It says they didn't actually test with SLI/Crossfire though so it's a totally unfounded claim until it's tested. That might sound like I'm just nitpicking, and I'm not, but I'd like to see some real tests.
Also, we're talking about City Of Heroes. I've played that quite a lot and it has some serious issues when it comes to GPU scaling and it barely supports Radeon hardware too.
The problem is most reviews will not include these tests.
The claim though is not totally unfounded when you consider what being CPU limited means.
Take the following scenario.
The CPU sends data to the GPU to render. It then does the rest of it's work. i.e AI, physics, loading various data, handling user input, sound etc.
The cpu takes 6/100ths of a second to do the work. Whilst the CPU is doing this the GPU is rendering the scene with the data it has been sent by the cpu. Lets say it takes 2/100ths of a second to do this.
This would leave the GPU idle for 4/100ths until the CPU sends the data for the next frame.
Now lets say you add another GPU. The 2 GPU's can now render the scene in 1/100th of a second. The CPU will still take 6/100ths of a second.They are now both idle for 5/100ths of a second.
In a simple scenario like above you will not have gained by adding a second gpu.
In real games that are cpu limited there may be areas where the gpu takes longer to do its work than the cpu and becomes gpu limited for that part. In that case you main gain slightly by having a second gpu but only for the odd few frames. If you are heavily cpu limited the addition of a ppu would reduce the load on the cpu enabling it to complete its work faster and send data to the gpu faster and increase fps.