I just spent 2 days having the same problem. It is 100% an Intel software problem. The enabling/disabling of the driver in device manager works all the time. It may be a moody cable sense issue/voltage sense issue on 1 of the channels that the driver is not responding to correctly, but it is clearly fixable in software on the driver side. It doesn't happen in Linux at all with the same mainboard.
Mine has DVI and HDMI. The machine will start and run with only 1 of those outputs (either DVI or HDMI) when booted normally with loading the Intel driver at runtime, the other display on the other output works fine during POST, remains alive during the Windows welcome screen (W7x64), but then presumably when the Intel driver loads, one of the screens goes black, one of the displays claims no signal sensed on the monitor, so clearly a voltage sense problem in the software during handoff from Windows to the Intel Driver.
The cables I'm using have worked with several other computers (using the competitors embedded GPUs as well), and they worked fine in exactly the same configuration. I'd be perfectly willing to accept that Intel has a different voltage sense requirement, and therefore adheres to a specific cable spec. I'd be more than willing to spend $20 on some new cables if I knew they would work. But so far, its unfortunate that our friends at Intel aren't trying to figure out what the problem here is..... as it clearly is in the Intel HD Graphics Driver software package, unquestionably