Performance

Method Of Testing

Testing a network card properly is quite hard because you have a multitude of different variables that you have to factor in. In order to test this card I played on Danish servers that I know are plenty capable of providing a smooth game play. During benching I joined a server, played for 10 minutes first with my normal network card and then with the KillerNIC. While gaming I tried to be in the same areas each map and using the same playing style. All of the FPS stats were logged with FRAPS.

These game tests were repeated with and without various FNA applications loaded on to the KillerNIC. I was initially going to include a F.E.A.R online benchmark, but since major fluctuation occurred (caused by F.E.A.R's immature net engine) I couldn't really get any decent benchmarking data. The same goes for CounterStrike 1.6 since my PC could produce 100 FPS constantly. Offloading the network stack to the KillerNIC would have provided no measurable FPS gain. So instead of these benchmarks of the card I have decided to focus on what I felt while playing with the card and without. The things I focused on were FPS drops and fluctuating ping because these have a tendency to ruin game play.

To test the FNA BitTorrent client I had setup a small private torrent network between my own computer and some TPU servers. At first I would use uTorrent with the same torrent file (3DMark) loaded and with the same amount of connections to the servers. After this initial test I would load the torrent in the FN Torrent client and download it to my Transcend USB 2.0 thumbdrive. The files transfered via the BitTorrent were about 500 MB with a general chunk size of approximately 10 MB.

Test Setup:

• Intel P4 3 GHz Prescott
• MSI 875P NEO FISR Motherboard
• 1GB GeIL ONE DDR RAM
• 36 GB Western Digital Raptor 10K RPM hard drive
• Windows XP SP2

Internet Connection:

• 100 Mbit/s upload and download
• ISP Provider: Jay.Net
• Location: Denmark

Counter-Strike: Source

Counter-Strike:Source is based on the Half-Life 2 graphics engine. All the tests were done using a CSSNation configuration of both web engine and graphics engine. This means that all of the update rates were maxed out and the graphics turned down to low in order to get over 60 FPS in game. At these particular settings the CPU becomes the bottleneck because the geometry provides a constant load factor whereas the graphics at low put minimal strain on your graphics card.
With this particular setup we should see a decent increase in the amount of FPS in game because the KillerNIC can offload the network stack and therefore leaving the CPU to focus on the geometry calculations. I did all of the tests on the same server on the same map. The benchmarking data was collected via FRAPS.
Each benchmark lasted 400 seconds on the same team with the same amount of people on the server. To further increase the accuracy of the benchmarks I was at the same place on the map each round. As it turned out people’s movement on this FFA server was quite predictable so the benchmark data should be quite accurate.


As you can see from the graph above the performance increase was noticeable in game. I completed four sets of first playing without the KillerNIC and then with. After analyzing the benchmark data I was pleasantly surprised the performance increase was substantial. The impact on game play while having an extra 15 FPS was huge, everything felt more smooth especially when there was a lot of player movement on screen. Another thing I noted while playing was that the amount of packets transmitted to the server from my client increased a bit while using the KillerNIC (Approx 5-10 reports/second). This increase in data transmitted makes hit registration a bit better in some cases where you and the opponent are both in motion. The increase in the minimum amount of FPS was pretty sweet. I am certain that even though 45 FPS is a low frame rate I prefer having 45 instead of 30 FPS. The 15 FPS can easily make the difference between you winning or losing a gun fight in a first person shooter like CS:S.


The increase in minimal FPS definitely has a huge impact on how smooth the game feels. That can't really be said for max FPS because almost nobody can feel the difference between 89 and 101 FPS in game.


With an average increase of FPS of almost 25 FPS this card proves its worth. You can clearly feel the difference, and that coupled with the minimal FPS that's also 15 FPS higher makes this card worth the money from a gaming point of view.

Overall the KillerNIC contributed to a much smoother game experience without any stuttering even when there was a lot of player movement on screen.


The ping didn't drop at all by using the KillerNIC M1 in CS:S.

Battlefield 2

Battlefield 2 (BF2) is one of the most popular conquest FPS games on the market today. The game itself relies on a an aged graphics and net engine. While playing with the KillerNIC I noted a small difference in the accuracy of hitboxes and a small increase in FPS.


The increase of approximately 3 FPS of the minimum FPS in BF2 is alright. It's a bit harder to feel the difference between the KillerNIC and a normal network card in BF2, but hit registration was improved substantially even though BF2 has a poor net engine.


Max FPS in BF2 wasn't increased at all. This is probably because the "spawn" screen pops up the very instant you are declared "dead" in game. I'm sure that if BF2 had the same spawn and first person spectator mode as CS:S we would see the same increase in max FPS.


With an increase of 13 FPS on average this card has a huge impact on BF2's game engine performance. One of the things I noted while playing with this card was how much smoother the game seemed to run. Even while there was a lot of movement on screen the FPS didn't drop as much as with my ordinary network card.

While playing Battlefield 2 I noticed that hit registration was a bit better and that my overall latency to and from a server dropped by 1 ms and was more stable. My ping to a stable Danish server was 4-5 ms with a normal network card and 2-3 ms with the KillerNIC.


The improvement is probably due to the fact that the KillerNIC can reduce the internal latency that your client produces by getting data from and to game faster. During my FPS tests I played two rounds on the same team and tried to be where the action was all the time. The FPS increase was substantial just like in Counter-Strike:Source, however, the FPS increase wasn't as high.