I came acrossed this article at Overclockers.com and thought I would pass it on, I don't have a dual core CPU anymore to be able to test with, so I am hoping someone would give it a go and report back there findings here

Download "Core Temp" below
Core Temp

Original article:
Core Temp

Nice find! That is a very interesting read.

Ran it for a couple of minutes and the temps seem to fluctuate about 3 degrees, but the temps are never more than 2 degrees different (On an Opty 170).

Arnt the cores on the same die?

I have used that now and then, but I get up to a 20C difference between core temp and PC Probe. When I check the bios temp, it looks like PC Probe is the more accurate.

Thought i'd give it a whirl & i have 20+c difference in core temps according to this, but according to MBM5 i'm idling @ 25c

Attached Thumbnails

 

Here is an example: Core Temp is showing 50C or better and Asus PC Probe (which I am sure is better tuned for the sensors) reads 37C. Right now I have my rig backed down to stock speed and voltage. A reboot and look in the bios showed 35C, so I would trust PC Probe more than Core Temp.

Attached Thumbnails

 

From what I understand, all temp reading program (Including the Bios) will read from the same thermal diode.

Now as to how accurate they are, soley depends on the formula they use to calculte the temperature based upon the out put of that thermal diode.

My guess is, this guy uses a different equation than most, and I often wonder if it's more accurate!

The reason being, is when I did use an older version (No dual core support) of this on my CPU, it seemed to read at least 10-15c higher than any other program, including my Bios.

Now the clue to m,e it may be more accurate was when my CPU would always flake out (When Overclocked) any time it got above a "Reported" 50c most other programs, I had stability issues...odd that this was well below AMD's rating of 60c+ for my CPU?

Now you may say..but you were overclocked! and that may have some effect on it, because I have seen temps at stock speeds and voltages exceed 60c and it still kept running (Ran a cpu with heat sink only, fan disabled)

However, like most things in the corporate world, the "Threshold" that they chose of around 60c is very conservative, and would probably handle upwards of 70c for short to medium periods before damage would occur.

Chances are the CPU could will live at 60c fullly loaded for 5+ years so they would never have to worry about returns!

How long would it last at 70c loaded? My guess is probably only a year or two...probably not very good for warranty reasons!

I suppose the only way we will ever know what is somewhat accurate would be to remove the heat spreader spreader, and desing one of two way of measuring temps.

One would be to use a Mil.Spec calibrate thermocouple placed on the center of the die.

The other would be to make a "Swiss Cheese" like heat sink, with a fan mounted on the side instead of on top,and then use a calibrated laser thermal temp reader...you would have to shine it down the holes.

Probably to hard to do by holding it since we "Shake" too much, but I am sure you could use something like a "Holder" like device they use in quality labs when doing measurements...I would think it wouldn't be too hard to come up with something similar at tool supply store.

Again..this may not even be the best method, because we are only getting a surface reading, and we are realy looking for the "Internal" temps of the cores, and that may be much different.

Once agaiin, I was just doing a bit of "Daydreaming" here...



Ralph, Ralph...RALPH! you can go home now, class is over...

(For those of you who used to watch alot of Buggs Bunny "Looney Tunes" and are a bit older..from an episode called "Boyhood Daze")

Plot Summary for Boyhood Daze (1957)

Attached Images

 

Quote: Originally Posted by $SOLID$ Necro One would be to use a Mil.Spec calibrate thermocouple placed on the center of the die.

is that a real thing that is supposedly better than a themometer with NIST certification? it looks like the thermal diode readings typically have better than 5 degrees accuracy according to the manufactures. on the national spec sheet for the chip that reads the thermal diode it says +/- 3 degrees c. (first spec sheets that pops up on google btw) i've never been one to trust onbard thermal sensing either.

the app says my second core is 3 degrees hotter and in task manager it says it is doing more work. fun times

Again, I would refer to my previous post. I was at stock. At OC I get up to a 20C varience. The varience was the same with the Intel Bad Axe board I was running and the Asus board I am currently using.

My opinion, the CPU Temp program is Sh*t and isn't accurate. I had deleted it but downloaded it again just so I could show the temp difference. It is already deleted again.

Noted BA...I am glad to hear an opinion such as this, I am pretty open to listen to what others have to say.

I have ben told quite often I am a bit stubborn unless I see some hardcore evidence that makes me change my mind though, and it has to come from someone that is "In the know" (The last part I will take some exception too, even a nOOb can get me to change my opinion if there logic is valid, and not just talking out there "Blow Hole" lol)

Your opinion caries a lot of weight in my book, even if it differs from mine....I wish I had some more time and the hardware to investigate this properly..I only care about being accurate, not about who is right or wrong...too many people I see get confused by that on forums, and let that get in the way.

Even if you don't care for this program, accuracy aside, the effort was made to let us see the temps of both cores, that in itself merits kudos from me

Especially with C2D I have always felt weird about the differences in temperatures between monitoring apps. Personally I feel core temp is crap. There is a big thread over at XS which talks about core temp. (To lazy to search for it )

Cool util.. I found that Everest functions similar when set to display temps in the system tray.

The information I get seems to be fairly close between SpeedFan and CoreTemp... disgustingly high.

~49-53c on each core under token (<10% total) load. Case temperature 38c.

Is this sometrhing to be expected on a setup with a 4600+ under an AC Freezer 64 Pro (at full blast)? I tried turning up the case exhaust (120mm, from 1200RPM to 2200) and it only cuts about 3c off everything. CoreTemp announces 'TCaseMax = 59c", so I'm thinking letting it run up under full load will not be goodness.

Im a smelly spammer, please use your l337 skillz and hax0r me.

67.159.45.52

brosne@gmail.com