CPU running hot idle, cool when performing tasks?

[JamesAllen]

*UPDATE 4/17/11*
Somehow the BIOS set the CPU speed to 4.2 Ghz. I set it back to the standard 3.3 Ghz with 3.7 Ghz target for turbo. I ran the test again with the following results:

CPU-Z
Core Speed at 100% CPU Load- 3.4 GHz
Core Voltage at 100% CPU Load- 1.20 V

HWMonitor
Max Core Voltage- 1.23 V
Max Temperature- 76 C

RealTemp
Max Temperature- 79 C

This definitely helped bring down the temperature significantly (over 10 degrees), but it still has me a bit worried that I am getting a reading of almost 80 C at 100% CPU load.

-----

*UPDATE 4/17/11*
It seems that the flip flopping of temperature data was an issue with SpeedFan (possibly a bad install or download to begin with) because Real Temp gave me readings that made a lot more sense. I performed the Prime95's Blend Test with it and while I am still not enthused about the results, they are much better than before. I am now at 35-40*C with 0% CPU usage and I peaked at 90*C on one of the cores (the others were very close as well) at 100% CPU usage. This is still quite high and I plan to try reinstalling the cooler, trying a third temperature monitor to make sure Real Temp is accurate and it truly was SpeedFan, and I suppose I will get a new CPU cooler if all else fails. Any other suggestions are still appreciated.

-----

I finally finished building my PC and decided to run the tests suggested on the Tom'sHardware Step-by-Step Guide to Building a PC since all I could check before was the BIOS reading which showed my CPU running around 50*C. I chose to use SpeedFan to monitor my temperatures for the suggested test.

I installed Speedfan and gave my system a restart so I could start the test nice and fresh. I started up SpeedFan and began running Prime95’s Blend Test. I took a look at the digital readings and the graphical display and saw my CPU and AUX temps were both showing with a distressing flame next to their temperature readings with the AUX reading around 60*C and the CPU read from 80*C to 90*C from before the Blend test was started. Then right at the time of the Blend test, the CPU immediately dropped and stayed between 35*C and 45*C while the AUX dropped to around a constant 55*C. I ended the test, and then instantly, my CPU temperature shot up to 90*C and my AUX back up over 60*C.

This seems completely counter-intuitive to me. When my PC is idle, running absolutely no application the temperature is a dangerous 80*C plus. When I am running a stress test it drops by half to 40*C. What in the world could be happening? Why are the idle temperature in the OS so much higher than when I boot into the BIOS?

I am running the stock CPU cooler with the stock pre-applied thermal paste. Should I replace the thermal paste? The entire cooler? Is this even an issue with the cooler if it can reach 40*C?

Build
i5-2500K (no OC) with stock cpu cooler and pre-applied thermal paste
ASUS p6p87 B3LE
8GB 1066mhz
Corsair FS60 SSD
Samsung Spinpoint F3 1TB HDD
Antec 300 with stock 140mm and 120mm rear and top fans

 

[asantesoul]

Yes, you should replace the stock heatsink if you can... here's a fairly good one people like to use http://www.newegg.com/Product/Product.aspx?Item=N82E16835103065&cm_re=hyper_212%2b-_-35-103-065-_-Product
It's not expensive and it will do well

And yes, change the thermal paste if you can...some of these will be great
http://www.newegg.com/Product/Product.aspx?Item=N82E16835186038&cm_re=mx_5-_-35-186-038-_-Product
http://www.newegg.com/Product/Product.aspx?Item=N82E16835100007&cm_re=as5-_-35-100-007-_-Product

Probably go with the MX4....but, both are good..

And what bios are you running? Did you download the latest from the asus website? And as good as the sandy chips are, using the stock heatsink and running prime may not always produce favorable results..you need something a little better in a lot of cases

 

[JamesAllen]

If I change the heatsink then I would have to change the thermal paste. So, did you mean that if I don't want to replace the entire thing I should just try the thermal paste? How much of a difference can that really make?

I would really rather not install that heatsink if I can avoid it. It would require me to take apart about %90 of my PC to install, and that's not something I really want to do after I just put it all together.

