LinuxQuestions.org
Support LQ: Use code LQ3 and save $3 on Domain Registration
Go Back   LinuxQuestions.org > Forums > Linux Forums > Linux - Hardware
User Name
Password
Linux - Hardware This forum is for Hardware issues.
Having trouble installing a piece of hardware? Want to know if that peripheral is compatible with Linux?

Notices

Reply
 
LinkBack Search this Thread
Old 03-07-2013, 11:37 PM   #1
haertig
Senior Member
 
Registered: Nov 2004
Distribution: Debian, Ubuntu, LinuxMint, Slackware, SysrescueCD
Posts: 1,950

Rep: Reputation: 279Reputation: 279Reputation: 279
Question: CPU cooling


Related to this thread http://www.linuxquestions.org/questi...rd-4175452690/ that was originally about video cards, but morphed into a discussion on CPUs...

I am now looking for a CPU upgrade. I was undecided about upgrading the CPU until this evening. I found my system was crashing hard (total lights off, power down). It was very repeatable, I was using DVDShrink under Wine to shrink and re-encode a video. Lights-out total system crash every time after just a few moments.

The repeatability made me think it was software related. But the total lights out had me thinking it was hardware. I reinstalled Wine and DVDShrink with no improvement.

So I though, maybe thermal shutdown on the CPU? I figured the crash was just too repeatable to be memory, power supply, etc. Possible I suppose, but I think unlikely. I started looking at lm-sensors while running the video re-encode and saw my CPU core temps spike rapidly. Max safe temps for the CPU I have installed are around 65 Celcius. One of the cores spiked over 100 Celcius in a matter of seconds, with the second core not far behind. So I shut down DVDShrink before I had another crash. No telling how high the temps were getting when it was crashing. And I didn't want to experiment to find out! Thermal shutdown appears the most likely cause for my previous system crashes, at least in my mind. Any other comments?

The current CPU is cooled by the OEM CPU cooler that come in the retail box from AMD with the CPU. I used Arctic Silver when installing it (that was 4-5 years ago). This box has never been stressed before, serving only as a file server. So it is possible my CPU cooling has been massively inadequate since day one. The system was just never stressed enough to uncover that fact. That is my assumption at this point. I followed all the instructions for installing the CPU cooler and Arctic Silver very closely during the initial install. I *thought* I had done a good job, but maybe not.

Anyway (finally!), on to my question. Are the CPU coolers that come in retail packages with CPUs just trash to begin with? Or is it more likely I had some kind of bad installation? Are you supposed to routinely clean up the old thermal compound and reapply new and reinstall the cooler? If so, I have never done that in my life. Live and learn I guess.

My point being, when I run down to the store tomorrow to pick up a new (higher wattage) CPU, should I also pick up a third party CPU cooler? Or go with the one that will come in the box with the new CPU? The fact that I have probably cooked my current CPU and shortened its lifespan with the half a dozeon of so crashes I've worked through trying to debug things makes me think a new CPU is now more appropriate than ever. The old one may be roasted and half dead by now, although it still appears to be working fine.

Is this CPU cooler http://www.microcenter.com/product/3...rma_CPU_Cooler better than an OEM boxed one, or do I have to pay more money than this to get a decent third party one? Here's another, slightly higher priced potential choice: http://www.microcenter.com/product/3...sal_CPU_Cooler

Last edited by haertig; 03-07-2013 at 11:40 PM.
 
Old 03-08-2013, 05:38 AM   #2
yooy
Senior Member
 
Registered: Dec 2009
Posts: 1,071

Rep: Reputation: 125Reputation: 125
Quote:
Are the CPU coolers that come in retail packages with CPUs just trash to begin with?
no, you can still run 100% all the time (distributed computing project) on non-overclocked cpu with retail cooler.

Quote:
Are you supposed to routinely clean up the old thermal compound and reapply new and reinstall the cooler?
thermal compound should be cleaned and new applied before reinstalling cooler.


Zalmam is surely better and will result in lower temperatures.
 
Old 03-08-2013, 06:20 AM   #3
TobiSGD
Moderator
 
Registered: Dec 2009
Location: Hanover, Germany
Distribution: Slackware
Posts: 14,880
Blog Entries: 2

Rep: Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805
Usually the boxed coolers come with some sort of thermal paste already applied, so I wonder why you had to add thermal paste yourself.
But anyways, those coolers work fine, they are obviously not the best but should be able to keep the CPU within reasonable temperatures.
There may be several different causes for your CPU overheating:
- The cooler is not mounted correctly, so that it doesn't really attach to the CPUs surface.
- The layer of thermal paste may be to thick (it really should be as thin as possible) or dried out over time.
- Your cooler is blocked by dust/pet hairs or the fan isn't working. This is the most common cause for overheating systems, I would check this first.
 
