[H] Killer NIC Evaluation

I call shenanigans. My first hop ping (to my ISP) is less than a millisecond. That means that the lag contribution from my PC, my NIC, my gig-E switch, my firewall box and my cable modem all put together is less than one millisecond round trip. How can a NIC alone reduce lag that simply isn't there?? Frame rates? Sure, it could have an effect if it had a better driver/API and more acceleration features available in hardware.

$250 for 1-5fps more? Buy a video card instead.

Were these tests ABX'd (double blind)? Was enough data collected to ensure statistical significance?
 
spock,

I love the tech. questions, so I'm happy to answer.

One of the little known facts of game design is: How you ask windows for data matters A LOT.

The windows network stack can add 2-15ms or more to the latency of a game. Everything from buffers and copies adds latency to more drastic things, like WHEN the game asks for data, or WHEN windows informs the game that data is available.

Ping and Traceroute commands aren't running "in-game", so aren't dealing with when the game asks for data and when windows tells the game data is ready.

When we (Bigfoot) say that Killer bypasses the Windows network stack, it means a lot of things: one of which is that our hardware delivers a hardware interrupt directly to the WSP layer (the functions that games call). So polling games run faster, and blocking/interrupting games get their data faster as well.

If there are further questions I'm happy to discuss, also check out this link for more:
http://www.killernic.com/KillerNic/KillerAboutLLR.aspx
 
tomstomper said:
spock,

I love the tech. questions, so I'm happy to answer.

One of the little known facts of game design is: How you ask windows for data matters A LOT.

The windows network stack can add 2-15ms or more to the latency of a game. Everything from buffers and copies adds latency to more drastic things, like WHEN the game asks for data, or WHEN windows informs the game that data is available.

Ping and Traceroute commands aren't running "in-game", so aren't dealing with when the game asks for data and when windows tells the game data is ready.

When we (Bigfoot) say that Killer bypasses the Windows network stack, it means a lot of things: one of which is that our hardware delivers a hardware interrupt directly to the WSP layer (the functions that games call). So polling games run faster, and blocking/interrupting games get their data faster as well.

If there are further questions I'm happy to discuss, also check out this link for more:
http://www.killernic.com/KillerNic/KillerAboutLLR.aspx
Click to expand...
All well and good, but you guys need to seriously reduce the price of this card. Reading through the responses on this board, one theme stands out - "nice, but I'd upgrade every other piece of hardware in my system before I'd consider this card". You also need a PCI-E version. Even if you don't need the bandwidth, most people don't have a PCI slot to spare. My $.02.
 
tomstomper said:
spock,

I love the tech. questions, so I'm happy to answer.

One of the little known facts of game design is: How you ask windows for data matters A LOT.

The windows network stack can add 2-15ms or more to the latency of a game. Everything from buffers and copies adds latency to more drastic things, like WHEN the game asks for data, or WHEN windows informs the game that data is available.

Ping and Traceroute commands aren't running "in-game", so aren't dealing with when the game asks for data and when windows tells the game data is ready.

When we (Bigfoot) say that Killer bypasses the Windows network stack, it means a lot of things: one of which is that our hardware delivers a hardware interrupt directly to the WSP layer (the functions that games call). So polling games run faster, and blocking/interrupting games get their data faster as well.

If there are further questions I'm happy to discuss, also check out this link for more:
http://www.killernic.com/KillerNic/KillerAboutLLR.aspx
Click to expand...
I'd like to know what range the average gamer here is seeing in Windows Network Stack Latency. With just about everyone in here with fast CPUs, Video cards and tons of RAM, I would venture to guess that we're in the 2-5ms range. I usually ping in the 15-35 range in BF2 and there is no way I'm getting 20-15=5ms ping from the server.

