Single bigger dice can be more complex than making multiple smaller die. Also, yields of many smaller die tend to be higher than complex large die. Occam's razor really.
Printable View
Oh so this will be theyr quad-sli -.-
Well its just 2 x G92 GTS in 1 card, they should release G100 instead of this crap.
I hope it's more than just "2 x G92's" -- my problem with all the AA at super high resolutions, is that it's sometimes just a texture fillrate benchmark. Take a look at some of the benchmarks here (the Sapphire active cooling method is a good read, too):
http://www.elitebastards.com/cms/ind...1&limitstart=4
/rambling begin/
Nvidia is a fill-rate monster, but not necessarily the best at shader/ALU ops. While I'm digging some of the newer games out there (BioShock, CoD4, and Crysis), they seem to be more focused on AA+hires than more realistic physics, AI, gameplay, and visual effects -- although CoD4 is one of the best shooters I've played in a long time. I'm pretty content playing at 1680x1050 =)
We're in the 1st generation of DX10 games; game makers are still trying to figure out how to "do it fast" -- the game makers can't keep up with the evolving hardware =) we've had multi-core machines for years, and developers are still trying to "do it right" cpu side.
/rambling end/
Why only 30% performance increase? Possibly due to Ahmdal's law (google it), but I'm still not sure what benchmark they are using -- 30% faster in fill-rate, shader, or ALU ops?
What I want from the graphics cards is more performance with similar/lower power consumption, heat, and footprint -- is that too much to ask? =P
XFX offering I think.
http://www.tomshardware.com/2008/01/...force_9800gx2/