Old 03-08-2013, 09:33 AM   #4
H_TeXMeX_H
Guru
 
Registered: Oct 2005
Location: $RANDOM
Distribution: slackware64
Posts: 12,928
Blog Entries: 2

Rep: Reputation: 1266Reputation: 1266Reputation: 1266Reputation: 1266Reputation: 1266Reputation: 1266Reputation: 1266Reputation: 1266Reputation: 1266
You don't need a new cooler unless the one that came with the box is not working.

Just remove the CPU cooler, clean all the paste off using either 99% isopropanol or whatever solution comes with the thermal paste, and apply a small amount of paste to one side and clamp the cooler back on properly. Make sure the cooler is attached very well, otherwise it won't work properly. Some of these coolers that come with the CPU are hard to attach, but it can be done.

http://www.techpowerup.com/articles/overclocking/134

You also don't need the best thermal paste. Anything containing zinc, aluminum, or silver oxides will work just fine. The silicone and copper based pastes are not good, so don't get those.

I change the paste every few years, because it do notice that it starts to wear out.

Last edited by H_TeXMeX_H; 03-08-2013 at 09:35 AM.
 
Old 03-08-2013, 02:06 PM   #5
haertig
Senior Member
 
Registered: Nov 2004
Distribution: Debian, Ubuntu, LinuxMint, Slackware, SysrescueCD
Posts: 1,950

Original Poster
Rep: Reputation: 279Reputation: 279Reputation: 279
Well, now you guys have gone and made me do it. Between this thread and a few others I have recently started. Went down to MicroCenter this morning. Wow, what a huge store - never been there before! The intend was to buy that Phenom II 6 core CPU. Did that. But "also picked up" a new WD Cavier Black terrabyte drive, a couple of Gb more memory, some new fans, etc.

This afternoon will be "upgrade day" for me.
 
Old 03-09-2013, 09:20 PM   #6
gradinaruvasile
Member
 
Registered: Apr 2010
Location: Cluj, Romania
Distribution: Debian Testing
Posts: 478

Rep: Reputation: 100Reputation: 100
A question: have you looked at the BIOS provided cpu temps?
A good test is to boot the system and go in the BIOS and look at the temperatures provided there. If the coolong isnt working well, you will see the temps raising rapidly, and in some cases leading to thermal shutdown (50-55 C is an acceptable temp here). This test excludes other causes that might have to do with issues like these.

The reason is that, unlike the OS, the BIOS does not use dynamic freqency scaling so the CPU is throttled to the max the whole time.
Also, as above mentioned already, make sure there is no dust in the fan/cooler. Best method of clearing the dust from the cooler and the case is by using compressed air (a compressor is ideal) to blow it out.

PS Make sure you do a BIOS version check/upgrade before you upgrade your CPU.

Last edited by gradinaruvasile; 03-09-2013 at 09:23 PM.
 
Old 03-10-2013, 12:25 AM   #7
nimnull22
Senior Member
 
Registered: Jul 2009
Distribution: OpenSuse 11.1, Fedora 14, Ubuntu 12.04/12.10, FreeBSD 9.0
Posts: 1,571

Rep: Reputation: 92
I would also suggest to use nice utility "lm-sensors". You can measure temperature with it. For example mine i5 gives:
Physical id 0: +32.0C (high = +80.0C, crit = +99.0C)
Core 0: +32.0C (high = +80.0C, crit = +99.0C)
Core 1: +32.0C (high = +80.0C, crit = +99.0C)
Core 2: +32.0C (high = +80.0C, crit = +99.0C)
Core 3: +32.0C (high = +80.0C, crit = +99.0C)
 
Old 03-10-2013, 09:20 AM   #8
onebuck
Moderator
 
Registered: Jan 2005
Location: Midwest USA, Central Illinois
Distribution: Slackware
Posts: 10,883
Blog Entries: 1

Rep: Reputation: 1307Reputation: 1307Reputation: 1307Reputation: 1307Reputation: 1307Reputation: 1307Reputation: 1307Reputation: 1307Reputation: 1307Reputation: 1307
Member Response

Hi,

Depending on the age and use of the machine the thermal compound can dry out. When this happens the paste hardens so the thermal transfer is not uniform between the CPU surface and the HSF.

I use denatured alcohol with a lint-less towel or quality shop towels to clean both surfaces. A quality compound (follow manufactures specification) should be used by applying enough to the center of the processor then use a plastic card(old credit card) or plastic razor blade to spread the compound even across the surface. Be sure that you do not apply excessive amount of compound as that will cause other issues.

When I place a HSF, the unit will be parallel but a few degrees off center then place slight amount of downward force on contact then twist the few degrees back to square HSF. That way the compound will be unified between both surfaces. Then secure the HSF.

