KingCody
Posts: 964 +8
Hi all 
I am hoping that this may help some people who are wondering about the accuracy of software based temp monitoring programs (MM5, speedfan, etc.)
My software temp sensing experiment:
my case has a cheap 3-probe temp sensor built into it. a while after i got it i borrowed an infared thermometer (Extech 42500 Digital IR Thermometer) from my brother in-law to see how accurate my case sensors were. which I could then compare to software based temp monitoring programs.
i had one of the temp sensors glued to the PCB of my AthlonXP 2600+ with the tip of the sensor physically touching the CPU die. also, I wired the case's LCD temp display with a 12v power pack from an old police scanner so that it would stay on constantly, and not shutoff with the PC.
since I obviously could not verify the temp of the die via the IR with the heatsink on, i removed the heatsink (and thoroughly cleaned the die).
to avoid overheating the CPU, I set my mobo set to shut down when the CPU reached 65C (the lowest setting possible). the max die temp for that processor was 95C so I wasn't too worried about it.
I turned on the PC and within seconds it shut down. according to the BIOS setting, that would mean it should have heated up to 65C... yet my temp sensor only read 56C and the IR read 58C. (i will assume that the IR is more accurate than anything else i had access to) that's a difference of 7C. after a few more tries i had found that my temp sensor was always 1C-2C lower than the IR reading (a negligible amount).
I now trusted that my case's display was accurate to within 1C-2C of the IR, which was better than I expected.
i then reinstalled the heatsink and booted into the BIOS. the BIOS read a constant 43C, and my LCD display read a constant 39C (the laser would have read 40C-41C). so that proved (to me) that the BIOS was reading the temp about 2C too high. no big deal, but this wasn't stressing the CPU
i then booted into windows and monitored it through MM5 and speedfan. i ran halo, F.E.A.R., and superPI. the software temps were all over the place. speedfan read 3C-8C higher than my LCD. MM5 stayed more consistent but still read 6C-7C higher.
the bottom line... software temp monitors are not accurate enough to rely on. they can be useful as a "before and after" comparison tool to see if a change in your cooling made any difference, but that's about it.
I hope this was helpful (or at least entertaining.. lol :haha: ) hopefully i got all the numbers right because i was going off a notepad file which has the temps on it (i had done this experiment quite some time ago)
cheers :wave:
I am hoping that this may help some people who are wondering about the accuracy of software based temp monitoring programs (MM5, speedfan, etc.)My software temp sensing experiment:
my case has a cheap 3-probe temp sensor built into it. a while after i got it i borrowed an infared thermometer (Extech 42500 Digital IR Thermometer) from my brother in-law to see how accurate my case sensors were. which I could then compare to software based temp monitoring programs.
i had one of the temp sensors glued to the PCB of my AthlonXP 2600+ with the tip of the sensor physically touching the CPU die. also, I wired the case's LCD temp display with a 12v power pack from an old police scanner so that it would stay on constantly, and not shutoff with the PC.
since I obviously could not verify the temp of the die via the IR with the heatsink on, i removed the heatsink (and thoroughly cleaned the die).
to avoid overheating the CPU, I set my mobo set to shut down when the CPU reached 65C (the lowest setting possible). the max die temp for that processor was 95C so I wasn't too worried about it.
I turned on the PC and within seconds it shut down. according to the BIOS setting, that would mean it should have heated up to 65C... yet my temp sensor only read 56C and the IR read 58C. (i will assume that the IR is more accurate than anything else i had access to) that's a difference of 7C. after a few more tries i had found that my temp sensor was always 1C-2C lower than the IR reading (a negligible amount).
I now trusted that my case's display was accurate to within 1C-2C of the IR, which was better than I expected.
i then reinstalled the heatsink and booted into the BIOS. the BIOS read a constant 43C, and my LCD display read a constant 39C (the laser would have read 40C-41C). so that proved (to me) that the BIOS was reading the temp about 2C too high. no big deal, but this wasn't stressing the CPU
i then booted into windows and monitored it through MM5 and speedfan. i ran halo, F.E.A.R., and superPI. the software temps were all over the place. speedfan read 3C-8C higher than my LCD. MM5 stayed more consistent but still read 6C-7C higher.
the bottom line... software temp monitors are not accurate enough to rely on. they can be useful as a "before and after" comparison tool to see if a change in your cooling made any difference, but that's about it.
I hope this was helpful (or at least entertaining.. lol :haha: ) hopefully i got all the numbers right because i was going off a notepad file which has the temps on it (i had done this experiment quite some time ago)
cheers :wave:
