Killer NIC gets an early review

Bigfoot Networks’ Killer NIC “gaming network accelerator” has yet to hit stores, but the card has landed in IGN’s labs and been the subject of an early review. We already had a brief look at the Killer NIC’s features earlier this month, and IGN’s review mostly covers the same things. However, the folks at IGN also ran a preliminary benchmark with fairly interesting results:

Based upon some rough benchmarking, without the Killer NIC our test PC (AMD FX 62 2600, ATI Radeon X1900 XTX, 2 GB RAM) runs F.E.A.R. at 1024x768x32 with details maxed and 2xAA-2xAF at an average of 74 fps though two 15 minute sessions. Our ping to server ranged from 40-80ms, and our average worked out to 67ms over the two sessions.

After installing the Killer NIC we got back on the same F.E.A.R. server and played on the same map. The improvement with the Killer NIC was noticeable and impressive. Pings to server dramatically improved, generally ranging from 17-50ms, averaging 45ms over another two 15 minute sessions. FPS also got a nice bump, jumping our average from 74 to 86 fps.

That said, as IGN itself points out, benchmarking the Killer NIC accurately is a difficult task. Online games are a very volatile environment, and differences in ping or frame rates could very well be attributed to changes on the game server or in gameplay itself. Even if the Killer NIC is indeed responsible for the lower ping and higher frame rates, though, gamers might find the card’s announced $279.99 price tag to be a pretty serious deterrent.

Comments closed
    • Aihyah
    • 13 years ago

    i sometimes get as low as the teens in counterstrike(source) using an intel pci network card and dsl. generally 30 ping is what i can expect, but lower isn’t rare at all. so i don’t see what the fuss is about. its not exactly a new card either. it was 50 bucks when it came out, so its nothing special.

    • Shintai
    • 13 years ago

    y[

    • sluggo
    • 13 years ago

    How accurate is the ping-o-meter in Counterstrike? I was playing on a server last night as was getting a solid 100-102ms ping in-game. I logged out and pinged the game server and got a solid 68ms. This may mean something or it may mean nothing, depending on what’s actually being measured and how well it’s measuring it.

    On the surface, though, if there is indeed a 32-34ms delay in processing packets while in 3D apps, maybe it bears looking into.

    • alphaGulp
    • 13 years ago

    Here’s how a psychologist might evaluate the card:

    Have two people with identical machines in separate rooms play 1 vs. 1 deathmatch on a public server. Keep track of the # of wins.

    Take one machine and replace the NIC with the “Killer NIC”. Take the other machine and either overclock the GPU or replace the GPU so that you get the same boost in FPS.

    Tell both subjects that they received the “Killer NIC” and have them resume play. Track the # of wins.

    Repeat with 20 pairs of people to get a nice statistical sample and publish.

      • lyc
      • 13 years ago

      20 correlated samples is hardly “nice” in statistical terms…

        • alphaGulp
        • 13 years ago

        What’s not nice about it? Based on my hazy recollections from college stats courses you generally need a minimum of 12 sample points to be able to calculate anything meaningful in statistics.

        The number of necessary sample points is dependant on the complexity and number of factors you are trying to eliminate. Looking for correlation between a particular disease and say, being left-handed, would require a huge sample because there are so many other factors that have an impact and because it takes so many years for the symptoms to appear.

        If you play a game against someone 20 times, you don’t feel like you have enough data to spot a trend?

        Besides, multiply 20 pairs of players by, say, 20 matches and you have 400 sample points… Or do you believe two players need to play 10 thousand matches before you can discover any trends?

    • alphaGulp
    • 13 years ago

    Another way to test it:

    Have two identical machines (with synchronized clocks) playing a multiplayer game on the same _internet_ server. Keep a log of ping times and compare to see if the ping times are close to one another.

    As an aside, communication for this type of game is only from client to server, never from client to client, so being on the same public IP and same local network should not be an issue.

    Anyhow, assuming that the ping times are close to one another during the play session, replace one NIC with this “Killer NIC” and play again, comparing the results once more.

    • stmok
    • 13 years ago

    It runs Linux? I’m sold! ๐Ÿ˜€
    I’m gonna turn it into a firewall!

    Although…If it ends up costing MORE than that Snapgear firewall-NIC solution (which also runs an embedded form of Linux), then I wouldn’t bother…

      • Jigar
      • 13 years ago

      Offcourse u r sold …. to buy that card u will have to sell urself

    • drsauced
    • 13 years ago

    OMG, I cannot believe people are taking this thing seriously…

      • Shining Arcanine
      • 13 years ago

      A CPU can only process one thing at a time unless it is a dual core processor and the developers of <insert game here> are so talented that they are able to parallelize things such that physics, artifical intelligence, networking, the game engine, etcetera can all run in separate threads with any two running at the same exact instant in time in a dual core processor and the networking thread getting priority over all of the other threads. But since that is extremely rare, things will occur one at at ime with networking being underprioritzed. This underprioritization could delay a PC from sending out network signals when it should, causing lag. Logically speaking, having a processing unit dedicated to processing network traffic would ensure that does not happen, such that it would be more optimal than having the CPU handle it even in the best cases of programming on the part of game developers, because when networking overhead is handling by a CPU, other threads (e.g. the kernal, graphics driver, etcetera) could cause a slight delay in processing network traffic, which a dedicated unit would not have. Although, when game developers implement a 100% optimal software solution, the gains from having a dedicated processing unit would be around the point of diminishing returns, but given the difficulty of multithreading things and the huge inefficiency of optimizing compilers, the chances of that happening are extremely low.