Back in the Quake 1 days I would boot from DOS and load a DOS based TCP/IP stack to avoid the Windows Stack.That made a huge 30-50ms difference (300ms to 250ms :eek: ). That was on a P120 with 64mb of ram. WIth the speed on hardware today, I don't think the "Stack Latency" theory is significant for the average cost conscious gamer. Of course that only MHO.

If you can only see the true Windows Stack lag in game, why couldn't we just ping the server IP, get a 100 ping average, then hop into the game and record the same data. CMD PIng AVG - Game PIng AVG = Windows Stack Latency?
 
quick question:


Has there been, or do you plan observations on the other side? How are peoples reactions to a player using this card in their server?
 
L1ght said:
I would be all over one of these IF:

-they were less than $100

Nice article BTW.
Click to expand...


Exactly,sell it for 99$ usd and I am in for one for sure.225 to 250+ is way too much.
 
QuakerOatz,
Preface: I'm in the Eng. dept. so I can't speak to how the card is marketed, etc.

What I can say though is that the Calculation you propose is awesome. We do that calculation all the time (on a wide variety of systems + cpu's), and we do even more than that, we look at 'why' a particular game's ping to the server is different than the in-game reported ping.
(and why in some games case, it is within 2-3ms of the reported ping).
Across a wide variety of CPUs (including high-end Intel Dual Core's), some games are 2-3ms (or 1-2) and some are 15ms or more.... [hence Rook's question]

2 Major Categories of games are revieled (and I hope this answer's Rook's Technical question [though the snake oil comment is a bit harsh IMO].
1.) Games that poll: generally this type of game makes a network call to the OS once per Frame (e.g. FPS would be max packet rate)... so at 60fps, that's 1 packet every 16.6ms (that's the best the game could do due to polling). A large number of games fall into this category and PING now relates to FPS in a direct way! Killer will increase FPS for this reason: polling type games do not traverse the network stack because Killer will interrupt when data is ready, so a "poll" returns in 1 cycle. Now your FPS is going up, right, and so is your average ping potential! So if your FPS is now 100fps, your ping can be 10ms (a 6ms savings).
MOST GAMES are type 1.

2.) Games that are truly multithreaded (a small fraction of games out there: Quake 4 being one of them). This type of game uses "blocking" or "call-back" WSP functions in their game networking. This type of game is asynchronous from the FPS, so the only FPS benefits a Killer will provide are from freed up CPU power (which will vary based on throughput)...less than type 1 for sure. Ping benefits are there more consistantly than in type 1 though, because in this model Killer's ability to Interrupt directly from hardware to the Application (a very unique feature, because of the NPU), becomes a true hardware interrupt from the game. So, the network stack latency is completely bypassed yielding a consistent Ping benefit.

In summary, what I think makes Killer a decent upgrade for the price (everyone can decide that on their own), is that it will always have benefits regardless of other upgrades [CPU, GPU, RAM, etc]... (because of Type 1 and Type 2 games). So, for that reason, it'll live longer in my system than my old 7800 did... (especially since it now runs my firewall while I game, which I love)

Hope this helps, glad to answer more questions as long as I'm welcome.
 
Forgot to mention, the analysis I just did was for RX only. TX works kinda the same, and so, the total benefit (round-trip-time benefit) would be TX + RX benefits.

That's why when customers talk about Killer, they usually talk about 'hit registration'.
 
L1ght said:
I would be all over one of these IF:

-they were less than $100
-they get rid of that stupid heatsink

Nice article BTW.
Click to expand...

