• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Info regarding high temperatures, causes/solutions, and reading from the CPU diode

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
galador said:
He moved the cpu and HS to another Mb and the temp now says 30C idle 46 load. Problem solved :D .


That works. I bet either he started reading via the CPU diode, or maybe accidently adjusted the voltage for the CPU. Volts can have a big effect on the heat a CPU puts out. Here are some examples:

Ok...nevermind. I forgot the equation. When I go home I will look it up.
 
Streifenkarl said:
Nice... but how do i read the CPU diode with my motherboard (see sig)?


Ok, first go in your BIOS and tell me if you have 3 or more temp readings in there. If you have two for the CPU, then you can read via the CPU core.

There should be three:
1.) Case
2.) CPU something or other (socket)
3.) CPU something else or other (core)

or in any other order

If you only see:

1.) Case
2.) CPU something or other


...then there are extremely high chances you cannot read via the CPU core diode.

Once you figure that out, we can go about configuring MBM to read that. :)
 
@md0Cer said:
*WORK IN PROGRESS*


CAUSES FOR HIGH CPU TEMPERATURES (and solutions!)

Heatsink Seated Improperly
It could be that your heatsink is seated improperly. If so, your idle temperature on your CPU diode will be high, and your load temperature will be even higher and by a good 10+ degrees and rise instantly (roughly in the blink of an eye type of "instantly" ). How to solve this would be:

  • Take heatsink off
  • Clean off CPU and bottom of heatsink with 91 percent Isopropyl Alchohol
  • Apply some silver sompound to the heatsink, rub it in and then off,this is rumored to fill the microscopic cracks and stuff in the heatsink.
  • Apply a thin layer to the CPU core or heat spreader, make it even but thin as possible..not too thin..but thin.
  • Make sure you don't get any dirt or anything in!
  • Carefully put the heatsink back on

Heatsink and/or CPU heatspreader not flat!
If your heatsink and/or cpu heatspreader is not flat, you will see high temperatures and a large gap between load and idle temperatures on your CPU core. How do I fix this?

  • Take heatsink off, take CPU out if lapping the heat spreader.
  • Find a perfectly flat block, or surface.
  • Get some sandpaper, from 400-1000 grit (you can find these high grit count sandpapers at autoparts stores)
  • Place the sand paper on the flat surface, hold the heatsink or CPU surface on the sand paper, and rub it. Make sure it is flat. Your goal is to get them perfectly flat.
  • Start sanding with the lower grit, like the 400. Remember, your goal is to get it perfectly flat, so make sure you are not tilting anything, and make sure the sand paper is perfectly flat against the surface, or against your perfectly flat sanding block (none that I know of unless you take something that is glass and really flat or like a thing of polished granite. You can sand in any direction, it does not matter as long as it is flat, if it is diagonal, horizontal, vertical, or circular. If I do something that is not circular I do it in the opposite direction, like vertical after going horizontal so the small scratches are not all the same direction, and thus it could be smoother and more flat.
  • Look so see after a few seconds if the scratch marks from the sandpaper cover all of the surface, just the middle, or just the outsides. When it does not cover a certain area...that means it is not flat. Keep going until it is flat, then move up gradually to the 1000 grit.
  • Clean off CPU and bottom of heatsink with 91 percent Isopropyl Alchohol
  • Apply some silver sompound to the heatsink, rub it in and then off,this is rumored to fill the microscopic cracks and stuff in the heatsink.
  • Apply a thin layer to the CPU core or heat spreader, make it even but thin as possible..not too thin..but thin.
  • Make sure you don't get any dirt or anything in!
  • Carefully put the heatsink back on

Look at my post above for some lapping links :)

Thermal Paste/Grease too thick!
One of the most common causes for high temperatures is putting on a layer of thermal compound that is too thick. It depends on how flat your heatsink is. If you have a horribly unflat heatsink, maybe a thicker layer would be required for it to make contact. Usually a thin layer is better. You still want to be able to see thermal compound on there, but make it as thin as possible. Sometimes a good way to tell is if there are numbers or letter on the core or heatspreader. If you are just about to the point of being able to see the numbers, but just a wee bit thicker. That is usually ideal, and is quite thin.

