[tacos]

For me the issue comes down to whether or not Microsoft does anything useful with this data (probably not, if 20 years of NVIDIA blue screen driver failure logs, Windows 8 and OneDrive are any example of how 'big data' impacts Microsoft product quality) versus how many comments I have to read where joeblow52 is personally offended that Microsoft dares to learn what his compile time plus 999,999 other compile times, divided by a million, equals.

[prewett]

How, exactly, are you thinking that Microsoft is going to fix nVidia's buggy drivers? They can collect all the data they want, but at the end of the day, it's nVidia's driver.

[tacos]

I worked there. Ways we solved these sorts of problems include: hardening the other side of the API/HAL when appropriate/possible, simplifying the driver model so that mere mortals could write drivers, writing our own drivers and overwriting known buggy ones for companies that couldn't get their shit together (usually network vendors), adding workarounds to the OS not to use certain features of certain cards, flying external engineers to lavish parties and our driver development labs and compatibility labs and providing one on one engineering development assistance from senior kernel developers, providing free testing of drivers for known problems before release, rolling fixed drivers into Windows updates, providing marketing funds as reward for fixing problems, and not using NVIDIA in the Xbox 360 after using them in the original Xbox as punishment because they were personally responsible for over 80% of blue screens in Windows for the preceding five years.

Sadly the motivation was often to ignore the data or watch it get spun by some jackass with the exact wrong agenda. It's just software, there's always a way to fix things if you really want to.

[sundvor]

Nice work, thanks for the insight!