Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Your first link is broken. It could be a mistake for all that I know though. I'd be more inclined to trust gaming benchmarks since that's something we'd see in the real world. Could the XBench/Cinebench results be factoring in more then video processing? I know XBench tests your whole system, without being open that link it's impossible for me to say if they're separating our the scores or just giving you one giant lump score.
 
Discrete refers to the 9600, integrated to the 9400.

EDIT: haha I didn't see the word 'performance'. I'm tired and need to go to BED! Cheers.
 
With all due respect to Arstechnica, we stopped publishing the OpenGL test results from Cinebench because it stressed the CPU as much as GPU (and ran at only 800x600). The Xbench OpenGL test was also of little value since it only utilizes one function in a small window.

In our testing at Bare Feats, we published results from 3D Games at 1440x900 and High Quality settings since they use a complex set of OpenGL functions and more video memory. That's why you see a greater gap between the 9400M and 9600M GT in our graphs.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.