    • notfred
    • 13 years ago

    If someone wants to review this properly, it needs to be done under controlled settings on the network side. The way to do that is to play on a LAN that is not connected to the Internet, but simulate the WAN latencies and packet drops of a typical Internet connection using something like the NISTNet Network Impairment Simulator ยง[< http://www-x.antd.nist.gov/nistnet/<]ยง

    • Vaughn
    • 13 years ago

    Even if this thing drops your ping from say 60 to 30ms and gave u 20fps say 90 from 70fps. Is it worth almost $300 dollars ?

    and #21 that is an excellent idea, this needs to be benchmarked on a LAN setting and a Public internet server so people can see what the deal is.

    • dukerjames
    • 13 years ago

    Here’s an idea on how to bench this card,

    Get 2 same spec machines on a lan and start playing games, 1 with the Killer card, the other with onboard, both should be in the same game session, record average ping/stats/fps in both machines. Swap ONLY the network card and play again, record stats.

    If the card really does work then it should show better average numbers in both test.

      • sativa
      • 13 years ago

      true, but its tough because you can’t easily duplicate the same network play. Reviewers need a way to emulate key and mouse activity so that they aren’t running from a timedemo, but are actually ‘playing’ the real game in the same manner each time.

      • Firestarter
      • 13 years ago

      to get similar gameplay on both the machines, you can play on one and set the other to observe you. play for 15 minutes, switch pc’s, play some more. switch NICS and repeat

      if this card isn’t all bull, you should be able to tell by the numbers.

    • sativa
    • 13 years ago

    y[

    • blitzy
    • 13 years ago

    good luck benching your latency on a public network, traffic is never gonna be the same… better have a much larger sample size than two sessions if u wanna even have a rough guesstimate

    • Bensam123
    • 13 years ago

    So uh did they use the same server for both tests?

    You know they could also benchmark it by running a series of pings through command prompt while a game is just running without actually being connected to a server (run a normal benchmark while a preset number of pings run in the background). It would simulate a game enviroment and the ping associated with it without the ‘volatile’ nature of online games.

    • Bauxite
    • 13 years ago

    Hi there vapid press release people, my ping from NIC to x86 router is <1ms @ 1gbit with a very low cpu util, wake the hell up.

    TR please aquire one and show it for the BS it is…just please don’t actually pay from your own wallets.

    • 5150
    • 13 years ago

    Forget TR reviewing it, for $280.00 I could get more than 10 more fps.

      • thecoldanddarkone
      • 13 years ago

      I think I would get a new video card before getting that card, or sound card, or new motherboard, or processor, well I don’t think I would ever spend 300 dollars on a nic…
      edit…
      Just in case your wondering, I know your being sarcastic.

      • Thebolt
      • 13 years ago

      I don’t think FPS is what’s important here, the fact that it boosts the frames per second is a perk but at the same time I don’t think it matters at all because it only raises the bar where the CPU is the limiting factor, when that’s the case with newer CPUs it doesn’t matter how many frames you get because it’ll be smooth in any case.

      If you want more FPS then yes, get a new video card. If you want lower pings and have a tree in your back yard that grows money then you should get this card.

        • thecoldanddarkone
        • 13 years ago

        lol, that was good. ahhh the money tree.

    • Krogoth
    • 13 years ago

    Spend the money on a better ISP or broadband modem/router which will help reduce latency by a sigificant amount if the existing unit is sub-standard.

    Anyway, when latency gets to >50ms it begins to depend upon how the fast data moves around your system’s interial buses not the network connection. I know this first-hand from LAN’ing with fast systems versus slower systems.

    The bottom line is that “Killer NIC” is nothing more then a casturated server-grade NIC (PCI bus = auck!). It’s actual benefits only come into play when you are transfering a crapload of data(100+Mbps) with a lot of connections which is clearly not found in gaming or torrenting on the average broadband connection.

      • Bauxite
      • 13 years ago

      Wow I forgot this puff of smoke was PCI…

      At least go x1 so you can pretend its good enough for gigabit and today’s enthusiast boards.

      Honestly though, $280 would probably be better spent trying to bribe your local city council to put more pressure on the local telco/cable utilities to bring fiber to all residences a la fios.

    • CampinCarl
    • 13 years ago

    TR needs to review this against some “regular” gigabit ethernet cards. Including “onboard” ๐Ÿ˜›

      • continuum
      • 13 years ago

      Hell, I think I have a Xytratex XL-1 here… :-p

    • StashTheVampede
    • 13 years ago

    Seeing a benefit in FPS would be one thing, but couldn’t “gamers” buy NIC cards that are already available? There *has* to be NIC cards with dedicated chips for maximum performance, right?

    For the netplay gain: getting a linux based router and configuring it for UDP priority would do you a larger benefit, right? Ultimately, it’s your bottleneck to the net — you might as well configure it to your benefit.

    • crabjokeman
    • 13 years ago

    I guess if “Fatal1ty” branded products are selling well, these should go pretty good too.

    • Arkham
    • 13 years ago

    Isn’t really only useful if your connection is being used simultatneously for other purposes, e.g. gaming while your kid sister’s filesharing and your brother is Quaking?

    • flip-mode
    • 13 years ago

    Wow! I want one … for me to poop on!

    • herothezero
    • 13 years ago

    Still needs an authoritative review from TR.

    • Thebolt
    • 13 years ago

    looks a lot better than I would have expected. If only it were under $100 it might be worthwhile assuming those results are correct.

    • Gandhi
    • 13 years ago

    For $280, upgrading the video card is a better use of the money.

      • wizboy11
      • 13 years ago

      You can get a X1900XT for close to that price.

        • swaaye
        • 13 years ago

        Or a Parhelia (hahahaha!!!)

Pin It on Pinterest

Share This