One note, do not scratch either the processor or HSF while cleaning. The compound is used to provide a uniform heat transfer between the surfaces by filling any imperfections but too much compound can get onto other components thus causing issues..

One other thing, if your CPU happens to have a pad that was factory installed. Remove the old pad. I use the processor packaging or old socket assembly to hold the processor while removing a pad or cleaning a processor surface. That way no pins or motherboard will be damaged while cleaning or applying compound.
From http://www.tomshardware.com/forum/10...hermal-compund
Hope that helps!
 
Old 03-11-2013, 12:42 AM   #9
haertig
Senior Member
 
Registered: Nov 2004
Distribution: Debian, Ubuntu, LinuxMint, Slackware, SysrescueCD
Posts: 1,950

Original Poster
Rep: Reputation: 279Reputation: 279Reputation: 279
I found that the OEM cooling heatsink/fan/thermal_paste that came with the new CPU I installed was just fine. Below are the results of transcoding and reauthoring a 7.4Gb DVD to 4.5Gb. I was amazed that it only took 5 minutes. I'm quite happy with the results (CPU temp, CPU use, and time to transcode). CPU is a Phenom II X6 1045T 2.7GHz six core. Stayed at 29 degrees Celcius and 48% CPU use pretty much the entire transcoding time. At idle, this new CPU runs at 14 to 17 degrees Celcius. That's cold (57 to 62 degrees Fahrenheit) ... maybe I need to turn the heat up in the house!

http://img689.imageshack.us/img689/3...thor74gbdv.png

p.s. - The "psensor" app that did the above is just a simple graphing application for data collected with lm-sensors.

Last edited by haertig; 03-11-2013 at 12:47 AM.
 
Old 03-11-2013, 12:59 AM   #10
haertig
Senior Member
 
Registered: Nov 2004
Distribution: Debian, Ubuntu, LinuxMint, Slackware, SysrescueCD
Posts: 1,950

Original Poster
Rep: Reputation: 279Reputation: 279Reputation: 279
Interesting point to note: this same motherboard with it's previous CPU (Athlon X2 64 dual core 5200+) had lm-sensors displaying four temps - two temps for each core. With the new six core CPU, lm-sensors only lists a single temp. The only other significant difference is that when running with the older CPU, this box had 32-bit LinuxMint13Xfce installed. Now it has 64-bit, but still 13 and Xfce. Maybe lm-sensors changed between 32 and 64-bit? Or maybe the new CPU only reports the one temp? Not that it matters. I actually like only having to deal with one temp anyway.

I also think the thermal compound on the old Athlon CPU had hardened into stone over the years. I believe I used Arctic Silver, but I have built many systems and it could be I'm remembering wrong and this particular one used the OEM paste that came preinstalled on the heatsink. My system was experiencing thermal shutdowns with that CPU in place. I watched temps go over 100 degress Celcius with that old CPU before killing processes. No telling how high temps had gotten when it was shutting down. When I tried to remove the old heatsink when the CPU was still in the motherboard, no way. It was stuck solid to the chip. I finally had to pull the whole heatsink and CPU out as a unit (even when that little lever that I couldn't get to under the heatsink still had the CPU locked in place). Luckily the socket didn't appear to be damaged in the process and the new CPU is working fine. Someday I'll have to try to get the heatsink off the old CPU. But for now, they are solidly mated together in an anti-static bag in the spare parts box.

Last edited by haertig; 03-11-2013 at 01:07 AM.
 
Old 03-11-2013, 03:08 AM   #11
gradinaruvasile
Member
 
Registered: Apr 2010
Location: Cluj, Romania
Distribution: Debian Testing
Posts: 478

Rep: Reputation: 100Reputation: 100
I encountered cases such as this. You can separate the heatsink from the cpu with a cutter. It takes some patience though.

The multiple temp thingie - it really depends what monitoring chips/interfaces are accesible by lm-sensors. In some cases there are multiple interfaces that can be monitored and sometimes these are handled by different kernel modules. You can even get different reads because of different processing methods.
 
