The NVIDIA CUDA 4.0 Toolkit is designed to focus on parallel programming, and enable developers to port their applications to GPUs and x86 CPUs. The three new main features of the new CUDA includes. GPUDirect 2.0 (for peer-to-peer communication among GPUs within a single server or workstation.) Unified Virtual Addressing and Thrust C++ Template Performance Primitives Libraries. You can read more about CUDA 4.0 here.