Playing Soldat in a network environment (server hosted in the office), the Killer NIC in gaming mode would actually cause lag to a certain degree. Explosives would hit but not explode and often cause random effects later on (explosive hits, but carries on, and then lands on the opponent again). This lag appeared to be packet loss rather than actual lag which would point towards the network as a whole being problematic (especially as both machines connected to the server experienced the same ping and packet loss). However, switching to App mode alleviated the problem slightly.
Playing Soldat online was fine with the Killer NIC in either mode, or with onboard with no noticeable change between the contenders. Running either the FNApp Bitorrent client, or uTorrent at maximum speed between the three options did show that the Killer NIC GameFirst feature does work. It provides better pings, but when you are running at ~400ms, the loss of ~40ms is unnoticeable.
Running Command and Conquer 3: Tiberium Wars in a network environment wouldn’t work unless the Killer NIC was running in App mode, no matter which version of the Killer NIC drivers were used. It’s similar to when I have to change the mode of my X-Fi for gaming, requiring you to alt+tab out and switch the Killer NIC to the correct mode.
When running Battlefield 2142 with another client beside me with similar setups, apart from my rig was sporting the Killer NIC, neither PC had considerable difference in ping. Considering as we were acting as a team, both going into intense battles where packet loss and latency are likely to increase, there was little gain to be had when using the Killer NIC. The Killer shaved off a maximum of 2-3ms from my ping, which didn’t massively help when my ping is usual 30ms anyway.
Currently, the Killer NIC seems like a good way to waste your money, but I was determined to get a decent result from the card. I moved on to hosting a dedicated Battlefield 2142 server with one of my home PC’s (Athlon 4000+ X2 AM2, 2GB 800DDR2, onboard network), then connecting to it with two other similar PC’s with their graphics set to minimum to remove any load on non-networking components. The Killer NIC equipped PC saw a decrease of 1-3ms ping maximum difference.