I wrote about such ideas over two years ago. it seems the concept of offloading packet forwarding from CPU to GPU may have some merits, and if you’re interested in that - take a look at packetshader.
still however, hardware config needed to achieve that kind of performance feat is quite expensive. it demonstrates however that GPU, next to FPGA experiments, can also be viable way of forwarding in high-performance packet forwarding/routing based on PCs.
the problem is with services - things like ACL and QoS or MPLS are either not supported, or slow things down considerably. the other problem - is lack of support in the long term for running GPU under constant 100% load. balancing load vs performance may be however good choice.