CPU overheating with age?

[Macaw!]

Odd, my CPU was getting to 100C, so I bought a big CPU cooler (twice the normal size). That took it down to 80C, but it's gradually worked it's way back up to 100C again. Now I've slowed the CPU down to 85% of normal speed in the BIOS, and it's back to running at a more comfortable 70C. Something funny's going on here.

Yes, the heatsink is hot, heat is transferring correctly. The fan is working fine, hot air is blowing off it. The problem is the CPU is creating way too much heat. Faulty CPU? Faulty motherboard? Anything I can do apart from just replacing one or the other? The CPU is an i5 3570K and has been run flat out for years (BOINC).

The BIOS won't let me lower the speed any more. I've reduced the base clock from 100 to 95, it won't go lower than that. It ignores me when I lower the multiplier (which is odd as it's a K CPU which should be unlocked?) But I have managed to disable the turbo function, so it doesn't jump up to 38 multiplier. So I've ended up with a 3.2GHz CPU instead of 3.8GHz.

 

[Robzere31]

I'd start by cleaning the fan & heatsink off of all the dust etc & replace thermal compound.

 

[doyll]

Could you tell us what case, case fan configuration, and other components in the case please? Could be any of a number of things, for example it's possible the case intake vents have filters on them and they are full duct thus lowering case airflow and causing overheating.

 

[Macaw!]

Not noticed any instability. Voltage reading is correct, so I assume the motherboard's VRM isn't broken. The heatsink feels nice and cosy, so I think it's pulling heat off it ok. I'll take it apart now and stick new thermal paste on it. I've only had the new cooler for a few weeks, so the paste can't have degraded, unless it's really rubbish and has oozed out?

Just a generic thing bought elsewhere. But it looks twice the size of the stock Intel one I replaced:
*** No competitor linking - please read the rules *** - Thanks, Armageus

It's a big Xcase with LOADS of cooling, big side fan etc, no dust, recently cleaned.

 

[Armageus]

I'd start by just checking that the Heatsink is actually seated correctly, as that could account for the temperature increase e.g. if it has come loose.

However - the possible issue is that it is just a cheap generic heatsink - there is no guarantee that the base is completely flat, or that the design (which will be a poor copy of an existing heatsink) of the heatsink actually works adequately.

If reseating it doesn't get the temperatures under control, I would look at replacing it with a heatsink from a reputable manufacturer - something like either of these:

My basket at Overclockers UK:

• 1 x Raijintek Aidos Direct Contact CPU Cooler - Black= £19.99
• 1 x Cooler Master Hyper 212 Evo CPU Cooler= £28.99

Total: £58.88
(includes shipping: £9.90)




​

 

[Macaw!]

I would have thought £11 was enough for what's essentially just a block of metal and a fan. The part of it in contact with the CPU is about 50C when the CPU is 65C, so it seems to be taking the heat off it ok. This is then fed to the fins with two copper pipes, and the fins are about 40C. Sounds like everything is working ok, so who knows!

Just done some tests:
CPU core was 70C.
Thorough dust clean (of case intake vents - only 2 weeks of parrot dust!): 65C
Clean off paste and replace with a heatsink PAD (cut to size): 105C!
No paste at all, bare metal onto bare metal: 80C
New paste (not for CPUs, just generic electronics stuff (silicone) I happened to have (made by electrolube (sounds kinky!)): 65C again, same as paste from the cooler.
I'll look up some reviews and see if expensive paste is actually any better, as it does seem like it can make some difference, and pads are clearly utterly pointless, you're better with nothing at all!
.....According to reviews, expensive pastes like Arctic Silver only give you 2C difference, and the really good ones are electrically conductive! Not something you want oozing out onto the motherboard! I'll stick to the standard silicone paste.

 

[Armageus]

I would have thought £11 was enough for what's essentially just a block of metal and a fan.

You would think so, but quite a lot of R&D goes into heatsinks to improve the thermal performance through different technology and designs - cheap generic ones aren't forced to apply any of that, and just copy existing designs without understanding how/why they work, or without the quality control to make them work efficiently (e.g. only takes for the base of a heatsink to be slightly concave or slightly convex, and you have a differing amount of contact)

This is then fed to the fins with two copper pipes

The copper pipes are heat pipes and even they are fairly complicated devices https://en.wikipedia.org/wiki/Heat_pipe

New paste (not for CPUs, just generic electronics stuff (silicone) I happened to have (made by electrolube (sounds kinky!)): 65C again, same as paste from the cooler.
I'll look up some reviews and see if expensive paste is actually any better, as it does seem like it can make some difference, and pads are clearly utterly pointless, you're better with nothing at all!
.....According to reviews, expensive pastes like Arctic Silver only give you 2C difference, and the really good ones are electrically conductive! Not something you want oozing out onto the motherboard! I'll stick to the standard silicone paste.

Generally it doesn't matter how good the transfer between CPU and heatsink is, if the heatsink isn't capable of getting rid of the heat. Generic "white" thermal paste is fine for most applications - certainly a lot safer than the liquid metal TIMs

 

[Macaw!]

