Synthetic testing is a great way to measure cold hard facts, but its important we translate this into a real world environment so we can see any possible benefits from switching to a Killer 2100 Network card.
First we created a 2.7 GB folder of various files. Jpegs, database files and system files. Then we copied it across from machine to machine with a combination of the Intel and Killer 2100 Network cards in both machines.
Onboard to onboard gives us a solid reference figure of around 25 seconds. When we change one over to a single Killer 2100 we can shave a second of the time, and finally with two Killer 2100's sending and receiving we reduce the time to just over 23 seconds. A benefit of two seconds total.
When then used the Synology DS710+ NAS system we have tested before and put a Buffalo 128GB SSD in it. With the main 875k system using an Intel SSD in Raid 0 and the 980X system with a Kingston 128GB SSD as the main drive, we are in a perfect situation to check line saturation without being limited by the hard drives.
Our results are basically saturating the 1Gbps 16 port control switch we use as our network controller. A stunning set of results.
Testing online gaming is a very difficult thing – we have already shown that with Resident Evil and online activity that the situation is improved dramatically.
We fired up Call Of Duty, Bad Company 2 and several Valve games and fragged away for a full day. We noticed that once we got used to the Killer 2100 we never noticed lags or stuttering, but when we switched back to the onboard Intel solution that sometimes we would get drop out which was noticeable. This is obviously dependent on the server we connect to, but with the Killer 2100 it seemed to deal with worse conditions much better. It is something you adapt to, and you never know what you are missing, but with both NIC's connected to our gigabit switch, we would play for 30 minutes, drop out, switch over and try with the other. It really started to become noticeable, especially when I monitored CPU usage, even on such high powered machines.
CPU performance would drop by 2-5% when using the Killer 2100. This might not seem like much, but bear in mind this is with a 4.8 ghz 875k processor. The gains with a lesser processor would be more noticeable.