We started with our file transfer tests, which involved transferring large folders of files over a private network between two identical computers. When transferring a folder full of large files between the two computers the Killer 2100 did make a difference, if only a slight one. With the Killer 2100 involved in the transfer, whether sending or receiving the data, we saw transfer speeds of 11.1MB/sec. This result is slightly above the result seen when transferring data between the on-board controllers, but only by 0.3MB/sec.
The Killer 2100 had a problem when it came to organizing small files however, as we saw transfer speeds of only 5.69MB/sec when the Killer 2100-equipped PC was sending this type of file over the network. This was well behind the speeds we saw for the other two transfers and indicates that the Killer 2100 may struggle when bombarded by thousands of tiny files. It’s all very well removing the networking processing from the CPU but this isn’t a benefit if the replacement chip can’t cope with it.
Things were a lot less clear cut when it came to our games testing. Our ping remained constant while playing Team Fortress 2 regardless of whether we had the Killer 2100 installed or not. We also saw our frame rates drop with the Killer 2100 installed, a surprising result given that the card claims to keep your frame rates consistently high when gaming. The Killer 2100 had more of an effect in Call of Duty 4 though, with our ping dropping by around 3ms when using the Killer 2100. Frame rates also saw a small boost with the Killer 2100 installed.
Click to elnarge
Our time with CS:S did little to tip opinion either in favour or against the Killer 2100. We saw a tiny improvement in ping with the Killer 2100 equipped but our framerates remained constant. Our GRID testing was limited to measuring framerates as we couldn’t get the game to show us our ping. Unfortunately for Bigfoot our framerates remained the same when we added the Killer 2100 to our test rig.
But can you game while downloading?
We were intrigued to see how well the software network manager worked, so we fired up a torrent client to see if the Killer 2100 could still deliver acceptable frame rates and ping while other programs are demanding bandwidth. We did our testing in CoD4 while our client was downloading three (completely legal) 1GB torrents. Each torrent had around 100 connections and we allowed them to get up to speed before starting our gaming.
Needless to say, gaming was rendered impossible with the on-board network controller while our torrent client was running. Most servers we could see had a ping of 999 and would kick us immediately after we joined. This was acutely embarrassing.
We slotted the Killer 2100 back into our test PC to see if its whizzery could do anything to help us game while downloading. To test this we set the priority of the torrent client to low and the priority of CoD 4 to high in the Killer 2100 network manager software. Once in the server we were pleasantly surprised to see a ping of between 50 and 60ms, not too shabby given we were downloading in the background.
The bundled software is actually very good. Click to enlarge
Our happiness was short lived though: as soon as we moved around in the game we knew something was wrong. Even though our ping was reported by the game as playable, the experience was very laggy. For example, we’d walk forward through a door only to have the screen stutter before putting us back outside the doorway we walked through.
This kind of thing should be familiar to anyone who’s attempted to play on a borked server, and it obviously made the game unplayable. It’s intriguing that the Killer 2100 was doing enough work for the server to report our ping as 60ms, but it’s obvious that something was still stopping the game being playable. Whether our ping was micro-stuttering or we were losing too many data packets, we couldn't tell.
We got better results when we dropped out of the game and reduced the bandwidth for the torrent clients to a total of about a quarter of what was available. This allowed us to play CoD 4 as normal, with on a slightly higher ping than we had when not torrenting. This is a good result, and means you can happily download a new steam purchase while playing an old game to while away the time.
It says a lot that it was so difficult for us to see any tangible difference from the Killer 2100 despite our best efforts to test in a scientific and consistent way. Even in our theoretical tests where we could control the variables of our local, private network we saw little benefit from the Killer 2100.
When gaming we saw little difference between our (free) on-board network controller and the Killer 2100 in terms of ping and frame rate. The only advantage we found was that the Killer 2100 allowed us to download in the background while gaming - possibly handy if you buy loads of games via download services but haven't the patience to wait for those downloads. This makes it very difficult to recommend the Killer 2100 as, at the end of the day, everyone wants some kind of tangible benefit for their £68.
For the gamer who has everything, the Killer 2100 could be a consideration, as we did see very slight ping improvements in CoD 4 and CSS, after all. Having said that, you might see better improvements in online gaming performance by spending that £70 ditching your ISP and getting a more gamer-friendly one (pay close attention to whether, and how much, gaming traffic is throttled). Alternatively, invest in a new mouse or keyboard, or put the money toward a larger monitor.