As the article explains, G-CAT scores will be based on results in Doom 3, Half Life 2, and UT 2004, but raw framerates aren't the focus of Intel's new benchmark. Instead, Intel uses two separate models (Threshold and Bayesian) to predict and rate perceived performance.
The DH-CAT test will break down systems into three categories: Basic, HD, and Connected. "Basic" compliance means that a system will handle playing, recording, and transcoding standard TV or DVD content, "HD" compliance means that a system is capable of performing the same tasks on HDTV footage, and "Connected" level requires the system to be capable of DNLA-compliant streaming.
Even if you aren't particularly interested in a Intel-developed benchmark suite, it's safe to bet that G-CAT and DH-CAT results will begin popping up in various trade magazines and online publications. If the benchmarks gain much traction, we may even see them start popping up in PC makers' marketing materials. Obviously, there are a number of questions we could raise about the fairness and objectivity of such tests, but all bets are off until we see the final programs. Silent PC Review's article paints Intel's work in this area as neutral and user-centric. Hopefully G-CAT and DH-CAT will prove to be free of dirty tricks.