The benchmarking suite is only in beta right now, which is why no hard numbers have been released. He also points out in this article, or a related article, that he was hired to write a benchmark suite that would make the P4 look good. When he posted some articles on his website pointing out some interesting performance observations, he was canned. Yes, this may lead to bias, but I think we're all aware that the current benchmarking suites are heavily Intel funded, which makes them biased as well.
While some people may not do much multitasking, myself and most of my co-workers do. I always have several browsers open, a DOS prompt, my email client, pcAnywhere logged in to several machines, 2 FoxPro sessions, one for programming one for processing, ICQ, mIRC, and WinAmp. At home I'll probably have browsers, email, mIRC, ICQ, Money, Word, and a game of Minesweeper running. You SHOULD get very different results when benchmarking performance under these circumstances than loading these applications up one at a time. This isn't meant to replace current benchmarks, it should simply add another way to analyse performance, and give us end users another tool in our decision making suite.