CoreTemp

[75below0]

Anandtech just reviewed the Freezer Pro 7.
The author notes "In comparing results, please keep in mind the test results from the new cooling bed using CoreTemp are not directly comparable to earlier cooling results."
Guess he put that in b/c the Ultima 90 now reads +7C hotter under load than it did when he previously reviewed it.
I'm trying to figure out how much of this increase can be attributed to using CoreTemp as opposed to NvidiaMonitor and Intel TAT like he did in his review of the Ultima 90.
I'm checking forums to try to figure this out and I find

http://forums.anandtech.com/me...AR_FORUMVIEWTMP=Linear

in which CoreTemp reads this guy's CPU +12C hotter than his BIOS says it is?

I'm wondering if CoreTemp is a POS?

Thanks

 

[40sTheme]

Dang, just saw Anand's review... I must be doing something wrong with my ACF7P... Maybe I'll put the TX-2 paste I have on and reseat the sink.
CoreTemp is not a POS. It takes things straight from the DTS on the chip.

 

[yuid]

I also wonder if there is some issue with Coretemp reading high.

I have an E8400, arctic freezer 7 pro in an antec P180b case, and I'm getting an idle temperature of 49C consistently with Coretemp.

I tried disabling the fan speed control in the BIOS (so that the fan runs at max speed all the time), and the temperature stays the same.

When running prime95's torture test, Coretemp will indicate a temperature of ~62-63 tops (won't go any higher than that even after a few hours), and as soon as the test is stopped the temps go back down to ~49-50ish.

The fact that the temperature levels off pretty fast at steady-state high load seems to indicate that the heatsink is doing its job, but it seems strange that I would be getting such a high temperature in the first place...my room isn't that warm...

 

[Riverhound777]

Originally posted by: yuid
I also wonder if there is some issue with Coretemp reading high.

I have an E8400, arctic freezer 7 pro in an antec P180b case, and I'm getting an idle temperature of 49C consistently with Coretemp.

I tried disabling the fan speed control in the BIOS (so that the fan runs at max speed all the time), and the temperature stays the same.

When running prime95's torture test, Coretemp will indicate a temperature of ~62-63 tops (won't go any higher than that even after a few hours), and as soon as the test is stopped the temps go back down to ~49-50ish.

The fact that the temperature levels off pretty fast at steady-state high load seems to indicate that the heatsink is doing its job, but it seems strange that I would be getting such a high temperature in the first place...my room isn't that warm...

Should be a sticky on this somewhere.

http://forums.anandtech.com/me...=2147992&enterthread=y
http://forums.anandtech.com/me...=2148010&enterthread=y
http://forums.anandtech.com/me...=2145748&enterthread=y
http://forums.anandtech.com/me...id=28&threadid=2146935

 

[yuid]

Originally posted by: Riverhound777

Originally posted by: yuid
I also wonder if there is some issue with Coretemp reading high.

I have an E8400, arctic freezer 7 pro in an antec P180b case, and I'm getting an idle temperature of 49C consistently with Coretemp.

I tried disabling the fan speed control in the BIOS (so that the fan runs at max speed all the time), and the temperature stays the same.

When running prime95's torture test, Coretemp will indicate a temperature of ~62-63 tops (won't go any higher than that even after a few hours), and as soon as the test is stopped the temps go back down to ~49-50ish.

The fact that the temperature levels off pretty fast at steady-state high load seems to indicate that the heatsink is doing its job, but it seems strange that I would be getting such a high temperature in the first place...my room isn't that warm...

Should be a sticky on this somewhere.

http://forums.anandtech.com/me...=2147992&enterthread=y
http://forums.anandtech.com/me...=2148010&enterthread=y
http://forums.anandtech.com/me...=2145748&enterthread=y
http://forums.anandtech.com/me...id=28&threadid=2146935

Wow, thanks for the heads-up about that. I had a feeling something was off somewhere.

I wonder if the issue is with the BIOS (I have a DS3L like you), such that a BIOS update would make the temps report properly, or if it's something with Coretemp not recognizing the new wolfdales.

 

[Riverhound777]

Originally posted by: yuid

Originally posted by: Riverhound777

Originally posted by: yuid
I also wonder if there is some issue with Coretemp reading high.

I have an E8400, arctic freezer 7 pro in an antec P180b case, and I'm getting an idle temperature of 49C consistently with Coretemp.

I tried disabling the fan speed control in the BIOS (so that the fan runs at max speed all the time), and the temperature stays the same.

When running prime95's torture test, Coretemp will indicate a temperature of ~62-63 tops (won't go any higher than that even after a few hours), and as soon as the test is stopped the temps go back down to ~49-50ish.

The fact that the temperature levels off pretty fast at steady-state high load seems to indicate that the heatsink is doing its job, but it seems strange that I would be getting such a high temperature in the first place...my room isn't that warm...

Should be a sticky on this somewhere.

http://forums.anandtech.com/me...=2147992&enterthread=y
http://forums.anandtech.com/me...=2148010&enterthread=y
http://forums.anandtech.com/me...=2145748&enterthread=y
http://forums.anandtech.com/me...id=28&threadid=2146935

Wow, thanks for the heads-up about that. I had a feeling something was off somewhere.

I wonder if the issue is with the BIOS (I have a DS3L like you), such that a BIOS update would make the temps report properly, or if it's something with Coretemp not recognizing the new wolfdales.

