You guys do realize that all of these CPU temp progs read the same data from the same hardware sensors, right? Therefore it is pointless to say "HW monitor is better than speedfan" and "coretemp is better than HW monitor." As long as the progs are reading the same hardware sensors, they will give identical results (assuming the programmer didn't implement some sort of temp "correction" in his software). It isn't like these progs pull a temperature out of thin air and hope it's right. They only report what the hardware gives.
I have noticed that some of the Windows progs will pick up only some of my hardware sensors, while others will pick up different ones. I think it really depends on which hardware sensors the programmer has implemented into his software. There are tons of hardware sensors out there, which is why it's hard to find one prog that will have support for them all. Lm_sensors (Linux) is the best I have seen at reporting ALL of the sensors on my particular mobo. The website for lm_sensors has a huge list of supported sensors. I doubt Speedfan or Core Temp has this many implemented into the code.
One thing I have figured out is that my BIOS reports my CPU temp based on my "it87 hwmon0" thermal probe. Once I figured which sensor my BIOS used, I went into Linux (and Windows) and looked at what the temp for that specific sensor was. I noticed that my temps as reported in both Linux and Windows were ALWAYS 5 degrees lower than my BIOS, which suggests to me that Gigabyte did their own "temperature correction" -- that is, they made the BIOS report temps 5 degrees higher at all times for safety. They do this because they know that hardware sensors are not always 100% accurate (this is just the way it is due to design limitations). To remedy this discrepancy, I simply opened lm-sensors in Linux and set it so that my CPU temp always reads +5 degrees -- this makes it identical to my BIOS at all times. I think Speedfan in Windows gives you the same option for temp "correction." The problem is that you don't really know which sensor is which in many of these Windows progs, therefore it's hard to know which one to "calibrate" to your BIOS reading. I have no idea if the 5 degree addition the BIOS gives me is right, or if the lower readings straight from the hardware is. Either way, it's safer to use the higher reading, which I do.
One more thing: I am not convinced that while you are in BIOS that you are under "80% load." My BIOS temps (while I'm in my BIOS) are always the same
as my idle
temps in Windows or Linux. Therefore, I am pretty sure my BIOS when open is not at 80% load. Maybe it's true on some mobos, but it's never been true on any I've ever owned. I surmise this got started because people noticed their BIOS temps were always higher. In all likelihood this is not because the BIOS is under load, but because the manufacturer of the board "programmed" the BIOS to report higher temps than what the sensors actually report. Think about it -- does a simple blue screen and text really need 80% CPU power?
And, oh yeah, the guy who said the AC64 cooler is only good for stock speeds is wrong. Many people report massive overclocks using that cooler. I've never used the AC64 or AC7, but I have a cooler that's very similar in design to the AC and my CPU temp at a 900 Mhz overclock never goes above 45 degrees, even on an all-night orthos test. Granted, both my CPU and heatsink are lapped, but I have noticed no difference in temps after lapping. All that work for no temp decrease.
But to the OP: your heatsink is probably not mounted properly. You either didn't use AS5 right or you are getting too much air in between the IHS and the heatsink. Also, you need some case fans.
Basically, I echo the advice everyone in this thread has already given.