Different Core Temps In CPU

I have a AMD Opteron 180 dual core cpu running on a ABIT AN8 motherboard with Win.7 and the different core temps are pusseling me. I have tried with 4 different cpu coolers, stock cooler for Opteron 180, stock cooler for phenom 2, Coolermaster Hyper TX3 and Arctic Ice 64.
Using Arctic Silver 5 with each cooler. Temps are relatively even amongst the 4 coolers 5- to 6 ºC differences, but the thing that concerns me is, that with each of these 4 coolers installed the difference between Core #0 and core #1 is always around 10ºC. If core #0 is as low as 16 ºC core #1 would be 26 ºC with a max of 32 ºC with core #0 on 23ºC.
I know it is not the coolers nor the thermal compound as I use the same method on all occasions.
This leaves the motherboard or the cpu in my opinion.
Any other ideas ?? I would love to hear about them.
hawk22

 

Windows uses the first core extensively & all the time. So you can expect some difference in temps, but around 10 C seems to be excessive.

How are you measuring the temps ? What programs are running ?

 

I use CPUID HW Monitor and Core Temp and they both tell me the same.

 

Are you experiencing performance problems? If not, I would not worry about it. Many programs are not coded to support more than one core.

Or maybe the sensor is wrong. I would not worry unless the temps were too high, or performance was impacted.

Were those figures you gave just illustrations of the temperature differences? I realize you are approaching Winter where you are, but 16°C is 60.8°F and without some sort of refrigerated cooling, it is impossible to get a CPU below ambient (room) temperatures.

Are your room temperatures really that low? Even a core temp of 26°C is only 78.8°F and that is not even breaking a sweat for a CPU. 60°C for a CPU is just starting to approach the "very warm/hot" stage.

So if your temps are true, and the low 30s C are your highs, I envy you!

 

Hi and thanks Bill, no I do not have any performance problems, she is performing very well indeed. Ambient temp would be around 20ºC.

I have just done another check with Core Temp v. 0.99.8 and the results are as before,
Core #0 low 29ºC high 43ºC
Core #1 low 16ºC high 35ºC
that are the measurements with a load of up to 18%

CPUID Hardware Monitor

Core #0 16ºC 60ºF
Core #1 30ºC 85ºF
these are the lowest temps with Core# 0 going up to 25ºC

I have been doing more googling on this matter and have found that a 10ºC difference between cores is quite acceptable.

Also I have found that some ABIT AN8 series Motherboards have a history of
unreliable sensors.

I have a AMD Phenom II x4 965 and there is never more than 2ºC between the 4 cores but average temps run at 41ºC with a very mild overclock.

 

Something is funny. If your room temp is around 20°C, there's no way blowing 20°C air on a device will cool the device cooler than the air temperature. Physics and thermodynamics don't work that way on chunks of metal - especially when that device is generating lots of heat.

I suspect you have a slight accuracy issue with one or more of your sensors (not uncommon - they are pretty low-tech). I would not be concerned as long as it is consistent - that's the key because that still gives you a baseline to gauge from.

You are watching it, and that is the important thing.

Note I start getting nervous when my CPU temps hit 60°C (140°F) so your 41°C is great.

 

Thanks Bill, and yes I would agree with you on that one, blowing 20ºC air onto a piece of heated aluminium is not going to bring it down to 16ºC, not without a Peltier anyways.
When I played around with overclocking the Phenom II x 4 965 BE on the stock cooler under full load it shot up to 69ºC and that scared the pants of me.
With the same overclock now with a Coolermaster Hyper TX3 it will go to 59ºC and to my liking that is to high as well, so as I mentioned in my last post on a very mild OC she is cruising along at 41ºC.
hawk22

 

I have an ABIT KN9. Originally it was showing CPU temperatures about 10 or 15 degrees C below ambient temps (like 5*C ). I think you need to upgrade the chipset drivers. I now show temps around 30*C at idle, but I am not certain upgrading the drivers made the difference.

See if this is the model:
http://www.cdrinfo.com/Sections/Reviews/Print.aspx?ArticleId=16819
If so, you would need to update the NVIDIA NF4 chipset drivers.

Matt

 

Also...don't forget to check your readouts in the BIOS. Do a restart and check them there. They should be still close to what was read under Windows.

 

At idle!

