To utilize GPU in Gaussian 09, install the GPU-enabled version, update NVIDIA drivers, and include %UseCuda=True in the input file. For multicore usage, set %NProcShared=24 (or more, based on your CPU) to utilize logical cores and %Mem=28GB to allocate memory. Ensure CUDA Toolkit compatibility and specify fast storage for temporary files using %Chk=/path/to/disk.chk. Verify GPU and CPU utilization in log files post-run. Optimize settings through benchmarking and allocate resources wisely to balance performance. Ensure both Gaussian 09 and your hardware support GPU acceleration for specific computations.