A recent problem give me things to think about. Temperatures are extremely important for a stable computer, more if one plans to overclock. Several programs out there measure the temperature of our CPUs, yet... almost everyone gives different results... Which to believe? If they measure the same, how can the results be so different? Im talking about 10c variations between some of them...
Oh well... What do you use and why?
Message edited by bodhidharmazen on 07-26-2008 at 04:58:01 AM
CoreTemp is high & RealTemp is low. Speedfan is in the middle, & probably safer.
CoreTemp & RealTemp, although about a 10c difference, both report the same distance/delta to TjMax. Whatever the TJunction is set to, the distance/delta to TjMax stays the same.
So keeping the distance/delta to TjMax above 30c-35c would probably be a safer way to measure temperature.
By dead on, do you mean changing the TjMax or leaving it as is?
The CPU temperature maybe, but what about core temperature?
Since no one knows the TjMax, how can you be sure? On 45nm RealTemp uses 95c TjMax , SpeedFan uses 100c TjMAx, & CoreTemp uses 105c TjMax.
I guess since you can change TjMax in RealTemp & if you knew what the TjMax was, then you could calibrate RealTemp to be dead on?
By dead on, do you mean changing the TjMax or leaving it as is? The CPU temperature maybe, but what about core temperature? Since no one knows the TjMax, how can you be sure? On 45nm RealTemp uses 95c TjMax , SpeedFan uses 100c TjMAx, & CoreTemp uses 105c TjMax.
I guess since you can change TjMax in RealTemp & if you knew what the TjMax was, then you could calibrate RealTemp to be dead on?
Real Temp measures the core temp, not case temp.
Real Temp is the best by default because unclewebb, the author, has taken great pains to do actual testing of the tjmax for some of the processors. Don't believe me, read it for yourself.
It all depends on the machine I am using as all of mine are different and either require or work better with a different method of monitoring.
1. Desktop (X2 4200+)
- Sensor used: CPU core digital temp sensor (driver = k8temp)
- Monitor used: GkrellM
2. Notebook (C2D U7500)
- Sensor used: CPU core digital temp sensor (driver = coretemp)
- Monitor used: GNOME hardware sensor applet (screen is too small for GkrellM to sit off on one side)
3. HTPC (900 MHz Celeron Coppermine-128)
- Sensor used: ACPI thermal zone temperature reading from on-package Analog Devices sensor
- Monitor used: The machine is only hooked up to a CRT TV, a homebrew RS-232 IR remote receiver, and a wireless bridge, so I just execute a "cat /proc/acpi/thermal_zone/THRM/temperature" when I want to see the temperature. This sensor is known to be needing a certain offset that is outlined in the "Intel Celeron up to 1.10 GHz" data sheet, so I have to mentally add about 5 C to the temperature to get the proper temp, +/- 1 C.
---------------
UNIX is user-friendly- it's just picky who its friends are.
DRM is slowly killing personal computing, one Sony rootkit and TPM chip at a time.
Another alternative to feel safer, is to go by the Tcase sensor. That would be the sensor that your bios reads, MB utility App, SpeedFan also reads it. It tricky with Speed Fan since you will need to figure that out on loading your cores to watch the temps go up. Sometimes it labeled CPU, other temps it maybe Temp1 or Temp2, which can be re-named.
I only say this since Intel does publish the thermal spec of a CPU, and generally would be the Tcase max, in which you shouldn't exceed.
However I think its still niffy to be able to see the separate core temps.
On a side note, you can adjust the tjmax on Coretemp/RealTemp, and offset cores on SpeedFan. For Vista (since speedfan doesn't reconize my gpu temp) I use speed fan to show the Tcase temp, while coretemp shows, obviously the core temps. (edit: display on the taskbar)
So I don't exactly have a favorite.
Message edited by Grimmy on 07-26-2008 at 09:33:04 PM
