


For codecs that are built specifically with the GPU in mind (NotchLC, HAP, etc) it can actually make a big difference. Certain media servers and hardware are geared to take advantage of a GPU’s speed and efficiency. As well as reduce usage of RAM, PCI-E bandwidth, and benefit from a parallel workload between CPU & GPU.
CPUs are optimized for serial tasks. GPUs are optimized for parallel tasks. At any given moment a media server needs to perform several tasks in parallel.
With larger resolutions and multiple layers of media stacked on top of each other, composited live in the server, efficiency becomes more and more important.