PDA

View Full Version : In socket thermistor's (lack of) accuracy


NeoMoses
12-05-02, 07:23 AM
So I'm up and running on my new 8k5a2+, which has an in socket thermistor rather than reading the on die diode. I thought to myself, "how far off could it be?" running the same cooling, same CPU, same everything on an 8k3A+ was giving me full load temps of 34-38C depending on ambient temps. Right now I am sitting at 52C load (according to MBM5). I've tried re-seating this WB 3 times now and am sure it's seated well on the die.

I never realized these things were so innacurate. i was expecting to see temps drop if anything. Now I know that the range of accuracy isn't even +/-10C. More like +/-15C. These things suck! Looks like a Nexus is in my future, I hate not having reliable temps.

Moral of the story: in-socket thermistors are horribly innacurate. don't trust them at all. Epox really took a step backwards by switching from on-die to in-socket temps in this mobo.

paulie
12-05-02, 09:34 AM
Yup, in socket sucks big time. It makes chumps out there actually believe that their cpu is only 3c above ambient :rolleyes:
I remember people out there slagging the Abit ST6 for giving high readings, but it does reads on die, so it is accurate :D

NeoMoses
12-05-02, 10:07 AM
wow, this really got me thinking. all along, i've been giving advice like, "if your cpu temp is below 50C for an XP, it should be ok." A temp could be reported as 50C and actually be nearing 65C. wow! I'll bet many people have been led astray thinking their RAM or something else is holding back their overclocking, when actually it's CPU temp.