It's a very similar concept to the TCP offloading engines you see in many major servers. I do believe they could get the 10% latency decrease by handling all UDP on the card instead of the main board, ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results