Testing lag is something that really can’t be done. Bigfoot have released an application that does measure lag to a certain degree but a program made by a company that proves that their product is the best is something to be wary of. In fact, it seems more of an advert to their product than a analysis tool, as your results are duplicated with a couple of ms removed and put on the ‘if you had a Killer NIC’ graph if you don’t have one installed. This flips around if you have the Killer plugged in. It would be like if 3Dmark released a graphics card that got high marks on their benchmarks.
Even if there is no foul play with the program, it doesn’t give you raw statistics, and instead uses a graphical interface that doesn’t allow for easy comparison. Plus, it’s difficult to get exactly the same scenario for when you are using the Killer and when you are using an onboard solution as server load, network load etc. can all change drastically. I also found that it crashes easily which doesn’t increase my faith in the program.
Instead, I’ll have to rely on a blind test, which involves a second person either choosing to install the Killer or use onboard and then the tester decided whether or not there is a difference.
The first test that we want to show is the difference between the Killer NIC and an onboard nForce 630a solution. I’ll be transferring the USB test directory over our office network which contains a whole bunch of files – big and small – which amounts to 900mb. Timing how long this takes should allow us to give a real world mark on the throughput. The nVidia onboard will be running in ‘throughput’ and ‘CPU’ mode.
All of the tests are based on a 100mbit/s network, even though the Killer NIC can deal with gigabit networking speeds.
As you can see, the Killer NIC in either mode can’t contend with the onboard solution although a couple of seconds isn’t a massive difference and pretty much undetectable in normal activities. If you were constantly transferring very large files (>5GB) then these few seconds will mount up. While this test shows throughput, the Killer isn’t designed to be the best at this, and makes a sacrifice in this department to kick ass at latency.
The blind test was repeated constantly throughout an entire day
f Battlefield 2142 gaming. This time it was against an Asus P5N-DH Deluxe onboard networking solution which didn’t have the same nVidia optimisations as the 630a. The blind test consisted of playing for 30 minutes, then walking out the room and the Killer NIC either being left in (Game mode), the onboard used instead, or the Killer NIC put into App mode. The testing then continued and every time I wrote down my personal view of the lag. As I had no idea which network adaptor was being used, any placebo effect was removed.
Honestly, I couldn’t tell the difference between any of the options, which doesn’t bode well for the optimisations that the Killer NIC brings to the table. In either case, I didn’t notice any difference in frame rate which the Killer is meant to enhance. As most games fail to make the most of your components, or you are limited by your graphics card, releasing extra CPU cycles isn’t going to help. Bring in multi-core processors, and the Killer NIC doesn’t seem like the greatest idea going.
As the Killer NIC only eliminates client side lag (i.e. the time taken for the inputted signal to be transferred to the application) the gains should always be the same. For example, if it normally takes 4 ms to go from cable to game with an onboard NIC, and the Killer NIC removes this, you’ll always save 4 ms. This was not the case in my testing.