I wholeheartedly agree with you. I would purchase this Killer NIC if it cost $99.99. Getting rid of the big "K" heatsink would be a good way for cut the price by at least $30. And could be replaces with an aluminum non-fancy heatsink. I would rather have a Killer NIC than a Physix card any day of the week. :D :D
 
tomstomper said:
question [though the snake oil comment is a bit harsh IMO].
1.) Games that poll: generally this type of game makes a network call to the OS once per Frame (e.g. FPS would be max packet rate)... so at 60fps, that's 1 packet every 16.6ms (that's the best the game could do due to polling). A large number of games fall into this category and PING now relates to FPS in a direct way! Killer will increase FPS for this reason: polling type games do not traverse the network stack because Killer will interrupt when data is ready, so a "poll" returns in 1 cycle. Now your FPS is going up, right, and so is your average ping potential! So if your FPS is now 100fps, your ping can be 10ms (a 6ms savings).
MOST GAMES are type 1.
Click to expand...

well, if that is true, it bodes ... well... (let me be lame and say) "REALLY COOL FACTOR +10

but some of us (probably the ones that are bitching the most) are still running AGP boards, so spending 250 bucks on a NIC is laughable, especially when I have no trouble owning about 90% of any NME's I come across in any FPS (yes, I game too much, yes i've been at it since wolfenstien and before, but sure, I would be glad to try out this card... maybe for around 50 bucks though)

Maybe if you could explain why the cost is so high, that would help ease our minds a little. Specifically, can you just skim the surface, say per part?

With a product like this, I would assume much of your cost would be due to development and no necessarily processor and memory, like graphics cards do.

Sorry if I came across harsh, but skeptics get the answers.
 
adonn78 said:
I wholeheartedly agree with you. I would purchase this Killer NIC if it cost $99.99. Getting rid of the big "K" heatsink would be a good way for cut the price by at least $30. And could be replaces with an aluminum non-fancy heatsink. I would rather have a Killer NIC than a Physix card any day of the week. :D :D
Click to expand...

this post just BEGS the question... do we really need that big of a heatsink? How much wattage does this thing take up, and how hot will it get? Can we get some clock speed numbers?
 
tomstomper said:
QuakerOatz,
Preface: I'm in the Eng. dept. so I can't speak to how the card is marketed, etc.

Hope this helps, glad to answer more questions as long as I'm welcome.
Click to expand...

I've got quite a few technical questions for you:
1) Do you realistically see full multiplatform support between Linux and Windows (FNA/Firewall control/etc)?
2) How do you feel the offloaded stack will compare to the newer Windows Vista from-scratch implementation?
3a) Do you think it is possible to realistically build a dual/quad KillerNIC while still fully offloading the stack?
3b) Do you believe the KillerNIC technology could scale (given time of course) to 10/20Gbit as a possible competition to Infiniband/other 10GigE products?
4) Do you see any future for including FNA applications to assist in offloading other tasks such as cryptographic API's to accelerate protocols like SSL, or is the onboard processor perhaps too weak to handle this many intensive tasks?

Thanks in advance for any of the questions you can answer, if you can't answer a question for some reason that is very understandable.
 
First on costs.. I don't know the per-part costs, etc... I do know Killer has a 400Mhz NPU (embedded network processor), 64MB of DDR Ram, a custom Xilinx FPGA design, an 10/100/1000 Phy, 8Megs of Flash, EEPROM, and more. I also have worked on several Server Network Accelerator designs in my past, and none of them have ever sold for less than $1000.

Second, on the Heatsink... yes it is needed to dissapate heat from the processor, and it gets quite warm... the design was done to spread the heat evenly enough that the card would still work with near zero airflow (which is typical for a PCI card at the bottom of a case).

And then, on to the big list! (Great questions all, and thanks!)

1) Do you realistically see full multiplatform support between Linux and Windows (FNA/Firewall control/etc)?
Click to expand...
It works really really well, the list of 'helper' apps keeps growing, we've got a windows based: linux console, linux shell from cmd line windows, download, download+execute, and more (all via open source licenses). Not to mention the FNA Firewall... (and more to come)... our goal is ease of use to integrate linux apps into the windows desktop. This is one of my jobs at Bigfoot...so I know a lot about it.
http://www.killernic.com/KillerNic/KillerDownloadsFNApps.aspx

2) How do you feel the offloaded stack will compare to the newer Windows Vista from-scratch implementation?
Click to expand...
We've been looking at that (our CEO/CTO is), and so far everything comeing out of Vista is focused on TCP/IP. Winsock hasn't changed, only underlying TCP/IP.. (stuff that Linux has been doing for a while...). And games use UDP/IP... so NONE of the enhancements in VISTA apply to games (except a very small handful of rts games).

