Comment 110 for bug 6270

Yesterday I collected an (apparently) useful information regarding the i855GM (rev.02) hardware (which I own):

after a linux session I rebooted into windows xp, activated the TV-OUT and used it to watch for many hours. Then I rebooted into linux and just before Xorg rewrote the video buffer, I could see the previous linux session's buffer (with some minimal noise added). This had happened previously when rebooting from linux to linux, and can be explained because there is some RAM persistence in the i855GM.

But if the persistence happens equally when using another video driver, e.g. Windows i855 driver, then it means that the linux intel driver uses *different* RAM than the Windows driver, maybe some PIPE A/B difference?

It's even more interesting to note that on Windows I used the normal LCD display and also the TV-OUT, while on Linux only the LCD display which obviously must have been feeded by RAM which was not touched by Windows for many hours.

Can somebody explain this differently?