I'm using a Xeon W3520, which is identical to a Core i7 920 except that it supports ECC RAM. I have it running on an ASUS P6T6 WS Revolution motherboard which also supports ECC RAM, and I have 12GB (six 2GB DIMM Modules) of ECC RAM installed.
Windows 7 RTM works perfectly fine on this system.
The one thing I can't tell you is how Windows 7 would react to an actual soft or hard ECC error, since I don't know any way to actually generate one. I expect that it would at the very least recognize the hardware-detected error and either log it (for a soft error) or bugcheck or cancel the affected application (for a hard error), depending on what area of memory the error occurred in.
Actually, I poked around a bit and found some information on WHEA - the Windows Hardware Error Architecture, including this MSDN page that describes changes in Windows 7, one of which is:
WHEA supports Predictive Failure Analysis (PFA) for Error Correction Code (ECC) memory. Through PFA, WHEA can monitor one or more ECC memory pages that have encountered previous errors. When too many errors have occurred, WHEA attempts to bring the memory page into an offline state.
It's true that, in terms of raw memory functionality, the OS doesn't care whether it's ECC or not. But what's the point of running memory that detects errors if there's no way to find out when you have an error?
An important part of the ECC puzzle is finding out WHEN you have a memory problem, and the OS is an integral part of that process. You can't just rely on the BIOS to report errors since they may only occur when the OS is up and running and the memory is being stressed in a particular way.
Running ECC memory with an OS that doesn't report soft or hard ECC errors is kind of like running a redundant RAID with no way to find out if one of your disks has failed.
Running an OS which doesn't disable memory pages that have soft ECC errors is kind of like running a RAID array that won't let you configure a "hot spare" drive.