I am running the BIOS that was already onboard. All the MoBo drivers are updated, but I left the BIOS as is because it was running fine (other than this new temperature issue, which I do not suspect is BIOS related) and the updates to the BIOS did not contain any fixes for anything that I was having issue with.

What do you mean about "not always [producing] favorable results?" I actually saw a reduction in temperature under the stress test which is what is confusing me so much. If my temperature was hot all of the time, I would know it was simply the paste and/or heatsink. It is this nonsensical fluctuations that leave me perplexed.

 

[asantesoul]

lol..to be quite honest it has me puzzled as well..have you tried using real temp..or hwmonitor to check your temps as well? perhaps use those and see if then results are the same..

And yes, paste can definitely make a difference..I understand your reasons for not wanting to make too many adjustments..but, that's what happens when you put together a pc..sometimes things don't go as smoothly as you planned, and changes must be made in order to resolve them... changing the heatsink isnt difficult..simply turn off the cpu for a bit..remove the old paste with special cleaner..or, 100 percent isopropyl alcohol and re-apply new paste..what I would suggest before doing anything is downloading the two programs I mentioned and tell me if you still have high idling temps

 

[JamesAllen]

It seems that it was an issue with SpeedFan (possibly a bad install or download to begin with) because Real Temp gave me readings that made a lot more sense. I performed the Prime95's Blend Test with it and while I am still not enthused about the results, they are much better than before. I am now at 35-40*C with 0% CPU usage and I peaked at 90*C on one of the cores (the others were very close as well) at 100% CPU usage. I still feel like I would benefit from, and feel much safer with, a new heatsink and thermal paste, but I do not feel the rush for it as I felt before, nor the confusion that I had when these results were flipped. I will try reinstalling SpeedFan and/or installing hwmonitor tomorrow to double check the results.

Thank you!

 

[geekapproved]

If your not overclocking, you do NOT need to replace your cooler.

If your hitting 90c, the cooler was not installed properly.

Try CPUID HWmonitor and recheck temps, it can't possibly run cooler under load than idle, it's not physically possible unless your fan for some reason only spins under load.

 

[JamesAllen]

If your not overclocking, you do NOT need to replace your cooler.

If your hitting 90c, the cooler was not installed properly.

Try CPUID HWmonitor and recheck temps, it can't possibly run cooler under load than idle, it's not physically possible unless your fan for some reason only spins under load.

I did recheck with Real Temp and it read it "correctly" (I mentioned this in my response above, but maybe I should edit the original post). That being said, it was still putting out around 40*C at 0% CPU load and peaked at 90*C at 100% load. I will try reinstalling the cooler today.

 

[asantesoul]

Excellent...I figured the program you were using was proving inaccurate results..hw monitor and real temp are far better in providing accurate information...and yes, if you can re-install the cooler with different thermal paste that may prove to be quite beneficial...

Put a line on the chip going diagonally (not too much) and then attach the cooler...or, put a very small pea sized amount on the center of the chip and use a credit card, or business card to spread it evenly..then attach the cooler.. Also, check the bios and see how much voltage your giving the chip..it could be that if it's on auto (usually this is the default) the chip is getting more voltage than it needs, especially when under load..

Experiment a bit and then let me know what happens

 

[JamesAllen]

How do I know what voltage to set in the BIOS? Unfortunately, the ASUS P6P87 requires going into OC mode to tweak CPU voltages, and that is a kind of scary territory for me, but if it is necessary I can do it.

 

[JamesAllen]

I reseated my CPU cooler. I had not turned the pins the correct way, and so it was not completely locked down. Unfortunately, even after correctly setting it, the results of Prime95's Blend test were the similar with 35-45*C at 0% usage and 90*C at 100% usage.

 

[asantesoul]

if that's the case..then perhaps you may need to use a different cooler..or different thermal compound altogether.. As for the voltages, download cpu-z (1.57) and monitor the voltage while the computer is at 100% load.. if it's at stock, and it turbos to 3700 mhz (3.7Ghz) and has more than 1.3 v..then that's a problem..simply because that amount of voltage can give you a nice 4.5-4.7 overclock..

