First, I would be a VERY Suprised if Matrox would name their next gen as G1000. There is pretty heavy indications on the air that at least cards will be named after the codename. (which IMO doesn't sound bad... Matrox Parhelia is something that fits nicely to series of Impression, Millenium, Mystique and Marvel.
)
second, there is a general mix up with TBR, deferred rendering and tiled frame buffer. As someone here earlier stated, even G400 handled frame buffer in tiles instead of lines. But G400 wasn't deffered renderer. afaik, Glaze3D and Avalanche3D from Bitboys are doing it same way, but still they aren't deferred renderers.
There is a lot of rumours flying around, but most of them are BS.
to make some things clear, I am
expecting equal theorical pixel rate as GF4Ti4600 has and doubled theorical texel rate if both chips are running at the same speed. of course, practical speed is another case. Matrox has superior theorical memory bandwidth against nVidia counter part, but one thing that is question mark is the efficiency of their memory controller.
(though I believe that they have managed to make it pretty powerful. G400 had the best memory interface of it's time and it made most of G400 series cards most less memory bandwidth limited ones.)
Another thing is, their FSAA. which sounds pretty incredible in the light of comments and first "unnofficial" tests. So if you think that GF4Ti -series has fancy AA, what comes to quality and speed, just wait for the tuesday.
And as always, I can't be 100% sure if my information is correct so, maybe is better take this as a grain of salt. Anyways, on tuesday we should be a lot wiser.
