Well I think it's pretty well assumed NV34 doesn't do vertex shaders in hardware. Just for kicks, does anyone have 3DMark03 scores for a 9500 NP with software vertex shading forced? That might at least give us an overall ballpark where actual NV34 figures would play.
Then of course we adjust for 275/275 -> 250/200 (the latter being the reported clocking for the "Chaintech FX40" product). Plus the possibility that NV34 is missing more functionality (compared to NV30) than just vertex shaders (of course 9500 NP is missing HierZ too, isn't it?). Plus perhaps a fudge factor for unspecified inefficiencies of the NV3x architecture vs. R300 (to the degree an NV30 @ 275/275 would score worse on 3DMark03 than 9500 Pro, which isn't a sure thing but is probably likely). And, finally, for beta drivers.
Bottom line: I wouldn't dismiss these scores out of hand. When the Inquirer gets something right, it's often stuff like this (i.e. where there is little room for misinterpretation on the part of the reporter). It's likely that, by this point, many people at Nvidia's board-level partners and at PC OEMs already have access to this information, and the Inq certainly has sources among that group of people.