3a) Do you think it is possible to realistically build a dual/quad KillerNIC while still fully offloading the stack?
3b) Do you believe the KillerNIC technology could scale (given time of course) to 10/20Gbit as a possible competition to Infiniband/other 10GigE products?
Click to expand...
I believe that Linux + Hardware acceleration can do what you say, Killer itself can't (it doesnt' support 10GigE.

4) Do you see any future for including FNA applications to assist in offloading other tasks such as cryptographic API's to accelerate protocols like SSL, or is the onboard processor perhaps too weak to handle this many intensive tasks?
Click to expand...
Absolutely. The NPU we have has full cryptographic support built into hardware. We're looking at OpenSSL implementations now, for FNApps to use. (it's not simple though!)... It's an area of research I'm involved in daily. The encryption engines built into our FPGA and/or our NPU include AES, DES, 3DES, RSA, etc. (most of the SSL protocol).
 
tomstomper said:
Absolutely. The NPU we have has full cryptographic support built into hardware. We're looking at OpenSSL implementations now, for FNApps to use. (it's not simple though!)... It's an area of research I'm involved in daily. The encryption engines built into our FPGA and/or our NPU include AES, DES, 3DES, RSA, etc. (most of the SSL protocol).
Click to expand...

Thanks for all the responses, what I meant between Linux and Windows is running the card within a Linux environment (I see the Beta drivers don't support FNA/consoles?). The reason I ask is because I have one Windows machines, and too many Linux systems to count. I've got like 8 just in my apartment, which doesn't count all the ones I've setup for various groups. KillerNIC would have been an ideal solution for the HiperWall where we needed a lot of throughput/performance, but you guy's hadn't announced it and we ended up with 25 Apple's anyways :( Beyond that, it seems like it would be a highly ideal solution if you could say massively accelerate public key/shared key encryption for SSL. When I looked into it, it seemed like most ToE's started at like $500-800, going way up. And as far as SSL accelerators go... theres just nothing under like 3 or 4 grand that I found. So if you could combine both onto one card for $250, that's a bloody goldmine and I know I'd buy at least one. I'd probably buy one now if I could start writing apps on my Linux workstation.
 
Just wanted to throw my 0.02 in.

I bought one of the cards for a new system build based off the heavy advertising I see on HardOCP. I figured it must be a quality product since HardOCP is choosy about who they let advertise and what products they promote.

I can say without hesitation that the KillerNIC was the biggest POS I have ever used. It had driver bugs, my system was constantly hard locking, the pings were not noticably better and there was an annoying message in the taskbar about new firmware being available.

The firmware message was especially annoying as the flash utility did NOT work and you basically were in an endless loop of flashing / downloading firmware.

Had this review actually be a real one in the HardOCP labs they would have seen this piece of junk in action.

Read the reviews on NewEgg about this product then decide if it was just me or if this thing really is workthless.

-Nv
 
-(Xyphox)- said:
Wow, great article packed with tuns of info and incite
Gives a little better idea on just what this thing does
though i still think its not worth that price tag....
Click to expand...

I have to agree. While I'd love for there to be such a device on the market doing what this NIC does, I cannot help but not recommend the KGN due to less costly alternatives already on the market, which would likely give a similar level of improvement to gamers. Yes these other NICs are minus the KNA/Linux portion, but if you're not a linux developer or are only interested in improving your gaming experience then its a moot point. In other words: they're unnecessary feature that can be done without.

