My PC is top-notched all the time. Pretty basic configuration.
Motherboard Intel MB DP43TF
Processor Intel CPU Q8200
Memory 8 GB RAM
Power Enermax Modu80 650W (If I remember correctly)
Graphics Nvidia GTS 450
Lately a weird thing has started to happen.
On idle, regardless of temperature, the CPU fan suddenly raves up from 1K to 2.5-3K RPM, usually for a few seconds, and then decreasing.
It happens every few minutes and without a predictive pattern.
I might be able to connect the phenomenon to a power-cut that happened similarly to the appearance of this.
Things I've tried:
- Checking temps are OK with a utility (temps are OK)
- Resetting the CMOS by removing and reinstalling the cell (no change)
- Cleaning the sinks (no change)
- Load PC into BIOS to see if it got anything to do with OS (happens in BIOS as well)
I'm really stuck here so if any1 got something... I'll be thankful.
Software wise you may have a program running (in the background) that is causing a usage spike. Maybe try running a Linux OS from CD/DVD ans see if the same happens.
Hardware wise you could try reapplying some fresh thermal compound. Also consider the ambient temperature - a hotter (seasonal?) room may tip the temperature sensors over the threshold for higher fan RPMs.
I've mentioned that this phenomenon occurs also when entering the BIOS page and staying there.
The later suggestion is interesting though. The threshold thing. I imagine that raising the threshold is a good test to see if that solves this. That way I would be able to tell if the behavior was normal.
Any tip for setting the threshold or should I simply search and read about it?
If you are still using the stock cooler or something else with push-pins, it might be starting to loosen. Poor contact between the HSF and IHS would cause fan RPM to spike whenever the CPU has an activity spike.
A simple way to verify this would be to use monitoring software to see how quickly the CPU warms up under light/medium load and how quickly the fan ramps up with it.
In my case, I suspect the plastic pins and fan frame deformed just enough over ~3.5 years to no longer be capable of providing sufficient contact force. Replacing the paste only helped for a few days, the only thing that solved it for good was an aftermarket HSF.
Since I did not feel like messing with push-pins and have to worry about it loosening again, I simply went with the Hyper 212+.
Lastly, I sometimes get "voltage out of range" exceptions from the Intel Desktop Utility monitor. It's usually about the +3.3V and the Memory Controller Hub Vcc. Is that normal from time to time?
It would be a "little" odd for normal operating voltages to be "out of range" so this would be one of those eyebrow-raising moments where I would use a multimeter and see if I can test those voltages to determine whether it is a monitoring error (software or hardware) or real reading.
You might want to check your BIOS to see if the RAM is set to default voltages just in case. However, since you have an Intel board, I'm guessing you have no such option.
Basic electronics troubleshooting stuff: when you doubt the diagnostic software or remote sensor and cannot reliably be verified/calculated from alternate means, take a hands-on measurement if possible.
Is there any connection between the two?
Probably not. At 4+ years old, my best guess is that the stock HSF retention mechanism loosened up and needs replacement.
Some people (including myself) are wondering if this could be an engineered failure to make CPUs run into thermal throttling, annoy the heck out of people and make them upgrade more often.
If the memory controller voltage jumped to 32V, it would definitely be fried (it would probably take less than 5V to cause instantaneous failure) so this one is definitely a glitch.
The IOH temperature (assuming it is accurate and really stands for what it says it does) should not be that high. If your case does not have a front-bottom intake fan (the IOH is under the smaller heatsink near the bottom-right corner), you might want to try clearing the space around the IOH and adding a front fan first, maybe your IOH area is simply a little short on airflow. If you still get high IOH temperature (mine rarely reads above 50C), it might need a re-paste.
The CPU fan speed is a function of CPU temperature. Fan speed spiking should not happen if the HSF is making proper contact with the CPU since the HSF will prevent the CPU from experiencing nearly instantaneous temperature spikes.
Try running something like FurMark in a window and keep an eye on your monitoring tool's CPU temperature. This should peg your CPU fan to maximum speed within seconds until the benchmark completes or you cancel the burn-in test. If your CPU temperature stays below 70C, your CPU's HSF is fine and the annoying fan spikes are caused by obscure other factor. If your CPU temperature jumps over 85C then your CPU HSF is likely not making good enough contact with the CPU as I mentioned a few messages ago.
One last question about the CPU: what were the core temperatures during Prime95?
Since you have a system monitoring chip that occasionally reports bogus memory controller voltage, my next best guess is that it might have other bogus behavior which may include the fan spikes. In that case, you might have to live with it. Other possibilities include using a precision screwdriver to punch out the PWM pin out of the fan connector (or cutting the PWM wire), using an aftermarket fan controller, plugging the CPU fan on a normal 3-pin fan header or wiring the CPU fan to a molex connector either for 7V or 12V.