  • Clean off heatsink and CPU with 90% Isopropyl Alchohol, or that Arctic Silver cleaning thing. If you have to use 70 something % Isopropyl Alchohol (rubbing), it should work ok. Use a lint free cloth or paper towel that does not leave anything behind.
  • Rub some of the compound into the heatsink before starting, just a little bit, enough to fill the microcopic cracks. Couls just be a myth, but I do it.
  • Apply just a little bit to the CPU core or heatspreader, just a little, you can always add more.
  • Put on a glove or put your finger in a bag and rub it so it covers the whole core or heatspeader. If too much is on, just wipe off your bag or glove covering your finger with a paper towel and spread it around. Remember, have a thin and consistent layer, but not too thin.
  • Attach the heatsink! Don't let it sit around exposed for a long time, you don't want dirt on it, or cat hair, etc.


More to come :)

I updated :D
 
Last edited:
@md0Cer said:
That works. I bet either he started reading via the CPU diode, or maybe accidently adjusted the voltage for the CPU. Volts can have a big effect on the heat a CPU puts out. Here are some examples:

Ok...nevermind. I forgot the equation. When I go home I will look it up.

Ok, just as an example, my 1.47GHz Thoroughbred B at stock settings throws off 49.3 thermal watts. If I ran it at 1.75Vcore, (an easy mistake in the BIOS, if set on automatic, if it recognized it as the wrong CPU that could be an easy mistake too) that would be 67.2 thermal watts. That could yeild quite a temperature difference.

Wanna see how much of a difference that would make with it overclocked?

At 2.53GHz at 1.75 volts it would make 116.4 thermal watts. At 2.53GHz at 2.0 volts it would make 152 thermal watts. Quite a difference with just a voltage increase eh?
 
Good info. Thank you.
I agree that some people worry a bit too much.
My comp info is in my sig. I have been running this way for over a year or more.
 
craigiz1 said:
Good info. Thank you.
I agree that some people worry a bit too much.
My comp info is in my sig. I have been running this way for over a year or more.


Thanks. At 43°C that is nowhere near a damaging temperature.

I really wish I could find that other guy's post about this. He explained it really well. I think it was either NookieN or Dissolved, or someone, I am not sure who. I will continue my search for that other thread. I just updated that post in this thread and added Dirt/Dust. :)
 
@md0Cer said:
Ok, first go in your BIOS and tell me if you have 3 or more temp readings in there. If you have two for the CPU, then you can read via the CPU core.

There should be three:
1.) Case
2.) CPU something or other (socket)
3.) CPU something else or other (core)

or in any other order

If you only see:

1.) Case
2.) CPU something or other


...then there are extremely high chances you cannot read via the CPU core diode.

Once you figure that out, we can go about configuring MBM to read that. :)

All i have is the CPU temperature, the system temperature and the Northbridge Temp.
 
Hmm... That is wierd. My "system temperature" is my northbridge temp.

Ok. I will have to do some reasearching on your motherboard. If I don't get back to you in 24 hours, start pestering me via PM.
 
Captain Newbie said:
Nobody reads stickies anymore ;) Remember: Posting and You.


Stickies are invisible. Either way, I don't care if it is stuck or not. I just make the threads so I can link people to them and not have to re-type the info each and every time. If it gets stuck or not is out of my control. ;)

Posting and you? what?
 
Ok, I apologize for not finishing up the MBM configuration part. You can expect an update soon. There is also another thread I am working on that displays the manufacture's maximum recomended CPU casing temperatures for each CPU they make as well as thermal wattage specifications for CPU's. A vcore upper limit thread can be expected sometime before 2007. ;) j/k I will jump right into that as soon as I get these threads done!
 
I just want to point out that even the OC is not limited by the temperature at all :eek: It maybe sounds insane, but I got two proofs. Neither one of mine CPUs are need a one single Mhz step-down from their max. OC when temp hit 86C (most I tried, then it get too scary even for me :D )
Second, undervolters found, that hot CPU can live and fold hapilly and stable with LESS VOLTAGE, that it then need to start, when it is cold :-/
(translated - you can use AXP-M at 0.9V and 1200Mhz, but it need 1.0 to 1.1V for core to start, when it is cold. When you heat-up it, it live hapilly with 0.9V again)

I think all this could be explained when we look at physic and see, that semiconductors does NOT increase their resistance to elektricity, but they does DECREASE it.

Pretty funny, right?