In my opinion, as a network engineer and long-time gamer, if Bigfoot wants to cater to "gamers", then I suggest they make another version of the card without the KNA/Linux bit and with a much reduced cost. There are far too many elements of a PC that can be upgraded with a similar cost and add much more to improving the performance of a PC, and this is one of the biggest competitor's the the KGN. Many are looking at the KGN as if other NIC vendors are their leading competition, when in reality the true fight is all about ROI, whether consumers realize it or not. We even saw this is the review when Kyle said "There are very likely other upgrades you can make to your computer for the same investment that will give you more in return."

Bottom line, if you're looking to improve the performance of your games, you could just grab a $100 or $200 server-class part and get the same kind of performance increase that was seen with the KGN.

I'd love to see a review that compares the KGN to other similarly priced NICs, "consumer" NICs (SMC, Linksys, Netgear, etc), and a myriad of integrated NICs. That's the review that's going to indicate what sort of ROI is possible. Right now all we've seen is that the KGN does add benefit. However, compared to other hardware in the networking field how does it stack up in a price/performace ratio. I'd be willing to bet money that there are more than a handful other NICs that will offer similar benefits for much less money.

After all is said and done, the KNA/integrated Linux is a nice feature. However, to those that won't take advantage of it it might as well be shovel-hardware that only serves to increase the cost of the card. If you use it then its a nice feature that's going to cost you.


**Edit**
With the TCP/IP stack being offloaded to the one on-baord, how does Vista's improved stack bode for the KGN's future in regards to benefits of offloading the stack?

**Edit 2**
I want to make one thing clear: the KGN is a good product and does what they claim. There is no doubt that it will improve you games, to a degree. My bones about it is that there are alternatives available that add similar performance to games which are less costly. If you like the other features of the card then you cannot simply look at the price of alternatives and their impact on games and still make a fair comparison--basically you'd be comparing apples to oranges.

**Edit 3**
Why no benchmarks with controlled networks? Random testing using an internet connection cannot give you a firm indication as to what kind of latency or bandwidth benefits the card actually offers. All you do know is that between tests from internat NIC and Killer is that there were differences for certain tests. Without eliminating the obvious quirks inherent in networking, particularly WAN/MAN paths, how are we to know that the networking benefits weren't just a blip on the ISP's part? Bottom line, plug in some back-to-back systems and give it another try. This will also help in testing the offloaded stack.
 
Theoretical question:

Say you're on a regular server along with 10 other players, we all have the same Killer Nic card.
We all have good ping times etc, life is good and the only person with lag is the guy in Colorado with satellite internet who gets kicked for dragging everybody down.

Who has the overall advantage?

What I'm getting at is if everybody buys this card, aren't the benefits over other players nullified because everybody has the same advantage?
I'm not sure if I answered my own question there but maybe someone can see what I'm getting at.

Lose the heatsink, the retail packaging and throw an OEM-style version out there and I think more people would sit up and take notice.

Kudos to BFN for innovation, it seems that nobody seems to want to take a risk and this is a good reminder of the early days of the Amiga where something new and interesting came out every week.

I'd like to own one of these cards, I really would.
I can't pay, let alone justify even half of the $200 asking price but maybe in time if the price comes down I can be up there instead of just lauding it over satellite-internet-guy (Who, incidentally is still kicking my ass).
 
what about a killer nic vs a nice pci nic...?

if it were 50 bucks and didnt have that n00b pre mod looking hs on it
 
Phonic said:
Theoretical question:

Say you're on a regular server along with 10 other players, we all have the same Killer Nic card.
We all have good ping times etc, life is good and the only person with lag is the guy in Colorado with satellite internet who gets kicked for dragging everybody down.

Who has the overall advantage?

What I'm getting at is if everybody buys this card, aren't the benefits over other players nullified because everybody has the same advantage?
I'm not sure if I answered my own question there but maybe someone can see what I'm getting at.

Lose the heatsink, the retail packaging and throw an OEM-style version out there and I think more people would sit up and take notice.

