They aren’t far off form majority of Benchmarks. And knowing Intel’s knowledge and access to Cinebench I’d wager that wasn’t an accurate presentation with a version of Cinebench that is a little beyond Cinebench 10 and not really comparable. They were probably running a Cinebench that was using the R11 rendering core instead of the R10 one which would throw off the numbers, or perhaps they tweaked the benchmark file. just by using R11 and turning off AA I can get my Q6600 to render the scene in 27 seconds instead of the almost 2 minutes it usually takes, that really changes the benchmark. I know when We did the benchmarks using the 24 core we had to use a modified version of cinebench that used the R11 core simply because the R10 core was limited to 16 processors, so it wouldn’t represent 24 very well.
Here is an example showing how several people with older CPu’s got comparable scores by using the R11 renderengine with cinebench. The problem there is the bench marking standards are based of an certain engine and when that engine is dramatically improved the original benchmark stats are no longer accurate so values will always be extreme. Gives you a good idea how much better R11 is compared to R10 for rendering though.
to date I think the best cinebench I’ve seen from a third party source for an I7 was in the mid 20k range, 24-26k on an overclocked system, but then thats for a single physical cpu, while the macs are dual. I wish apple would give the actual benchmark instead of saying 1.3x faster than before.