Silverback
Based on the reatail heatsink, which has a C/W of .5(and this hsa been tested to be around .51-.55 on Internal Diode/Pentium 3 tests), a ~60W processor would run at 30C over ambient.
With a better heatsink, like a Swiftech or an Alpha with a C/W ~.3, the rise over ambient would be ~20C over ambient. That's with the best cooling units. So 10-15C over ambient would need water cooling to be true.
Also note that while overclockers.com says that +-5C is needed to adjust socket-thermistors, it isn't true with XP chips. The NEW PCB further blocks heat from the backside measurements. It knocks socket-thermistors to a level where they are even more inaccurate.
Mike
P.S. If you're gonna say Overclockers.com is in LA lA land, please note that the cooling articles are done by Joe, not that rambling fool Ed.
P.S.2. TO further refute your claims of slight inaccuracy, i've done a chart based on HardOCP(which has one of the most terrible testing methods) versus the Overclockers.com DIE simulator tests. I know how the die tests are done, and they're of pretty damn good accuracy for getting "die C/W".
With one exception of the millenium glaciator II(where the difference between DIE and socket-thermistor is -1.7), the average "miss" is
-9C, with the worst being -13.8C. Quite a bit more than the +- (And i've NEVER seen a mb that needed a + adjustment) 5C.
Chart comparing Socket-thermistor to DIE temps
And yes, this thread is ismply to ask "are my temps okay". HOwever, with your statements claiming that socket-thermistors are accurate is so misleading that I have to post against it. With a statement like yours, its no wonder why people with different "reading" mb's are always worried to crap about their temp readings.