Kudos to BFN for innovation, it seems that nobody seems to want to take a risk and this is a good reminder of the early days of the Amiga where something new and interesting came out every week.

I'd like to own one of these cards, I really would.
I can't pay, let alone justify even half of the $200 asking price but maybe in time if the price comes down I can be up there instead of just lauding it over satellite-internet-guy (Who, incidentally is still kicking my ass).
Click to expand...

If everyone has the same advantage, then yeah you won't be the only LPB. The overall game experience would be better though, because your opponents aren't lagging either. It sucks to play against someone who's blinking across the screen randomly. Reducing gameplay latency is an admirable goal, for sure.
 
Yuck...

The Killer has some good features but nothing my Linksys/Linux WRT router can't do, not to mention the Cisco 827ADSL... (QoS, stateful inspection, etc)


What you really should have done is put in a $30-50 intel/broadcom gigabit nic with TCP offload enabled. All of a sudden it will be exactly the same if not better FPS/ping than the Killer. There's a lot more overhead on a linux SBC than on a offloading workstation or server nic. If there wasn't then you would see the server market going nuts over these things, which they aren't.


I would also like to see a test where you take a bunch of old cisco 10mbit routers and create a hell of a routing path to get 50-60ms latency internally between the test system and the "test server"

Then run the MP game through this path that you control and has the same latency every time. Do it for motherboard NICs, workstation nics ($50) and server nics ($200) then the Killer ($250)

hopefully you can script an action then run it every time with each different nic and measure the latency and FPS.

That will be the only real way to tell. There is too much variance going out through any public connection to tell if the Killer really does shave off 10ms on every connection. I would not be surprised to find the $50 intel nics with TCP offload outperform the Killer.

Also most game data traffic is UDP which has very low overhead anyways. If anything all these reviews of this product show is that you really should disable your onboard networking and get a cheap workstation card unless your onboard is a decent hw solution with TCP offload, QoS flow control, etc...
 
good discussion going on here everyone.

Just want to respond to the comment about our flashing tool and early drivers: there were some problems. We've worked out a bunch of kinks in the last 3 months; and so I'm sorry that the NvidiotPrime had trouble, we've worked out a bunch of those issues (ESPECIALLY FLASHING!) since then. We've not had those kinds of issues for a month+.

Anyways, the early kinks of our small company are worked out now, and now we're just adding more and more features.

Finally, regarding less costly alternatives: I don't think there are any, really. TCP/IP Checksum offload is NOT a feature useful for gaming... and it certainly doesn't implement Windows stack bypass. Let alone FNApps, which are by far becoming increasingly interesting... the BitTorrent client is due out any day now! (stay tuned)
 
Glad to hear kinks are worked out. My main disappointment was all the hype and rave ratings on HardOCP about 2-3 months ago. I built a new C2D system on the latest ASUS 975 Intel (WS 64 Pro) platform and threw in the KillerNIC. All it did was lead to system lockups, random blue screens, hard drive corruption (used onboard Raid on mb which got fried on several of the blue screens), and indigestion. I kept my cool and went to your website. I found a FAQ that said you were aware of the flash issue and to keep flashing over and over until it works. I was like WTF. That is not a solution. (Tried flashing 30 straight times -- same issue. I want those two hours of my life back). It also said you were "working on another solution". I dont mind being on the cutting edge, but not for $279 and not when my built in NIC did the job for free and without the added "bonuses". I bought into the hype and I feel like I got abused. If you've cleaned up your products in the past 1.5 months then great. I wish Kyle and Co. would have tested this thing in their lab. Had they gone through a complete install and test then you would not be open to my criticism. But, sadly, that was not done so all we can do is take you at your word. If Kyle does actually test and install one of these then I will accept any rating they give it at that point. Until then this looks like a sad case of advertising $$ getting a rating from H|OCP. Maybe that is what this site has come to. Who knows....
 
If I want to spend $225 usd on a gaming related upgrade, it'll be on a graphics card.

