let me see if I can clarify some of my replys as they might have been a little unclear....
DeepScience said:
Through the interface, not on board the card. Which implies that if you used two cards with four processors the SLI system would, apparently, have to portion out a quarter of the screen to each processor.
basicly, I agree with you.
yet as it stands right now, SLI isn't always doing a 50% portion for each GPU.
there is also other options then just split frame SLI.
Alternate Frame Rendering (AFR), is also a posability and a reality for many games.
and although everything we say on quad SLI is speculation...
I'm sure nVidia or gigabyte would try to use a driver that alows 4 GPUs to render 4 full frames in AFR....if the game allows this.
JeffP said:
I don't believe this will be the case. Each processor will share the 256MB memory. This makes the work much more efficient then each having to load the same information into the memory it uses to split the workload. This is what Intel has been strugiling with in their dual core setups but IBM with their Risc series and Sun Microsystems have been doing it for years in their multi-core setups.
the major problem with your theory is that nVidia didn't paper launch this product....
gigabyte did.
since the card is basicly a SLI solution in a single card...
and all info points that way...
I don't see gigabyte utilising a technoligy that not even nVidia provided.
the research needed and the time nessasary for such a working idea would mean that this card would take forever to come to market...
if gigabyte did it on it's own.
in other words, everything looks like each GPU will utilise it's own set of memory.
this is how SLI works for two seporate cards, and I don't see this changing for this "all in one" card.
yet we all could be wrong. lol
mica