Update for NVIDIA only : I refined a bit CVT Timings (again), for the second table they almost fit perfectly with the original table, for the first one, values are closer. It takes in count the frequency of the mode we overwrite, so the pixel clock should be the same (as there is no pixel clock value in the second table, it's either calculated or fixed within the vbios.)
The first modes of the second table aren't patched anymore (the 320x200 ones.)
I'm wondering if, despite the patch, the actual mode it sets is the original one, and not the resized one. From CEOS' photos, the first mode of second table is patched with native mode (320x200 -> 1920x1080), and the mode it is about to set is this one (#271, ie 0x10F is standard 320x200x32@60Hz). CEOS reported that the set mode was 200x200 like (probably 320x200 actually.)
The one time it (almost) worked, the patched mode was 1920x1200, so it seemed good! Well... with a 120 pixels garbled band at the bottom, because buffer's height was 1080.
Eberts had a band too because it patched the mode from 1280x1024 to 1280x800. The buffer was 1280x800, but the actual mode was still 1280x1024!
I'm not sure I'm explaining it right...
So CEOS, I hope that the new one is working better, and I'm waiting for at least as much photos as the last time to posted. having so much info at a time really made my day. (I don't need much...)
@Azimutz, I don't think I can do much more about GMA, I went on Andy's Thread on InsanelyMac about is patch for 945GMA (wich includes 915resolution from prasys and gma graphicsEnabler from nawcom.) It apparently does not work for everyone. So I guess I'm stuck at the same point as he is.