Following on from the review I did on the BigFoot Networks Killer 2100, BigFoot Networks contacted me to ask if I’d be interested in a revisit with a new card and their support. I was more than happy to do so as was disappointed to not get the card working. This would rule out both the previous card being faulty and/or anything I may have done in error. So, with this, BigFoot Networks sent me out a brand new retail Killer 2100 as well as some documentation on the 2100 and how they recommend it to be reviewed.

BigFoot Networks Killer 2100 Network Card

BigFoot Networks Killer 2100 Network Card

I mentioned in my first review that due to Avast’s real-time shields, the QoS feature in the Killer Manager is negated. BigFoot Networks called this an unusual setup and recommended I disabled or removed Avast to “better reflect” the configuration preferred by you, our readers. I personally believe this to be a fairly typical setup as there are over 130 million registered users of Avast. During testing I will, however, disable the real-time shields and see if it makes any difference. The discussion also touched upon what BigFoot Networks’ aim is to do with the Killer 2100 and, indeed, all of their products. This is to recognise and prioritise game traffic, get the packet unwrapped and delivered to the game (bypassing the Windows network stack) within “one Windows time-slice”. BigFoot Networks then continued to say that the default way of doing it requires “multiple Windows time slices” to do the same task. I asked BigFoot Networks if they had a whitepaper or something with some more information as to this but, unfortunately, they do not. They did state that this would easily be shown by the results of the GaNE test and, one would assume, also in the real world.

Upon receiving it, I installed the card and the latest version of the software. After the mandatory reboot, I was told by both Windows’ Network & Sharing Centre and the Killer Network Manager that no cable was plugged in. My Gigabit Netgear LAN switch disagreed as reported a connection of 1 Gbps. After rebooting again, both became happy that a cable was plugged in and the Killer’s software popped up to tell me it wanted to update the card’s firmware. I said it could and it started erasing the flash, followed by writing to it. With no confirmation to let me know the firmware flash was successful I risked a reboot. The flash had been successful as I could browse the internet but taking a look at the Killer Network Manager it told me the card was disconnected – however, Windows disagreed and was happily using it. Again, this required a reboot to fix where everything became content it was working.

The card worked for a while and I played some games with the plan to begin testing the following day. However, later in the afternoon I lost nearly all connectivity. All web-pages stopped loading, MSN and Steam disconnected but Skype remained connected. I knew my internet connection was fine as both my phone (via Wi-Fi) and another computer had no problems connecting to the web. A reboot seemed to fix the issue. Unfortunately, this seemed to have only been a temporary fix, as the problem soon returned, this time being more stubborn with a reboot not solving it. Pinging my NAS (on the other side of my Gigabit switch) and pinging the router (across the switch, then via powerline adapters) both showed up a 25% chance of timeouts. Disabling my anti-virus also did nothing to help so the only option left to me was to switch back to onboard and email BigFoot Networks.

Within a couple of days I had a reply from BigFoot Networks’ Technical Support Manager asking me to run their diagnostic utility and send off the resulting text file. After running the utility I replied with the said text document and some screenshots further highlighting the issue. Shortly after, a reply came through asking for my network topology to see if any hardware could be eliminated due to a possible conflict. Without moving my whole setup, the first thing and easiest thing to do was bypass my switch. This meant I was connected to the router via only the powerline adapters. Interestingly, this caused web-pages to start working (though, I must admit they did seem slower to load) as well as pinging the router/google.co.uk/bbc.co.uk to no longer have errors. Trying Youtube and speedtest.net showed me that flash content wasn’t working but it was an improvement. To double check if this was my switch, I went back to using it and noticed exactly the same as the above. Having done some playing around I decided to see what my ping was in a Team Fortress 2 server I’ve been racking up a fair few hours on in the past week or two. Throughout the half an hour, or so, I was playing I kept taking note of the ping and it averaged out to the 30ms mark, which is a ~50% increase when compared to onboard (19ms). This caused me to stop playing and continue trying to narrow down the cause of the problem.

