DarkDust - Boost supports thread level parallelism, whereas PPL and TBB provide a task based abstraction on top of a thread pool and take the standard library's algorithms and containers approach. This removes a lot of the headaches associated with scheduling and resource management. Some task based parallelism features are now starting to be supported in C++11 too, notably std::future and other async related features.
Ghita - The PPL does not use managed extensions, it is purely native code but it is not all template based. The Windows implementation runs on top the Concurrency Runtime, ConcRT. See http://msdn.microsoft.com/en-us/library/gg663535.aspx for more details. Intel's implementation provides a source code compatible API for a subset of the PPL. See http://threadingbuildingblocks.org/docs/help/reference/appendices/ppl_compatibility.htm for a discussion of TBB/PPL compatability.
Alexey - The Intel version on Unix runs on top of their runtime. On Windows the TBB 3.0 can also run on top of ConcRT, depending on how you link it. See here, http://software.intel.com/en-us/blogs/2010/05/04/tbb-30-new-today-version-of-intel-threading-building-blocks
The book is available on Amazon etc and the content is also available on MSDN
http://msdn.microsoft.com/en-us/library/gg675934.aspx