Running the BIOS Setup Menu is probably the least demanding of all tasks we can ask of our computers. Also, computer parts heat up fast, but also cool down fast. And since you have to reboot to get into the BIOS Setup Menu, your system has had a few seconds of zero activity which allows the system and key components more time to cool. So the temps you see in the BIOS Setup Menu will be the lowest you will see them.

If high in the BIOS Setup Menu, you have problems. If acceptable or low in the BIOS, that's good, but not a conclusive indication of cooling performance while under load.

If you changed nothing else, I bet it was the drivers.

I think checking ABIT's website for new chipset drivers/BIOS update is an excellent suggestion.

 

Thanks guy's I will do that.
hawk22

 

Sorry about the delay.

No Matt, that is not the one that I have, and I have re checked the chipset drivers and I do have the last available installed.
This is my board:

http://www.abit.com.tw/page/en/moth...hp?pMODEL_NAME=Fatal1ty+AN8&fMTYPE=Socket+939

This board comes with the ABIT Guru Clock, for those who are familiar with the ABIT board, while in the BIOS the Guru clock shows CPU 37ºC and the same is the case for the CPU temp in the BIOS.
Neither of the two will show the Core temps core#0 or Core#1 only CPU temp.
But from my experience there is not a great deal of difference in general between core temps and cpu temps.

So I guess that I can take the 37ºC as roughly my CPU or Core Temp in idle.
And I am happy enough with that.
hawk22

 

I don't install Guru. I would only install it if I thought I needed it.

I use SIW (System Information for Windows) and SiSoftware Sandra which give me both core temps (currently 24 and 27*C).

If you took it that the Abit readouts were the basis and the dual core readouts may not be (completely) accurate, but always get quite a variation between the temperatures of the cores, there could be a slight glitch when you installed the CPU heatsink. The thermal paste could be slightly thicker over one core, a small air pocket or something like that. One tip I found for when you are installing the heatsink was just before applying the clamps, push down on the h/s and twist left and right. [Not too hard of course, to avoid cracking the motherboard.] This thins the compound and can push out any air bubbles.

I do it every time now. You can really only wiggle the heatsink slightly. If I think a CPU has temperature problems, I undo the clamps and do the wiggling, it shouldn't disturb the thermal compound unless you pull on it and break it's seal. If the temperatures were worse afterwards, I would remove the heatsink, clean it and start again. You only improve your technique for applying the compound. Some compounds cure hard and you couldn't wiggle them without breaking it and that could depend on how long the compound has been installed.

If the problem keeps bugging you, set aside some time to try and overcome it. Replacing the thermal compound needs "technique" rather than just following instructions. If you pull a heatsink off a CPU, look at the surfaces and see if you can add to your own technique of applying the compound.

As you would know, the aim is to get the thinnest possible layer of thermal compound, that does not have any air bubbles trapped in it.

Matt
PS: This might be a tip...I go into the BIOS and set the "Alarm" and "Shutdown" temperatures down slightly, then if the system overheats and shuts down, I KNOW there is a temperature problem

 

If you can squirm the TIM (thermal interface material) by pushing down and twisting the HSF, then I would suggest you have too much TIM applied. Thinner is better. The idea is to fill in the microscopic pits and valleys in the mating surfaces of the heatsink and CPU die to remove any trapped air that may act as an insulator.

It is important to note the best heat transfer occurs with direct metal-to-metal contact. So any extra TIM is in the way, and counterproductive. I agree 100% with Matt that applying TIM requires a technique and some practice. For what it is worth, here is my canned text on applying TIM.
***********

An often misunderstood and sometimes overlooked critical hardware component is thermal interface material or TIM. TIM is typically seen as a thermal pad on a CPU heatsink, or in paste form. It may also be called thermal grease, silicon grease, heat transfer compound, thermal paste, heat sink compound, or goop. There are probably several more names.

The purpose of TIM is to ensure all the microscopic pits and valleys in the CPU die and heatsink mating surfaces are void of heat trapping air, maximizing surface to surface contact. Any excess is too much and gets in the way, and can actually be counterproductive to the heat transfer process.

The 4 Most Common Heatsink Fan (HSF) Assembly Mounting Mistakes:

1. Failure to use TIM
2. Used too much TIM
3. Reused old TIM
4. Did not clean mating surfaces thoroughly before applying TIM

Materials Needed: One clean plastic shaft Q-Tip (cotton swab), acetone or 91% isopropyl alcohol (Note - most rubbing alcohol is 70% and leaves a film. 91% alcohol can be found at your local drug store), clean scissors, can of compressed dusting gas, and the TIM. I recommend one of the new generations of non-metallic TIMs such as AC MX-2, Tuniq TX-2 or, OCZ Freeze or the venerable silver based TIM, Arctic Silver 5.