As for the voltages in the bios..just set it to manual..and you can adjust it yourself...but get cpu-z first to check the voltage at load

 

[JamesAllen]

I ran Prime95's Blend Test up to 100% load using HWMonitor, RealTemp, and CPU-Z with the following results:

HWMonitor
Min Temp- 36*C
Max Temp- 88*C
Max CPU Vcore (at Max Core Speed seen in CPU-Z)- 1.34V

RealTemp
Min Temp- 28*C
Max Temp- 92*C

CPU-Z
Max Core Speed- ~4200Mhz
Core Voltage (at Max Core Speed seen in CPU-Z)- 1.312-1.32

Does this mean that my PC is automatically overclocking my CPU to 4.2Ghz and that is what is bring up my temperature?

 

[asantesoul]

yeah...your cpu is going straight to 4.2 Ghz..did you adjust the multiplier in the bios? lower the multiplier to 3.7 gh.z if it isn't already... and that much voltage for a 4.2 ghz oc is a bit high..especially considering your using the stock cooler...

set the turbo boost/multiplier to 3.7 ghz..and if you can set the voltage to 1.29-1.3...maybe 1.28...and test again with prime for stability and temperature...the most important thing right now is to bring down the overclock to 3.7...and check the temps at load..don't worry so much about idle for now

 

[JamesAllen]

I have no idea how it got set to the 4.2 Ghz multiplier. I set it back to the standard 3.3 Ghz with 3.7 Ghz target for turbo. I ran the test again with the following results:

CPU-Z
Core Speed at 100% CPU Load- 3.4 GHz
Core Voltage at 100% CPU Load- 1.20 V

HWMonitor
Max Core Voltage- 1.23 V
Max Temperature- 76 C

RealTemp
Max Temperature- 79 C

This definitely helped bring down the temperature significantly (over 10 degrees), but it still has me a bit worried that I am getting a reading of almost 80 C at 100% CPU load. I will likely add an aftermarket CPU cooler in the near future, but I am no longer worried about the CPU being the issue.

I know you said the core speed would be at 3.7 GHz, while I only witnessed 3.4 GHz, but I believe that the speed is correct based on other forum replies which indicate that with all 4 cores working in turbo mode, the speed will be 3.4 GHz rather than the regular 3.3 GHz or the Turbo 3.7 GHz.

 

[JackNaylorPE]

I reseated my CPU cooler. I had not turned the pins the correct way, and so it was not completely locked down. Unfortunately, even after correctly setting it, the results of Prime95's Blend test were the similar with 35-45*C at 0% usage and 90*C at 100% usage.

That's a problem. The thermal pad was not set properly and my guess is that you therefore do not have an even spread of TIM. If you want to try the stock cooler again, you'll have to do one of the following:

1. Remove the cooler and CPU, clean the CPU and cooler w/ isoprophyl alcohol, let dry, apply the new TIM and reinstall.
2. Grab a replacement stock cooler with intact thermal pad .... since hardly anyone uses, a local shop or computer user might be willing to give ya one. Clean the CPU as described above and reinstall

Your numbers still just sound wrong .... note that ya shouldn't use multiple monitoring programs at the same time as they can screw each other up. Many sites have reported running the SB processors to 4.4 and above using the stock cooler.....Guru3D for example says they did 4.6 at < 70C but as it was published pre -release of SB, as has been pointed out by Leaps-from-Shadows .... how do we know it was the stock cooler or was it the special Intel tower cooler ? Looking at the THG review published 23 days after SB's release ( http://www.tomshardware.com/reviews/sandy-bridge-overclocking-efficiency,2850-14.html ), I gotta think the review sites are getting different "stock coolers"

http://www.guru3d.com/article/asus-sabertooth-p67-tuf-review/10

Though have no way of knowing what they used, it's irrelevant to my point as what I wanna get at here is the bold portion:

Make sure your processor is properly cooled (we used the stock Intel cooler and forced the fan to 70% RPM)

Try going into the BIOS and upping the fan speed control.