BFN Killer 2100 with shroud removed

BFN Killer 2100 with shroud removed

Grabbing another switch I have (this time a 100Mb one) I, again, tried some speedtests and loading web-pages. With the help of a Firefox add-on to display the page’s load time, I was able to confirm what I had noticed earlier and that was that pages took a couple of seconds longer to load than via onboard. This was tested using several fairly large web-pages making sure to bypass the browser’s cache. When switching back and forth between the three setups (no switch, Gigabit switch and 100Mb switch) I saw the Killer not working with the Gigabit switch at all. Having tested with various web-pages and noting no difference between no switch and the 100Mb switch (while with the Gigabit switch just did not work), I went on to trying out some games. Call of Duty 4 was my first game of choice and I chose to first connect to a server I’ve been playing quite a bit on. However, this failed to connect and so did other servers from the master server list. This remained true for all of the setups, apart from onboard which connected straight away regardless of the switch used, if any. Having a look at the Killer Manager showed an ICMP ping but not one for UDP traffic. In an attempt to bypass the powerline adapters, I chose to bridge the Killer with my laptop’s wireless connection. This made no difference as to UDP traffic and with myself completely stumped I emailed the above off to BigFoot Networks in the hope they could shed some light on the issue.

A reply was waiting for me in the morning with a request to know the specs of my computer (any overclocking etc.) and that BigFoot Networks would try to get back to me with more information. Before they could get back to me, I experienced a BSOD with the stop code 0xD1. This, upon analysing the mini dump file, pointed to the driver “Xeno7x64.sys”.  While I wasn’t using the Killer 2100 at the time, due to the previous outlined problems, this seemed to be caused be the Killer’s driver. When I informed BigFoot Networks of this, their response was that my power supply was not adequate or the Killer NIC was faulty. Unfortunately, I did not have another power supply to test with and neither did I believe this was the issue, as in nearly two weeks I had had no other problems which could point to my power supply not being enough for the system. To reduce the power draw from my components I removed the overclocks and tried again, but this proved to make no difference. However, through further testing of setups, the Killer 2100 managed to shake its previous problem of UDP traffic not working. Despite the Killer Manager still reporting no ping for UDP traffic, it seemed to be working and I could get some testing done.

When I first received the card, BigFoot Networks sent me some documentation as to how they recommend the card is tested. This involves two computers (one as a “listen” PC and the other, the “test” PC, with the Killer 2100 installed), a Gigabit switch and the GaNE (“Game Network Efficiency”) test program. Both PCs are connected via the Gigabit switch, with the “test” PC having both the onboard and Killer 2100 network cards connected at the same time. The GaNE test is something created by BigFoot Networks to demonstrate the improvement their network cards have over onboard. It uses the User Datagram Protocol (as many games also do) to send 100-byte packets every 50ms (both of which they deem the average from examining games) and time the latency for each network interface card. I, however, didn’t run the GaNE test and this is for a few reasons:

     1) I disagree with its representation of game traffic – my reasons for which are below

     2) I cannot meet the recommended setup as the Killer 2100 does not work with my Gigabit switch

     3) BigFoot Networks never sent me the GaNE test, it was not attached with the other documentation and I never got a reply to the email I sent requesting it.

I stated above that one of my reasons for not running the GaNE test was me disagreeing with its representation of game traffic. To back myself up on this I used Microsoft Network Monitor v3.4 to capture the network traffic of four games: Team Fortress 2, Left 4 Dead 2, Call of Duty 4 and Battlefield: Bad Company 2. This data was then copied to an Excel spreadsheet where I performed a frequency distribution on it, which showed the following:

Game

Sample Size (No. Of Packets Captured)

Percentage of packets of size 91-110 Bytes

Percentage of packets of size 51-150 Bytes

Team Fortress 2

1048576

19.60

47.62

Call of Duty 4

281406

46.48

81.04

Left 4 Dead 2

10099

17.55

61.59

Battlefield: Bad Company 2

24152

19.58

64.45

To put it another way, the most common size for game traffic is around the 100 Byte mark. This, however, never makes up the majority of the game traffic. This information (as well as the full spreadsheet with my data and calculations) was sent to BigFoot Networks but I never heard back.
While the Killer 2100 may well not be tweaked towards the 100 Byte packet size but towards all possible sizes, only testing that area shows only half of the story. Given more time and a larger library of games I would analyse more games to see if they follow the same pattern.