To hell with it, I'm running at 85% clock speed (slowest the motherboard will go). If it gets worse I can turn down the processing % in BOINC (which is actually quite good - if I set 75% max, it'll throttle to keep OVERALL % at 75, even if another program is using a lot). If it gets bad enough to be too slow to use, or crashes a lot, I'll get another motherboard and CPU, then experiment with these in another machine and find what's broken. Something is definitely wrong as the heatsink is twice as good as the standard Intel one (it got 20C cooler when I swapped them over a month ago), and the Intel one used to be good enough. I've never known a CPU to break, but then it was overheating for who knows how long when the watercooler pump failed a year ago and I didn't know about it.

 

[doyll]

What Armageus said about non-known brand heatsink /cooler is spot on. £11 is ultra-cheap for a CPU cooler .. and off-brand to boot means it could be the copper pipes may not even be working heatpipes, but just copper tubes with closed ends that look like heatpipes. Try re-mounting stock cooler and see if it work better.

Sorry to be blunt, but from someone who buys a cooler like you did "big X case with lots of cooling" could mean almost anything. I doubt you have monitored the air temp going into cooler to know how much warmer it is than room ambient.

Get a name brand cooler like Aidos Armageus suggested and try it. At least then you will know for sure if it's cooler or CPU related.

Here is a guide to how case airflow works and how to setup case airflow so it cools well.
https://forums.overclockers.co.uk/t...-i-put-my-temp-sensor.18564223/#post-26159770

 

[Macaw!]

The cheap cooler removed 20C from the CPU temp over the stock cooler, so it's clearly pretty good. There was nothing wrong with the stock cooler, I checked the fan and heatsink gunk etc etc on it. the cheap cooler is warming up nicely to temps close to the CPU core, and hot air blows off it.

I couldn't be bothered finding the model of case, it's an Xcase but they don't make them anymore. It's a large full tower case with loads of room, and two front fans, two top fans, a rear fan, and a large side fan. Air comes in the front and the side and exits at the back and the top. Two exhaust fans are right next to the CPU cooler's outlet, and the big side fan blows directly onto it.

 

[Big.Wayne]

Macaw,

it's weird that the new cooler knocked 20°C off the CPU load temps but then gradually lost its cooling power so that the CPU temps reverted back to 100°C ?. What is the vCore reading under load?

 

[Randomface74]

Some thoughts:
I'm not much of an overclocker, but there's more than one voltage setting.
There's also a setting for fan profile on some mobos, check it's not in silent mode or something.
There's also a setting to reduce the voltage when the cpu isn't under full load, is that happening?
Is it under full load? what are you doing with it?
On my MOBO when I set XMP it changes way more than it should, check for that.

Post full spec.
Post photo(s) of BIOS.
Post photo of inside case.

 

[Macaw!]

Yes I don't understand losing the cooling ability at all. Mind you, some BOINC programs tax the CPU more, so it could be it just happened to be running something easier when I got a lower reading. I've updated the BIOS (on advice from another forum) and wished I hadn't. First Windows wouldn't boot, then I managed to stop it displaying any BIOS screen at all. I thought I'd killed my computer. Never ever flash a BIOS unless it's a new machine you're setting up with no data on it. I think I'll have a look around and select a top of the range cooler. Meanwhile I've set BOINC to use 2 of the 4 cores only. The reason the BIOS stopped displaying at all was when I reduced the base clock down to 80% and the multiplier from 34 to 30. I guess the CPU doesn't like that. So I've put it to stock speeds, and told BOINC to run 2 cores out of 4. If I tell it to use 75% load, the temperature fluctuates by 20C, which I'm sure is very bad (thermal stress).

 

[Macaw!]

Some thoughts:
I'm not much of an overclocker, but there's more than one voltage setting.
There's also a setting for fan profile on some mobos, check it's not in silent mode or something.
There's also a setting to reduce the voltage when the cpu isn't under full load, is that happening?
Is it under full load? what are you doing with it?
On my MOBO when I set XMP it changes way more than it should, check for that.

Post full spec.
Post photo(s) of BIOS.
Post photo of inside case.

All voltages are on auto or default. Fan is on normal. Turbo etc is on defaults. It's under full load running BOINC (the project SRBase calculating prime numbers taxes it the most). I have 1600MHz memory so I have extreme memory turned on. It doesn't seem to change anything non-memory related. All these settings have not changed for 3.5 years, so it can't be any of those that's causing the problem.

 

[Randomface74]

Auto voltage is generally not what you want - it tends to default too high.

 

[Macaw!]

In my experience lowering voltages (and clock speeds) tends to make things not boot or crash. I like to leave voltages alone! Anyway, the settings are as they were when it was working fine for 3.5 years. So I'm going to switch the blame to the cooler and buy a big one. It used to have a watercooler which was amazingly good, then I switched to the stock cooler when that broke, and was probably mistaken about it's abilities due to what happened to be running on the CPU at the time.