Note: I gotta wonder if perhaps many of us are just being overly cautious wanting to keep temps in lo 70's on our rigs. We have become accustomed to the temps of our GPU's increasing to 100C over the years but we seem to be stubbornly set at 72 for CPU's. I have seen nothing published one way or the other. Regardless, the temps you are seeing are way outta whack.

Bit late to ask this..... but have you confirmed that fan is spinning, fan is properly plugged into CPU fan header on MoBo and that temperature and fan speed appear at least to be accurate in BIOS ? Does fan speed increase as CPU temp increases ?

And yes, update the BIOS. It's really no big deal....download the new BIOS, unzip it to a location on your HD where you can easily find it. Load Asus Update and tell it to load the BIOS from a file and point it to the file you just unzipped .... no biggie.

 

[JamesAllen]

I didn't think about the improper seating causing the pad to spread unevenly (I was only considering the area of contact). I ordered Arctic 5 Thermal Compound and will clean off my current CPU cooler and apply the it. If that does not work, I will buy an aftermarket CPU cooler and I will already have the thermal compound I would have needed to purchase anyway.

I will also try upping the CPU fan's RPM (I know it is spinning BTW) and see if that makes a difference.

 

[asantesoul]

yeah..trial and error my friend..that's the best thing to do..when you get the new thermal paste..try and put a small amount on the cpu (smaller than a pea size) and spread it with a card..and see if that helps with the temps..but definitely if you can at some point..get another cpu cooler..but, for the time being just try and enjoy the computer...it does really hot..but, that's with prime 95..a program that does what other programs and games wont do to your cpu..so, in the mean time..i say monitor temps..and enjoy the computer until you get your AS5 and/or new cooler

 

[JamesAllen]

So it would seem. I tried upping the fan speed. It was already set to 100% if temperature reached 70 C, but I set the minimum to 70% for anything over 20 C where it was previously set to only 20%. This had no effect, even at lower temperatures. Makes me think its it an issue with the paste or heat-sink since an increase from 20% to 70% speed on the fan should have had at least some effect on the on the temperatures.

At least now I have solved the CPU speed issues, CPU voltage issues, and eliminated the fan speed as a possibility. Really only leaves the heat-sink or thermal paste, which like I said would be my next logical guess anyway. Thanks for all of the help, and I will update again once I have tried the thermal paste on the existing stock cooler after it arrives in a few days.

 

[asantesoul]

Excellent..in the mean time..relax..and enjoy the computer!

 

[JamesAllen]

Removed old thermal paste and applied Arctic 5 yesterday. Unfortunately, the temperatures are still the same. I have heard Arctic 5 can have a 200 hour curing time, so it is possible that it will drop a few degrees in the coming week, but it would appear that I may just have a bad cooler.

I just want to make sure this is not an issue with a bad CPU. I feel that these temperatures are reasonable enough that the poor cooler should explain them, but if the CPU could be the cause, I would like to find out as soon as possible. How can I know if the CPU is defective and is in fact the culprit?

 

[asantesoul]

You could try a modest overclock to see if the temps get EVEN higher..are you still using the stock cooler? If so, I think it's time to get a new cooler..Also, not sure if i mentioned this..but, try plugging the cpu fan into a different part of the mobo..other than the cpu fan header..let it run on another fan header and check the temps

 

[JamesAllen]

Best answer selected by jamesallen.

 

[JamesAllen]

I know this topic is old, but I promised I would update it once I got the new CPU fan/cooler. I finally purchased a ZALMAN CNPS9500 AT 2 Ball CPU Cooling Fan/Heatsink. I installed it today, with Arctic Silver 5 and the results are promising.

The cores are idling in the 30's which is reasonable consider the temperature outside is at the high end of the range, and the temperature inside is at the low end. These temps are slightly lower than I was seeing with my stick cooler and fan in 18C weather.

At load, the cores range from 66C to 69C which is about 30C cooler than when I found my initial heat issues, and about 10C cooler than when I reinstalled my stock heat sink with Arctic Silver 5 (fixing the initially poor install and terrible thermal pad) and fixed my voltage/overclock issues.

Thanks to everyone for all of your help! I can now be 100% confident that I am not hurting my computer when I am forcing it to render my dozens of layers of video.