Translated it means that temperature should help OC and not be problem. Sounds like the exactly oposite that some of the best OC guys with nVentiv chilling does... I don't know yet why and how to explain this, but Im sure that if there is well-enought capable watercooling that keep the CPU's at say 50-60C, then their OC could be even slightly higher.
Dunno why everyone seems "battle" for too low temperature.

When I put my chip on direct sunlight, the laser measuring thermometer says taht the core is about 40C hot. Many users think that 40C on iddle is bad. Now come on - just sun generate the same temp of the core, so... :D

PS. Im sure that every firm making money on cooling things will love to kill me for these thoughts :D
 
trodas said:
I just want to point out that even the OC is not limited by the temperature at all :eek: It maybe sounds insane, but I got two proofs. Neither one of mine CPUs are need a one single Mhz step-down from their max. OC when temp hit 86C (most I tried, then it get too scary even for me :D )
Second, undervolters found, that hot CPU can live and fold hapilly and stable with LESS VOLTAGE, that it then need to start, when it is cold :-/
(translated - you can use AXP-M at 0.9V and 1200Mhz, but it need 1.0 to 1.1V for core to start, when it is cold. When you heat-up it, it live hapilly with 0.9V again)

I think all this could be explained when we look at physic and see, that semiconductors does NOT increase their resistance to elektricity, but they does DECREASE it.

Pretty funny, right?

Translated it means that temperature should help OC and not be problem. Sounds like the exactly oposite that some of the best OC guys with nVentiv chilling does... I don't know yet why and how to explain this, but Im sure that if there is well-enought capable watercooling that keep the CPU's at say 50-60C, then their OC could be even slightly higher.
Dunno why everyone seems "battle" for too low temperature.

When I put my chip on direct sunlight, the laser measuring thermometer says taht the core is about 40C hot. Many users think that 40C on iddle is bad. Now come on - just sun generate the same temp of the core, so... :D

PS. Im sure that every firm making money on cooling things will love to kill me for these thoughts :D

Correct. However, a lower temperature does help with stability thus letting you increase your clockspeed on the same voltage. A 1°C difference probably wont, but I noticed some real nice gains dropping my Thoroughbred B from about 85°C to 55°C. Now, I must note that a newer CPU generally does not take as well to temperatures as older ones. This is also reflected by the maxTDP ratings of the newer CPUs. I am working on a thread listing those ratings.
 
good article. I kind of realised this a while ago. I was worried about 55c, then I though, wait, if AMD says 85 Max, they are probably being on the safe side. What the hell?
 
Stoanhart said:
good article. I kind of realised this a while ago. I was worried about 55c, then I though, wait, if AMD says 85 Max, they are probably being on the safe side. What the hell?

Correct! :clap:

Im not sure if you saw my post in that thread in Intel CPU's about the highest temperature:

I ran my Thoroughbred B at roughly 82°C 100 percent of the time for about a year with no problems. That was at 1.875Vcore and 2.43GHz (only was able to crank it up to 2.53 until I got the temps waaaay down).

Once I had a fan die and my last MBM log was either 122 or 126°C before the computer shut down due to instability.

Usually nasty things only happen to a CPU if it exceeds the manufactures recomended maximum casing temperature by about 50°C and stays there for a while.

Now of course, CPU's differ, a few of the less strong may crap out just a few C above that limit.

Now of course if your CPU is rated for 85°C and it craps out at 84°C, RMA it. The manufacture warranties them up to that point as "operating within design limitations."

Now the manufactur gotta go
 
technically, you shouldn't RMA in most cases, since to get them that high, you are most likely running them over stock voltage and over stock frequency. That voids warranty.
 
Stoanhart said:
technically, you shouldn't RMA in most cases, since to get them that high, you are most likely running them over stock voltage and over stock frequency. That voids warranty.

Correct. However not everyone has super duper overclocker's cooling setups. I remember a buddy buying a motherboard for his Dell and seeing the temperatures. With very quiet low flow cooling and an aluminum heatsink, things can run pretty hot at stock settings.

Anyway, what I should have said is that a CPU is warrenteed up to that level the manufacture lists for that specific CPU.

I am working on a database with that information listed. When I get that done I will probably post a link to it in here. :)
 
Title changed from

"Too many of you are too worried about CPU temperatures!"

to

"Info regarding high temperatures, causes/solutions, and reading from the CPU diode"

by @mdOCer's request. -- macklin01
 
Back