could design thermal solutions. I also measured max performance of AMD cores and smoked a few.

Also, the substance between the die and thermal solution is referred to as thermal grease. It is one of the few substances that has excellent thermal conductivity and low electrical conductivity. The distance between the die and thermal interface is critical. At one time, this gap tolerance was 8 mils +-3 mils. There may be newer thermal materials and different specifications now.

---