Intel has recently made available for free (albeit to the Intel developer partner program members) a performance toolset for real-time graphics applications, called GPA (Graphics Performance Analyzers). The current version of this is 4.1. These analyzers operate by collecting data during the run-time of the application and then offering visualizations of this in terms of performance over not just the CPU but also the GPU.
This is a tool particularly useful for resource-heavy applications such as (most notably!) games and is well worth checking out. More information by Intel themselves here, including not just a technical specification/manual of the toolset but also how-to videos and examples plus case studies of current games and game engines using the software to great effect.