Originally Posted by GoldenBoy
Okay, now I'm really confused - I thought when overscan was turned off was when the display was doing 1:1 pixel mapping, not the other way around.
A 1080 source has 1920 x 1080 pixels in the image. The SXRD sets (like all 1080p rear projection TVs except wobulating DLP's ... which only have half as many pixels) have a chip (in the SXRD case it's an LCOS chip) with an array of 1920 x 1080 elements that correspond on a 1:1 basis with the pixels in your display. So the 1920 x 1080 image you provide via a 1080p signal is mapped 1:1 with the display elements.
However, not all of these pixels are visible ... the sets are intentionally designed to have a small amount of "overscan" ==> about 3% in the case of the SXRDs, which masks any irregularities in the picture at the edges. Note that this has always been the case for televisions ... in fact CRT's have notably more overscan (typically 5-8%). This overscan is built into the optics of the set ... so it cannot be eliminated electronically (like it could be with a CRT).
It is possible to electronically resize the picture to display more (or less) pixels ... but if you do this, you lose the 1:1 pixel mapping. The concept is the same as using a non-native resolution on any flat-panel monitor. If, for example, you adjusted a 1920 x 1080 display source to be completely visible, you would be mapping that display onto the 1840 x 1036 (to use bulb hater's settings) visible pixels ... so you would NOT have 1:1 mapping.
What the modern video drivers let you do is provide a 1920 x 1080 signal to the video card; but only allow Windows to use a somewhat smaller desktop (1840 x 1036 in bulb hater's case). The pixels are mapped perfectly 1:1 (since the electronic signal to the display is 1920 x 1080) ... but you're only using the ones that are visible (if you set the display correctly).
Some video drivers make it easier than others to set the right parameters ... ATI & nVidia both have special HDTV sliders to let you adjust the size of the desktop; Mac's do it automatically when it's connected to an HDTV; etc. But in any event it's best to simply use the visible pixels and NOT try to "correct" for the overscan.