Dave B(TotalVR)
Regular
Rambus is a serial memory interface right?
So I have a board with up to 4 chips on it. With the board running one chip, I have one XDR path to the 4 RAM chips, which are connected one after the other right? (like the RIMMS on old p4 mobos)
So when I add chip 2, is the XDR bus configurable enough to see the second chip is in, so give me a path to the second 2 memory chips. So now, with 2 chips I have chip 1 connected to ram 1+2 and chip 2 to ram 3+4.
When I put all 4 chips in, is it possible for the XDR interface to be configured again so that each chip has its own XDR path to its own RAM chip.
Now if you give each GPU chip 3 more XDR lines, so the chips also communicate with each other (and can hence get data across the right bus and other stuff)....
Surely thats a great way to produce a family of graphics boards with easily configurable rendering power brackets.
Surely that, in itself, is a great reson to jump on the XDR bandwagon as far as 3D graphics is concerned.
no?
So I have a board with up to 4 chips on it. With the board running one chip, I have one XDR path to the 4 RAM chips, which are connected one after the other right? (like the RIMMS on old p4 mobos)
So when I add chip 2, is the XDR bus configurable enough to see the second chip is in, so give me a path to the second 2 memory chips. So now, with 2 chips I have chip 1 connected to ram 1+2 and chip 2 to ram 3+4.
When I put all 4 chips in, is it possible for the XDR interface to be configured again so that each chip has its own XDR path to its own RAM chip.
Now if you give each GPU chip 3 more XDR lines, so the chips also communicate with each other (and can hence get data across the right bus and other stuff)....
Surely thats a great way to produce a family of graphics boards with easily configurable rendering power brackets.
Surely that, in itself, is a great reson to jump on the XDR bandwagon as far as 3D graphics is concerned.
no?