Have you ever wondered why Core 2 temps people report vary so greatly. I have but after reading a post by CompuTronix over at OC forums, I now understand why. It would seem most motherboard manufactures do not program their boards to correctly report CPU case temp, AKA Tcase temp. On top of that temp reporting software may report core temps, AKA Tjunction temp incorrectly. 15C or more is not uncommon. This means people actually have no idea what their processor temps are. The upside is CompuTronix has come up with a method to calibrate Tcase and Tjunction temps reported by speedfan so users can be more confident of their temps. This also will allow people to more accurately compare temps reported by others. http://www.ocforums.com/showthread.php?t=543522
To be able to compare Idle and load temp results we need a way to eliminate variables. The best way to do this is to test using a common method. After some thought I came up with a method that should eliminate most variables and allow people to compare their temps with others.
Things that need to be known and or controlled.
1) What OS is being used.
2) What software is being used to report temps. Speedfan is the best bet as it allows compensation factors to be saved within the program. No matter what software is used the reported temps must be compensated for by using CompuTronix method.
3) What software is being used to load the CPU's. Prime95 is well known and free. Must be run using CompuTronix method.
4) Motherboard used, model, revision, bios version. Different boards may over or under volt cpu's so it needs to be known for comparison purposes.
5) CPU model and stepping.
6) Frequency and voltage used during the test.
7) Heat sink used as well as compound used.
8) CPU fan and case fans at full speed.
9) Ambient temp = room temp measured near case intake fan using an accurate thermometer. To keep case cooling out of the mix all tests should be done with case covers off. Room temp should be kept as close to 22C as possible so results can be more easily compared. As long as ambient temp is correctly measured and tests are done with the case covers off in theory we should be able to add or subtract the ambient temp difference from Tcase and Tjunction temps keeping the comparisons accurate.
Temp results needed
CPU temp = Tcase temp at idle and under load.
Core temp = Tjunction temp for each core at idle and under load.
Ambient temp = Temp measured using an accurate thermometer positioned near case intake fan.
My results.
XP pro sp2 32 bit
Speedfan 4.33 beta 37 used to report temps. Calibrated using the Computronix method.
Prime95 ver 255 use to load CPU's. set per Computronix method.
MSI P35 Neo2-FR ver 1.1 bios 1.6
E6400 L2 stepping at 3ghz 1.325 volts
Retail intel heatsink using AS5
All fans at 100% case covers off.
Idle temps
CPU 27C
Core0 38C
Core1 39C
Ambient 21C
Load temps
CPU 54C
Core0 67C
Core1 67C
Ambient 21C
Compensation factors I ended up needing for correct temp reporting.
subtracted 7C from reported CPU temp AKA Tcase temp.
Added 15C to reported Core temps AKA Tjunction temps.
Note: Once I started using the latest speedfan beta I no longer
needed to add 15C to the Tjunction temps as speedfan beta 37 correctly reports
Tjunction temp at least in my case.
Keep in mind when changing some variables your reported temps may become inaccurate so a recalibration may be needed. Examples: Bios update, speedfan update or similar. Also keep in mind that CompuTronix method only improves temp accuracy it does not give you 100% accuracy.
Please post your results
Note: all credit should be given to Computronix and his great guide!
To be able to compare Idle and load temp results we need a way to eliminate variables. The best way to do this is to test using a common method. After some thought I came up with a method that should eliminate most variables and allow people to compare their temps with others.
Things that need to be known and or controlled.
1) What OS is being used.
2) What software is being used to report temps. Speedfan is the best bet as it allows compensation factors to be saved within the program. No matter what software is used the reported temps must be compensated for by using CompuTronix method.
3) What software is being used to load the CPU's. Prime95 is well known and free. Must be run using CompuTronix method.
4) Motherboard used, model, revision, bios version. Different boards may over or under volt cpu's so it needs to be known for comparison purposes.
5) CPU model and stepping.
6) Frequency and voltage used during the test.
7) Heat sink used as well as compound used.
8) CPU fan and case fans at full speed.
9) Ambient temp = room temp measured near case intake fan using an accurate thermometer. To keep case cooling out of the mix all tests should be done with case covers off. Room temp should be kept as close to 22C as possible so results can be more easily compared. As long as ambient temp is correctly measured and tests are done with the case covers off in theory we should be able to add or subtract the ambient temp difference from Tcase and Tjunction temps keeping the comparisons accurate.
Temp results needed
CPU temp = Tcase temp at idle and under load.
Core temp = Tjunction temp for each core at idle and under load.
Ambient temp = Temp measured using an accurate thermometer positioned near case intake fan.
My results.
XP pro sp2 32 bit
Speedfan 4.33 beta 37 used to report temps. Calibrated using the Computronix method.
Prime95 ver 255 use to load CPU's. set per Computronix method.
MSI P35 Neo2-FR ver 1.1 bios 1.6
E6400 L2 stepping at 3ghz 1.325 volts
Retail intel heatsink using AS5
All fans at 100% case covers off.
Idle temps
CPU 27C
Core0 38C
Core1 39C
Ambient 21C
Load temps
CPU 54C
Core0 67C
Core1 67C
Ambient 21C
Compensation factors I ended up needing for correct temp reporting.
subtracted 7C from reported CPU temp AKA Tcase temp.
Added 15C to reported Core temps AKA Tjunction temps.
Note: Once I started using the latest speedfan beta I no longer
needed to add 15C to the Tjunction temps as speedfan beta 37 correctly reports
Tjunction temp at least in my case.
Keep in mind when changing some variables your reported temps may become inaccurate so a recalibration may be needed. Examples: Bios update, speedfan update or similar. Also keep in mind that CompuTronix method only improves temp accuracy it does not give you 100% accuracy.
Please post your results
Note: all credit should be given to Computronix and his great guide!
Facebook
Twitter