Killer NIC K1 and Custom BitTorrent Client Tested

NetworkingNed writes "The new Killer NIC K1 is the successor to the much debated original Killer NIC card that offers the same features at a lower price: this time for about $170 or so. Not cheap, that's for sure. But in this review at PC Perspective, not only is the new card tested under the drastically updated Vista networking stack with improved results, but the free BitTorrent client that runs on the Killer NIC is reviewed as well; with it you should be able to download torrents without affecting online gaming performance. Enough to warrant a $175 network card?"

Comments

Obvious

[AndroidCat]

The NIC has its own processor, will run a Bit Torrent client and save to its own USB drive.

But will it run Linux?

Re:Obvious

[solafide]

Yes it does. Scroll down in the PC Perspective article to the FNA=Flexible Network Architecture section - it's about a page down. It talks about how the card is basically a miniature computer, running Linux.

Spend the money elsewhere

[tomhudson]

If you're so worried about bittorrent degrading your performance, save your money - haul out that "obsolete" 1-2ghz machine and you won't have to leave your main box running (and costing electricity) when you seed.

Re:Spend the money elsewhere

[snowgirl]

I said it before, and I'll say it again. If you buy this card, then you deserve Everything You Deserve To Get(tm).

I'll still say that you don't need to trademark every silly thing that your card does "special". Like, "This case now with SafeCorner(tm) so that you're likely to get less BloodNStuff(tm) on your NetworkBOOST(tm)"

It just sounds way too much like you're marketing snake oil...

Screw the "NIC" aspect but..

[antime]

$170 for an embedded (Coldfire?) computer on a PCI card is not that much, and it could actually be useful for other tasks like monitoring, logging and administration. The on-board FPGA could also be used to offload some processing jobs, but it probably doesn't have too many gates.

Do they take your brain when you get one?

[khasim]

In every single one of the "reviews" of this card the "reviewer" has been a complete idiot.

Having never tested a network card (except to stress the maximum bandwidth of on-board solutions), especially one with claims of gaming benefits like this, I spent a long time finding a testing method that I was comfortable with.

Why not just setup a test network with a workstation with that NIC, a test server, a sniffer and some test scripts?

You image the workstation so you can start clean with each NIC you're testing.

You use the sniffer so you can see what is actually on the wire.

You use the scripts instead of doing anything manually because you want to remove the human factor as much as possible.

Online gaming is notoriously unreliable and unrepeatable as we all should know by now. Servers can be slower or faster based on the time of day, number of users online at the time; personal ISP connections can vary based on line quality, number of users in the area online at the time; global networks can go up and down and stream traffic anywhere at any time!

YES! Those are all the reasons why you run your own test server instead of adding additional variables to a test. So, are you going to do the test correctly?

For WoW, I selected a busy server, and attempted to play at the same time during a week day to try and always have a similar traffic level.

I guess not. Even with knowing every reason NOT to do that, you went ahead anyway.

I then used FRAPS to monitor our frame rates during the online game play and used the in-game ping monitoring for each title, reported every 10 seconds or so to another person writing the answers down. Each test was run 10 TIMES; nope, not kidding here. I wanted to be VERY sure that our results weren't a fluke, in either the Killer NIC's favor or not.

So what I'm wondering is why haven't we seen any REAL evaluations by people who know what they're doing? Do the Killer NIC people simply refuse to provide hardware to anyone who has a clue?

The on-board networking on the 975XBX2 motherboard was used for the non Killer NIC tests.

So you didn't even bother to test against a mid-range card? You used the chip on your motherboard.

Here is the torrent FNApp at work! You can see I have four files being downloaded at one time, though only two are transferring at the time. One of the things I wish BigFoot Networks had included was a transfer rate and maybe a way to see what you are actually uploading.

That's why you would use a sniffer.

In Day of Defeat: Source, I saw a 5% frame rate increase when using the Killer NIC versus the on-board networking on the Intel 975XBX2 motherboard.

And, once again, you didn't even go out and pick up a $50 NIC to compare it against.

The game did "feel" a bit faster, just as I reported in my testing under Windows XP, but once again, I am hesitant to put too much weight on that claim as it is such a vague and hard to verify point.

That's why you script the tests.

Under FEAR, I saw even more impressive results, especially considering that in Windows XP I saw NO change in performance.

And that didn't tell you something?

The model that supports FNApps is going to cost you around $179 retail, and Newegg.com has it for sale for $178 as of this publication date.

Seriously, you didn't test against a $50 NIC?

Re:No, but yes...

[tomhudson]

"It's not going to save any electricity. You rather have 2 boxes on while you're gaming instead of one, and while you're not gaming you still have one sucking electricity. There's no real energy savings here.

You might game for an hour or two, and download/seed for 24 - for 22 of those hours, your main box is off and not using electricity - and its more than likely that you can run the older box headless, saving even more juice; also that the video card in the older box doesn't run as hot ...

An even better solution...

[NerveGas]

Just get one of these. An external hard drive with built-in wireless networking and a built-in bittorrent client. No computer needed to download.

Set it up, let it leach off of an unsecured wireless network until the owner catches on, then switch to another one. No DMCA letters (at least not to YOUR door), and gaming performance on *your* network won't suffer at all!

Yes, that's bad in several ways. But it's still an interesting/funny thought!

That's the whole point of the story.

[DrYak]

But will it run Linux?

That's the whole point of the Killer NIC : It *does run* Linux.
The whole story can be broiled down to the Killer NIC being in fact a nice small router with loadbalancing/QoS/Pcket prioritizing. Plus a small server with it own mass storge pugable in USB.
The Killer NIC is nothing more than a glorified router shrinked to the size of a PCI card.
Once you get the basic idea there are only two quirks :
- It is sold completly ready to go. Whereas /. geeks lovingly tune their traffic shaping scripts to reach optiml balance between their latency sensitive application (VoIP, interactive SSH, gaming), their bandwith critical apps (File download, file sharing) and the rest (IMing, real-time meteo, etc.), this card comes "pre-tuned" so joe six pack has only to plug it to enjoy the benefits of QoS.
- As this is a PCI card and not a box that must communicated of the internet, the driver can use special hooks and directly tap into the Windows TCP/IP stack. Thus the router can sort and select packaets before they even leave the computer. Thus joe's gaming traffic gets put in front with higher priority than the traffic generated by the dozen of spywares/trojans/virus/spam zombies running in background.

Basically it's targeted to the same people who need quad-core CPUs : geeks who want to hack it, and clueless users who need to still have performance even when everthing is crawling under the load of crapware.

Re:No, but yes...

[arivanov]

Seconded - the thermal footprint of an average P3 after replacing the disks with modern ones is in the sub-50W range. The CPU depending on the model consumes 18-27W at max utilisation, disks are at most 10W each and peripherals rack up 10W or so on top of that. This is comparable to the thermal footprint of a 1GHz+ mini-ITX which is about as low as you can get with modern x86 hardware.

Compared to that a modern gaming capable system runs happily into the 400W+ territory. Even with all the advances in power saving modes on the peripherals and the CPU you are likely to find running an old P3 for router/firewall/P2P/file server/etc considerably more efficient compared to allocating these resources on your "main" box.

The only problem is the scarcity of CPU fans for P3s. There are none on the market. Athlon heatsinks/coolers for the older socket format often need cutting bits off and are also getting rare, so finding a suitable set to refurbish an old box may prove extremely challenging.