Or if it is the actual processors which is what I am leaning towards right now. Either the DTS is reading temps wrong, or it really is getting that hot indicating an issue with the IHS or the actual cores.

 

[imported_wonderwrench]

The only way to know your temps for sure is to calibrate your software to read correctly.
Here is a guide to do just that.
http://www.ocforums.com/showthread.php?t=543522

 

[DerwenArtos12]

Originally posted by: wonderwrench
The only way to know your temps for sure is to calibrate your software to read correctly.
Here is a guide to do just that.
http://www.ocforums.com/showthread.php?t=543522

No, the best way to tell temperatures is to use an external thermal probe. Depending on the software and what chip is being used to record and report temperatures there is almost no chance to be completely accurate.

 

[imported_wonderwrench]

Originally posted by: DerwenArtos12

Originally posted by: wonderwrench
The only way to know your temps for sure is to calibrate your software to read correctly.
Here is a guide to do just that.
http://www.ocforums.com/showthread.php?t=543522

No, the best way to tell temperatures is to use an external thermal probe. Depending on the software and what chip is being used to record and report temperatures there is almost no chance to be completely accurate.

Ok, so how do you plan to do this? Cut a grove in the CPU's heat spreader or cut a grove in the heatsink to place the thermal prob. Go for it if you want but I'm going to go with making the best of what I have without chopping my CPU or heatsink up. If you are mounting the probe somewhere else on the heatsink the temp reported would be useless.

The CPU core temps are what really matters and you are not going to read their true temp with a external probe. Software reads the core temps directly from the CPU. The reported temp in speedfan can be calibrate to be accurate to -+1 degree using the method in the guide.

 

[DerwenArtos12]

Originally posted by: wonderwrench

Originally posted by: DerwenArtos12

Originally posted by: wonderwrench
The only way to know your temps for sure is to calibrate your software to read correctly.
Here is a guide to do just that.
http://www.ocforums.com/showthread.php?t=543522

No, the best way to tell temperatures is to use an external thermal probe. Depending on the software and what chip is being used to record and report temperatures there is almost no chance to be completely accurate.

Ok, so how do you plan to do this? Cut a grove in the CPU's heat spreader or cut a grove in the heatsink to place the thermal prob. Go for it if you want but I'm going to go with making the best of what I have without chopping my CPU or heatsink up. If you are mounting the probe somewhere else on the heatsink the temp reported would be useless.

The CPU core temps are what really matters and you are not going to read their true temp with a external probe. Software reads the core temps directly from the CPU. The reported temp in speedfan can be calibrate to be accurate to -+1 degree using the method in the guide.

I read that entire guide and I do not believe that without any external calibrated sensor you can get an accurate reading. That review relies on the accuracy of the probes which is completely unknown. Do you really think that the temperatures on the sides of the IHS are much different than the top of the IHS? and if you're going to go through all the trouble of attempting to calibrate some 3rd party software so that all the softwares are consistantly reporting temperatures of unknown accuracy due to thermal probes that are in unknown locations reporting temperatures that we do not know the accuracy of, why not just remove the IHS and place a thermal probe on the chip either next to or between the cores. With it saturated in thermal paste you think the temperatures is going to be off by more than 1-2C?

Case in point: the new E8400 and 8500's thermal probes. If you can't trust one why do people think they can trust the other one?

 

[imported_wonderwrench]

In theory when you underclock a C2D to 1.6 ghz @ 1.25 volts the heat output is so low even a stock heatsink can keep Tcase temp near ambient temp, ambient to +1C. This is with case covers off. Using a known accurate thermometer to measure ambient temp placed near a case intake fan. This should give a fairly accurate reading. For example, with the system under test at idle for 10 min or more the Tcase temp reported by speedfan is 29C. Ambient temp is 22C. Add 1C to this and you have a Tcase temp of 23C. So 29C reported in speed fan - 23C corrected Tcase temp = 6. So so we need to enter a temp offset in speed fan of -6 so speed fan reports 23C. This may not be a perfect calibration but its way closer than doing nothing. On the system in my sig speedfan reported Tcase temp higher than Tjunction temp so I knew something was up as that is not possible. When I ran the Tcase calibration I found Tcase temp reported was 7C too high.

As far as Tjunction calibration goes. Intel claims the thermal diodes used are accurate + or - 1C. Software reads this directly from the CPU. The softwares accuracy is dependent on whether or not the software knows the Tjunction Max temp, either 85C or 100C. Its also possible for the software to read Tjunction temp incorrectly regardless. So calibration maybe needed. Since CrystalCPUID can read CPU registers and the guide gives us a way to read the correct register we can check Tjunction temp accuracy vary easily.

Temps reported by speedfan on my system before calibration were not even believable.

idle load
Tcase 34C 61C
Core0 23C 52C
Core1 24C 52C
Ambient 21C 21C

After calibration

idle load
Tcase 27C 54C
Core0 38C 67C
Core1 39C 67C
Ambient 21C 21C

The temps after calibration may not be spot on but are way closer than before calibration.
I now feel more confident knowing my Tcase and Tjunction temps are being reported close to there actual temps by speedfan. Before I had no way of knowing accuracy. If Computronix is correct in his theory my reported temps are accurate to + or - 1C. I can live with that no problem.