Old 03-12-2013, 08:06 AM   #12
TobiSGD
Moderator
 
Registered: Dec 2009
Location: Hanover, Germany
Distribution: Slackware
Posts: 14,880
Blog Entries: 2

Rep: Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805
Quote:
Originally Posted by haertig View Post
I found that the OEM cooling heatsink/fan/thermal_paste that came with the new CPU I installed was just fine. Below are the results of transcoding and reauthoring a 7.4Gb DVD to 4.5Gb. I was amazed that it only took 5 minutes. I'm quite happy with the results (CPU temp, CPU use, and time to transcode). CPU is a Phenom II X6 1045T 2.7GHz six core. Stayed at 29 degrees Celcius and 48% CPU use pretty much the entire transcoding time. At idle, this new CPU runs at 14 to 17 degrees Celcius. That's cold (57 to 62 degrees Fahrenheit) ... maybe I need to turn the heat up in the house!

http://img689.imageshack.us/img689/3...thor74gbdv.png

p.s. - The "psensor" app that did the above is just a simple graphing application for data collected with lm-sensors.
The reported temps for those AMDs aren't accurate, at least in the lower temperature range. You usually can add 10C go get the real temperatures. My Phenom II 1055T (2.8GHz) with Scythe Mugen II cooler reports temperatures below room temperature on idle, which is just impossible for something cooled with air.

Quote:
Interesting point to note: this same motherboard with it's previous CPU (Athlon X2 64 dual core 5200+) had lm-sensors displaying four temps - two temps for each core. With the new six core CPU, lm-sensors only lists a single temp. The only other significant difference is that when running with the older CPU, this box had 32-bit LinuxMint13Xfce installed. Now it has 64-bit, but still 13 and Xfce. Maybe lm-sensors changed between 32 and 64-bit? Or maybe the new CPU only reports the one temp? Not that it matters. I actually like only having to deal with one temp anyway.
Those CPUs report only one temperature.

Quote:
I also think the thermal compound on the old Athlon CPU had hardened into stone over the years. I believe I used Arctic Silver, but I have built many systems and it could be I'm remembering wrong and this particular one used the OEM paste that came preinstalled on the heatsink. My system was experiencing thermal shutdowns with that CPU in place. I watched temps go over 100 degress Celcius with that old CPU before killing processes. No telling how high temps had gotten when it was shutting down. When I tried to remove the old heatsink when the CPU was still in the motherboard, no way. It was stuck solid to the chip. I finally had to pull the whole heatsink and CPU out as a unit (even when that little lever that I couldn't get to under the heatsink still had the CPU locked in place). Luckily the socket didn't appear to be damaged in the process and the new CPU is working fine. Someday I'll have to try to get the heatsink off the old CPU. But for now, they are solidly mated together in an anti-static bag in the spare parts box.
Usually the trick is to twist the cooler until you can remove it.
 
Old 03-12-2013, 09:46 AM   #13
gradinaruvasile
Member
 
Registered: Apr 2010
Location: Cluj, Romania
Distribution: Debian Testing
Posts: 478

Rep: Reputation: 100Reputation: 100
I have an A8-5500 that reports 14 C on idle (room temp is 21-23 C). But if the CPU is loaded, the thermal reporting seems to starts working well, it goes up to ~50 and pretty much stays there (the fans are spinning much faster if goes higher, it seems it is a temp that it wants to keep).
I read somewhere that certain chips report accurate temp readings only if under load.

BTW my previous CPU was an Athlon II 250 and that reported the temperatures accurately (at least the same values as it did in BIOS).
 
Old 03-12-2013, 12:50 PM   #14
TobiSGD
Moderator
 
Registered: Dec 2009
Location: Hanover, Germany
Distribution: Slackware
Posts: 14,880
Blog Entries: 2

Rep: Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805Reputation: 3805
Quote:
Originally Posted by gradinaruvasile View Post
I have an A8-5500 that reports 14 C on idle (room temp is 21-23 C). But if the CPU is loaded, the thermal reporting seems to starts working well, it goes up to ~50 and pretty much stays there (the fans are spinning much faster if goes higher, it seems it is a temp that it wants to keep).
I read somewhere that certain chips report accurate temp readings only if under load.
The temperature sensor in those CPUs was designed to be able to prevent the chip from overheating, so accuracy is actually only needed in the higher ranges. That you can read out the temperatures is nothing more than a a little bit added value, so that you can check if your cooling system is working properly.
 
  


Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
cooling the CPU qrange Linux - Hardware 7 10-05-2011 03:40 PM
Computer cooling fans, possible water cooling ? H_TeXMeX_H General 31 06-19-2010 01:40 AM
Is anyone using a water cooling cpu color? abefroman Linux - Hardware 3 01-15-2006 06:36 AM
Compaq DeskproEN Cooling question. njbrain General 5 02-10-2005 04:39 PM


All times are GMT -5. The time now is 08:28 PM.

Main Menu
My LQ
Write for LQ
LinuxQuestions.org is looking for people interested in writing Editorials, Articles, Reviews, and more. If you'd like to contribute content, let us know.
Main Menu
Syndicate
RSS1  Latest Threads
RSS1  LQ News
Twitter: @linuxquestions
identi.ca: @linuxquestions
Facebook: linuxquestions Google+: linuxquestions
Open Source Consulting | Domain Registration