14 DAY TRIAL // Details about the release and specifications of the new G80 GPU still remain shrouded in mystery. The Inquirer states that the launch date has been postponed to early November, but if the company does not get the latest silicon for its new chips in time, early November is likely to become late November. Other sources claim that there will be short stock problems and G80 won't be available up until the end of the holiday season.
The new core would have a 400 mm die-size and, apparently, G80 is rumored to be a dual core GPU. However, G80 was designed to be a non-converged shader unit, unlike ATI's R600.
How about DirectX support? The Inquirer provides some interesting facts regarding G80's specs. It looks like Nvidia isn't focusing too much on the new DX10 technologies but G80 will surely do as much as two 7900s at DX9. ATI's R600 seems to do better at DX10 as G80 has only 1/3 of its die- size dedicated to DX10 specs.
Although it is not known whether G80 supports GDDR3 or GDDR4 memory, early prototype cards were using 12 memory chips on them, a quantity that would hardly be tolerated by desktop computers. Just imagine 384 or 768 MB of GDDR on a video card. Anyway, G80's pipelines clearly need more memory bandwidth and this seems to be consistent with the extra memory chips. In this case, the bus would be as wide as 384 b. Nvidia is yet to prove the cost efficiency of these configurations.
Will G80 powered video cards be energy efficient? Bit-Tech thinks it won't, as these cards will require their own power supply. And this will happen with ATI's R600, too. Better wait for R600's release before spending some $1000 on a SLI G80 configuration.