Intel C++ supports OpenMP thread creation. The operating system should always schedule a thread on a logical processor which is not busy, although some are better than others. I don't know what you mean by Bi-Xeon. The Pentium D EE has 2 cores, each with 2 logical HyperThreaded processors. If you are looking at the total capacitiy of the system, depending on the requirements of your program, you are likely to get 90% or more of the capability by running 1 thread on each core. Sorry if the Windows performance meter display is disappointing.
If your work is evenly balanced between 2 threads, and their performance isn't limited by some factor external to the CPU, such as disk access, you should be getting close to 100% usage of one logical process on each core. With HT enabled, Windows would say you were getting only 50% overall. In reality, it's much better than that.
Current released operating systems may not do a good job of spreading work evenly between the 2 cores, unless you disable HyperThreading. Operating systems support thread affinity calls, but those have to be programmed for specific types of systems, and run into trouble when multiple applications start up. Those aren't covered in compiler documentation.
