CPU temp fluctuating wildly

furpile

Expert Member
Joined
Jul 14, 2014
Messages
4,419
Reaction score
1,059
Location
Richards Bay
Hi

I have an i5 2500 (no overclock) that I have been using for the last 4 or 5 years. Using the stock intel cooler, my CPU temps were always low, like in the mid 30C range on idle and never more than mid 50's while gaming. I have had to replace my mobo about 3 times, and used generic thermal paste from Matrix. The last time was probably about 2 years ago. Every time my temps were were decent. I also got a GTX960 earlier this year and checked the temps quite a bit back then, everything normal.

About 2 weeks ago I randomly opened the monitoring program again (CPUID HWMonitor) and I see my CPU temp is fluctuating badly. Like every second or half second that it updates every core and the package temp has a different value. This is on idle in Win7, nothing running. The temperatures vary by up to 5 degrees between readings (less than a second apart), and in the last 5 minutes idling in Windows the range is from 33C to 53C. While gaming now the max it reaches is 70C sometimes, even though the CPU does not work a lot.

So I first though the crappy thermal paste has dried out. So I thought while I am at it, put a better cooler on. Ordered the Coolermaster Hyper TX3 Evo because it is cheap (R350 about) and the reviews are decent and you don't need to remove the mobo to install (and I don't overclock so I don't need major cooling). I also ordered Coolermaster Value V1 thermal paste, which seemed the best from what Wootware had available and I thought it should be better than the bit they give you with the cooler. I installed the new cooler tonight, cleaned it all with alcohol, applied a very thin layer on the CPU and the cooler base, and then a thin line in the centre (this is how it is described on the Artic Silver website). Turn on the PC and go into Bios, and immediately the CPU temp is on 39C. After a minute or so it goes to 40, then later to 41. This is already hot, but I boot into Windows and open the monitor again. And it is doing exactly the same thing (described above). As I type it is varying between 38C and 43C the whole time. I also felt the heatpipe directly where it comes off the CPU and it definitely did not feel warm, not what 40C should feel like.

So my question is, can it be the temp sensor that is faulty (all 4 cores and package)? Can it be that the CPU is not really that hot (reading wrong). What else can I check? The old thermal paste on the stock cooler seems to have spread well the previous time, and it was still a little sticky, so I am not sure that was the problem.

Other info:
Mobo temp is stable at 31C and GPU temp is also stable at 29C, it is only the CPU giving the problem.
 
So my question is, can it be the temp sensor that is faulty (all 4 cores and package)? Can it be that the CPU is not really that hot (reading wrong). What else can I check?
There are separate sensors on the CPU. Do they jump randomly (against each other) or have tendency to follow in the same direction? If later, then CPU reference voltage regulator is faulty or one of power supply rail is unstable. It could be either on-board noisy core voltage regulator, on-board 3.3V/1.5V regulator or 5V power pack rail. You need decent 100MHz osciloscope for this evaluation or just spare power supply, mobo or CPU for swapping.
 
Last edited:
There are separate sensors on the CPU. Do they jump randomly (against each other) or have tendency to follow in the same direction? If later, then CPU reference voltage regulator is faulty or one of power supply rail is unstable. It could be either on-board noisy core voltage regulator, on-board 3.3V/1.5V regulator or 5V power pack rail. You need decent 100MHz osciloscope for this evaluation or just spare power supply, mobo or CPU for swapping.

+ 1 Would do the PSU test for sure. Check the Watt rating on your PSU. Remember it takes time for the thermal paste to "cure".
 
Thanks for all the replies so far.

Try coretemp to see if it does the same.

Installed Coretemp now and it shows the same, just a second or so out of sync with CPUID.

There are separate sensors on the CPU. Do they jump randomly (against each other) or have tendency to follow in the same direction? If later, then CPU reference voltage regulator is faulty or one of power supply rail is unstable. It could be either on-board noisy core voltage regulator, on-board 3.3V/1.5V regulator or 5V power pack rail. You need decent 100MHz osciloscope for this evaluation or just spare power supply, mobo or CPU for swapping.

It is mostly random, when they jump just a few degrees. But when they go up a lot they all go up (like to 50C). The CPU voltage is also fluctuating according to both programs, between 0.961V and 1.256V. Not sure if this has anything to do with any power saving setting or anything? This happens while idle, does this indicate a faulty PSU?

These are the Mobo Voltages:
Capture.PNG

The 12V seems very low? These min/max values are over about 2 minutes while idling, you can also see the CPU voltage varying.

I don't have any spare parts to test, but I can make a plan if I need to. Just need a direction to test.