WARNING: Keep yourself grounded with the case to ensure there is no static buildup and discharge that might destroy any electrostatic discharge (ESD) sensitive devices. It is important to realize that the "threshold for human awareness" for a static shock is higher than the tolerance of ESD sensitive devices. In other words, you can shock and destroy a CPU, RAM module, or other sensitive device without even knowing there was a static discharge! Use an anti-static wrist-strap or frequently touch bare metal on the case to maintain your body at the same potential as chassis (case) ground.

Preparation: Power off and unplug the computer from the wall. Cut off one cotton swap near the end. Bend the plastic shaft about 1/2 inch from the cut end to make a nice little hockey stick. This is the working end of your TIM application device. Clean the die and heat sink mating surfaces with a soft, lint free cloth dampened (not dripping wet) with acetone or 91% alcohol. Do not let any fluids run down the sides of the CPU die. Clean skin oils from the working end of your applicator with the alcohol dampened cloth. Blast the surfaces with a quick blast of compressed air to ensure the surfaces are dry and no lint or dust remains behind. Do NOT touch the CPU die or heatsink mating surfaces, or the applicator's working end from this point on.

Application: Apply one "drop" of paste, about the size of a grain of rice, on the corner of the die and spread it out across the die with the applicator, like spreading icing on a cake. Spread the paste as thin as possible while ensuring complete coverage. It is easier to add more than remove excess. Remember, too much is counterproductive.

Note 1: Depending on the type of TIM used, some, such as the silver based compounds, can take 2 - 5 days or longer (depending on the power/heat up-cool down cycles) for the TIM to cure and reach optimum effectiveness. A 2 "“ 4°C drop in average temperatures may be realized after curing.

Note 2: A new HSF may come with a thermal interface pad already applied. Those pads consist of mostly paraffin, which is supposed to melt and squirm out of the way when the CPU heats up for the first time. Thermal pads are certainly better than no TIM at all, but they are not as effective as silver or ceramic based compounds. Do not use a sharp or metal object to remove the pad. A fingernail will work fine, removing any residue with acetone or alcohol.

Note 3: Do not reuse a thermal pad or paste. Always remove the old, cured TIM, clean the mating surfaces thoroughly, and apply a fresh application of new TIM.

Note 4: Thermal adhesive is a specific type of TIM used to permanently or semi-permanently glue heatsinks to devices that have no other heatsink mounting mechanism. Thermal adhesive is NOT intended to be used between a CPU and the CPU heatsink.

Note 5: TIM is also used to ensure maximum heat transfer to the heatsink from graphics processor units (GPUs), chipsets, graphics card memory modules, and other devices. Adhesive TIM, as mentioned in the note above, is often used on these devices as many do not have mounting brackets or holes to support a clamping mechanism. When mounting a heatsink to one of these components, the idea is the same; apply as thin a layer of TIM as possible, while still ensuring complete coverage.

See Benchmark Reviews 33-Way TIM Comparison or TweakTown TIM Review for additional information.
***********
​

 

Thank you Matt and Bill, I will do another re-seat of the cooler on the weekend and see if I will get a different reading.
I will keep you posted.
Thanks again
hawk22

 

Well, unless your current application of TIM is missing or very poorly done, you should not expect any major changes. 3 - 5°C at most is all I would expect. And note, that may not be achieved until a few days of curing has passed.

 

No surprises here, much as I had expected.
Re-seated the HSF using as always Arctic Silver 5, except this time I reverted back to my old method of applying the compound by spreading it very thinly over the cpu surface, opposed to Arctic Silvers recommended use of a uncooked grain of rice size dob in the middle of the cpu.
CPU Temp is steady on 31ºC and the cores are much the same as before with a low of 23ºC - 29ºC for core #0 and low of 13 ºC - 19ºC for core #1. With 10ºC between the cores as before, which brings me to the firm conclusion that the motherboard sensors are unreliable.
I will ignore the Core temps and just keep my eye on the cpu temps as usual.

Thanks again for your valuable advice that I appreciate very much.
hawk22

 

Sounds like a plan!