Multithreaded Rendering
It is a feature which allows DirectX to be processed via multiple CPU threads. This means that a dual-, triple- or quad-core CPU can have a higher utilization across all cores than DirectX APIs in the past. Historically the OS would load up a single core for commands to the GPU, in essence creating an overload on the first core and under utilizing the additional cores. With only one core issuing commands to a GPU, we have seen CPUs hold back the potential performance of the GPU. With Multithreaded Rendering, DirectX will take better advantage of all the available cores. This should result in a better experience for the multi-core user because of a faster processing pipeline and increased scaling.

DirectCompute
It is a feature which allows access to the shader cores/pipeline for Stream Computing (graphics acceleration) type applications and physics acceleration. One of the biggest technology breakthroughs of the past 5 years has been the notion that processing can be moved from the traditional CPU to the much more parallel GPU. Simply put, the CPU manages tasks sequentially; it accomplishes a task then moves on to the next task in a very orderly fashion and with tremendous speed. Today's CPUs can work at speeds of up to 108.8 GigaFLOPS (Floatingpoint Operations Per Second).

Ocean View: Real-Time Rendering with DirectX 11

A GPU is designed to work with many slower cores in parallel, giving a much wider vector -meaning a wider road for more cars to travel on - than a CPU. This allows tasks to be completed faster if the program or software is developed to take advantage of many-many cores, albeit slower ones.