IR will read shiny surfaces artificially low because of reflectivity. Also you are measuring the heatsink which is lower temp than tcase which is lower temp than tjunction.
If you put a piece of masking tape over the heatsink where aiming you can get accurate temp of heatsink...but on load there will be big gap.
But if you want to measure tcase, remove heatsink, undervolt and underclock (so does not overheat prior to reading), put thin piece masking tape on cpu (otherwise you will get artificially low temps) and put IR gun on top of cpu. That way you can calibrate your cpu temp to actual tcase temp if you desire (speedfan will give you cpu temp) and calibrate core (tjunction) temp ~5C higher than that.
To check accuracy of your sensors with IR:
In attached pic with tjmax of 95, tcase (IR) gives same temp as tjunction (software). Though we now know ~5C gradient exists at idle with no heatsink, underclocked state of measurement. So with correct tjmax 100, tcase would read 72 when tjunction read 77.
(reference articles for gradients are in realtemp thread)
Bookmarks