+ 1 Would do the PSU test for sure. Check the Watt rating on your PSU. Remember it takes time for the thermal paste to "cure".

This is a 550W psu, so it should be OK, although it is a cheap Matrix PSU. I will have to see if I can get another unit to test maybe.

Thanks again, will check again tomorrow.
 
GTX 960 on a 550W cheapie. disaster waiting to happen. 550W is nowhere near 550W on those.
 
GTX 960 on a 550W cheapie. disaster waiting to happen. 550W is nowhere near 550W on those.

Combined TDP of 960 + i5-2500 is 220W at full load which is not much. Not discounting a crap psu though, best to test with a spare or loan unit.
 
It is mostly random, when they jump just a few degrees. But when they go up a lot they all go up (like to 50C). The CPU voltage is also fluctuating according to both programs, between 0.961V and 1.256V. Not sure if this has anything to do with any power saving setting or anything? This happens while idle, does this indicate a faulty PSU?

These are the Mobo Voltages:
View attachment 356952

The 12V seems very low? These min/max values are over about 2 minutes while idling, you can also see the CPU voltage varying.
It doesn't looks like CPU. Core voltage is changing, as CPU is adjusting voltage, it could be normal for some CPU or a false reference voltage (as mentioned before) is forcing CPU to enter power saving mode.

However 5V power supply rail is terrible wrong. Minimum 5.718V? It is completely out of spec, you are in self-destruction zone. Stop using this power supply! It is your [-]third[/-] motherboard in a row and I strongly suspect your power supply is a reason for that. Another sign is your 12V rail. Also out of spec.

[EDIT] Corection: fourth motherboard.
 
Last edited:
It doesn't looks like CPU. Core voltage is changing, as CPU is adjusting voltage, it could be normal for some CPU or a false reference voltage (as mentioned before) is forcing CPU to enter power saving mode.

However 5V power supply rail is terrible wrong. Minimum 5.718V? It is completely out of spec, you are in self-destruction zone. Stop using this power supply! It is your [-]third[/-] motherboard in a row and I strongly suspect your power supply is a reason for that. Another sign is your 12V rail. Also out of spec.

[EDIT] Corection: fourth motherboard.

Thanks. I have not had the PSU since the beginning. The first 2 mobo's that stopped working were Gigabyte Z68P-DS3 models, each one after almost exactly a year. This was a known problem with that chipset and Rectron replaced it twice under warranty. The last time I did not even use the replacement because I got tired of reinstalling my OS every year. Bought an ASUS mobo and it has been happy since then. Not discounting the PSU as a problem, just saying that I am not sure it was related to the previous mobo issues. I will have a look at getting a PSU to borrow, or just get a new one and hopefully that is the problem. What size will be ok to run this? Will go for a bronze at least this time. PC has one SSD and 2 mechanical drives a this stage, the GTX960 (also the reason I went for this model as it doesn't use a lot of power) and the i5.

Thanks for all the other inputs as well.
 
Last edited:
OK, so I got the Seasonic model listed above. Just installed it into my PC. And it's the same.

Cpu temp is immediately at 40C booting directly into BIOS only. In windows it is still jumping up and down (between 36C and 56C now).

5V is still sitting at 5,8V. 12V is still just above 8V.

So obviously it is not the PSU that is causing the problems? Is the mobo damaged maybe?

With the new cooler I can feel the heatpipes directly off the CPU and it doesn't even feel warm, surely they should at least be warm if it is at 40C?

So what are the options now? I am not happy leaving it like this but I have now replaced the first two obvious components. Replacing the mobo will mean a new install of Windows which I do not want to do now.

I measured the 5V on a fan plug as well and it is at 5.08V, so there is a good chance the mobo is reporting wrong?
 
My 5V rail is also exactly 5.718


HWmonitor bug?

Seems like it. I am googling now and someone recommended HWInfo on tomshardware. Running this it shows 5V at 5.045V and 12V at 12.232V. So I probably didn't need a new PSU even though this one is a lot better than what I had.

Problem is even HWInfo shows the CPU temp varying the whole time. All 4 cores show an 8 or 9 degree variation.
 
It could just be a faulty sensor. If the cooler feels cool then just ignore it.
 
It could just be a faulty sensor. If the cooler feels cool then just ignore it.

It is just annoying that the temp is all over the place, I can hear the fan speed up every now and then as the temp goes over 50. I am afraid something might break, but the temp has never gone above 70C about and the PC has never shut down because of over temp. So it should be ok, but I still don't know.
 
Top
Sign up to the MyBroadband newsletter
X