After getting
ZERO response from either Anandtech or Anand HImself after their heatsink review, I have almost stopped caring.
My sig wont' fit another crap review. Too many characters.
This review clearly has even more problems than the Anandtech review. "they claim" that they are getting very accurate measurements with an under socket thermistor. Yes, accurate for where it is measuring temps, which is both a secondary heat pathway, and in this case, doesn't touch the backside of the chip at all.
Comparing to a laser, you're still measuring side of core temp. Which is likely lower than cpu core temp. The problem is, like longhorn says, they don't say how they compared it to a laser. With the handheld laser units, they are up to 2% off, and with the mb's the way they are, it is almost impossible to mount the gun in a way to measure the small gap between chip PCB and heatsink properly. A slightly off angle and you're measuring CPU PCB or Heatsink temp. This is almost an exactl replay of how Tillman claimed that "a thermistor touching the side gives the same results as the socket thermistor". A webmaster/reviewer telling you to trust something that is probably contrived.
Heatsink Testing Methodology. As measured in this series of tests, measuring from the side clearly isn't the same as measuring from the socket(and socket position matters largely, too, as explained later in this post).
If you go ot overclockers.com they compared temps between a kt7 thermistor and a thermocouple directly in the middle behind the core. The temp difference was anywhere from 8 to 10C too low for the kt7. Imagine going out another quarter inch
away from the core and you'll get even lower readings. for reference,
The kt7 thermistor touches the backside of the core edge
Also, i'm sure those resistors in the socket give off heat, which throws off the readings even more. I am not going to bother e-mailing tomshardware because its obvious that his reader base is probably even more "far-reaching" than anandtech's. I will also probably stop posting this information altogether, as thanks to this and other reviews I am awaiting an influx of people who will argue that the thermistor-in-socket is accurate.
Not to mention, this is the first review i've seen where the hedgehog does not end up near the top of the pile. Even in hte worst Hedgehog reviews, when done on platforms with good comparison hardware, it ends up tying the alpha's for top notch performance.
P.S. THeir PEP66 is an older-variant, probably without a socket A clip. That would decrease its performance quite significantly.
I think it is time for AMD to implement an internal diode. From what I understand of how it works, it should not be too hard for them to do. With high-heat level's(up to 70W for 1.2ghz t-birds), it is very important to get an accurate reading, even if that reading is uncorrected. It is much more useful for comparisons to be done on a diode, and you get none of the temperature compression you often time see on socket-thermistor solution.
Mike
Facebook
Twitter