To expand slightly on the second reason for not running the GaNE test, any other tests such as Netperf (or any of its equivalents) or timing the copying of files over the local network was because both the Killer 2100 and my onboard NIC are Gigabit capable. As the Killer 2100 didn’t work with my Gigabit switch I’d be forced to use a 100Mb switch which would be unfair for two reasons. It would not recommend a typical setup – savvy users aren’t going to use a 100Mb switch to connect two Gigabit devices (if they can help it). The other reason is that testing like this wouldn’t show which was better at whichever protocol (TCP and UDP) but the max bandwidth of the switch.

So what tests did I perform? Without being able to do the relatively synthetic benchmarks of GaNE and Netperf I was left with playing a couple of games and monitoring my ping throughout them. While scientifically this is not a good test – due to not being repeatable under exactly the same conditions and so on – it’s what the end user is going to do and see. They’re not going to run Netperf and are unable to run the GaNE test as BigFoot Networks only give it to reviewers. My two chosen test games were Team Fortress 2 and Call of Duty 4. These are highly competitive games (both featuring in the greatly popular i-series Multiplay LAN tournaments) and so are good tests where a low ping can make a great deal of difference.
The servers I frequent on Team Fortress 2 (a group of them based in the same location, run by the same clan) give me an average ping of 19 milliseconds when using the onboard network card of my Crosshair III Formula. When using the Killer 2100 I noticed the exact same average ping with the same highs and lows also seen. Using the “net_graph 3” command in the developer’s console shows the latency of the local network. It not only shows there being no difference but that there is very little latency added by the local network. Call of Duty 4 tells a similar story. One of the servers I have been frequenting over my time playing gives a ping of 27ms. The other server gives a 42ms ping, with the Killer being exactly the same on both servers. As I said I would at the beginning of this review, I disabled Avast to see if that made any difference and, as shown by the screenshots below, it did not.

So, to conclude, the Killer 2100 retails for ~£60 and comes with two main promises. The first is that it’s a plug and play device, to quote BigFoot Networks: “Simply plug in the card and install the software.  Killer™ 2100 automatically works with all online games – no mods, patches or updates required”. As I’m sure you can tell, this was far from the case for me in both of the encounters I’ve had with the Killer 2100. While I had the support of BigFoot Networks in the form of the Technical Support Manager being only an email away, something which I must say I appreciated, it still took some time before the card started working and I still have little idea what the original problem was or why it refuses to work with my Gigabit switch. The second claim that BigFoot Networks make is that the Killer 2100 “delivers fast networking with a new level of online gaming performance and a competitive edge over standard network interfaces”. This, again in my experience of the Killer 2100, was not the case due to it performing exactly the same as my onboard network card. Overall, and I’m sorry to say, I can’t recommend the Killer 2100 for either value for money or flat out performance.

Enjoyed this article? share it!

Add your own comment below. Please note that all comments are monitored by CCL.

Website User on 19th August 2011

Not sure why anybody would even consider buying something like this for on-line gaming. However fast your local network is (and with POE adapters in it, it will never be enormously fast), the internet connection is almost always going to be the bottleneck, especially when it comes to latency. As a user, you have no control how your packets are routed over the internet to the gaming server. Utilities like traceroute (or better equivalents) can help you see very quickly how many milliseconds it takes to traverse the LAN and parts of the internet.

Website User on 1st August 2011

Nice review, I purchased this card 2 weeks ago and believe it or not, you got further than I did. I tried this card on 2 systems, the first BSOD's constantly when installed and the second had problems realising the card was installed. Have been running through the sellers attempts to get this working and am currently waiting for an RMA as we appear to have exhausted al options now.