It took some thinking on this... but I do have to agree with Scott's point-of-view here. What this tactic is, is a double-edged sword.
The first "wave" of 'reviews' praises the GPU, and so the uninformed public will get a generally positive view of the APU design. Many people won't read follow up 'reviews' that add in all the missing benchmarks.
This is a double-edged sword, because if a site refuses to release such a highly selective "preview" then they will miss a very large portion of the initial readership that has been interested in the product. It seems like the majority of readers will read a preview elsewhere, and simply not read any further reviews. This will hurt any site that decides against publishing early, while also at the same time helping other sites that decide to fall in line and publish the limited story.
What AMD is doing here goes far beyond a typical "preview" release. In a typical "preview" article there are a few options: 1) a site can release information but not benchmarks of any kind, or 2) They get early access to the hardware and are free to run & publish any tests they wish on a preconfigured system. AMD has effectively found a third option... they can control what gets published, and in doing so present their product in the best possible light while completely hiding its shortcomings.
While what they have done isn't terrible, it is still bad and highly questionable. But more importantly, it sets a very dangerous precedent that WILL be taken advantage of by companies that are hard-pressed to push their products any other way.
Sure, a tech-enthusiast wouldn't be fooled by this. But are these sites publishing content for tech-enthusiasts, or Joe public?