If I want to run embedded linux on a small format device, I'll buy an early edition Linksys router.

As for the Killer NIC - I'm going to give it a big "uh... no thanks" - even if it was free.

It firmly fits into the same category that I place special-platinum-coated-fuel-efficiency-boosting-bling-aftermarket-car-parts. That special category that smells funny.
 
Markintosh13 (cool name), no problem, it's not for everyone. Stats reported by lots of different reviews don't put it into the category your talking about...imo.

Yah NvidiotPrime, I wasn't extremely happy with some of the issues you report either, but I can say it is way better now.

I think you are being unfair to HardOCP, whose reputation alone should hold merit... believe me, we don't have the budget (anywhere near it) to influence Kyle, not even Nvidia could I think...

But I can say Kyle HAS had a Killer (or 2) for quite some time, and still does I believe. The latest drivers and installer can be easily downloaded and tried out. I assure you the flashing issues have been worked out, and the drivers are light years more stable than before.

As for your lost 2 hours, as an engineer, i am sincerely sorry about that. On the bright side, our 30-day return policy (no questions asked) is still in effect, and I hope you made good use of it. [note: not all the resellers have this policy, but some do, as does Killernic.com itself].
 
So being there for the actual tests (I did the CS testing) I'd like to respond to a couple things.

In response to the "they must have done something to rig the testing" argument.
First off understand that we were there on a saturday and other than the 4 testers the only other people in the entire building was Kyle, and the two bigfoot guys (techy and PR guy).

When the switch was being made from one setup to another Kyle was looking over his shoulder ensuring nothing odd was being done and the PR guy was with the rest of us. During the testing both Bigfoot guys were commonly looking over our shoulders or off discussing the impressions with the last tester. There was little to no chance to be somewhere "creating a storm of packets" to disrupt the testing.

Finally the results were so close for both cards that it pretty much rules out anything like that being done. If they were going to skew the test wouldn't they have made the skewed results a little more obviously in their favor?

As to the specific differences in the games;

Kyle and I discussed theories for why we saw such clear differences in some games vs others and came up with some possible reasons. Keep in mind these are only theories and may be wildly off the mark:

CS:
we saw much more obvious ping differences when playing on a server with high ping rates to begin with. Initially I played on the fastest server I could find that had pings between 10-20ms. Playing on this server I saw almost no difference between the two cards in fact the session I reported as being better was actually the onboard NIC. Later when we tried a server with ~100ms pings the story was much different with the killer NIC session showing an average of 10-15ms lower pings and bullet reg being obviously better.

So why is that? We theorized that at the low ping rates the Killer didnt have much of a difference to make, basically that the connection to the server was already so good that the killer couldnt make much difference. However the high ping server offered much more room for improvement.

So with that said why didn't I notice a huge difference outside of ping even on the high ping server? My theory is twofold, 1st the network code in CS has had years in which to be optimized and refined to provide a smooth game to almost anyone at any connection speed. Second from day one CS has included lag reduction technology that seeks to try to equalize the game among all the players. These two things combined in my opinion are why we saw better results in newer games like FEAR and WOW and more subtle differences in the older titles like Quake and CS

At the end of the day I have absolutely no reservations about the objectivity of the testing or of the relevance of the results posted. Sure Kyle could have posted a bunch of benchmarks and numbers, but in the case of CS what good would that really have done you? A graph would have shown a clear reduction in ping when using the Killer NIC however that ping reduction made very little difference in game play, so what value does the graph have?

Personally the testing answered exactly the questions I had, namely what real difference do any of the claims or functionality make? Is it going to give me a huge edge in my online game? a very small one? or none at all?

As noted in the article it heavily depends on the game being played but unless you already have every other bleeding edge component and are looking for that last few percentage points of advantage then your money is probably better spent in the more common upgrade areas of CPU or video card.

If i was an avid WOW player I'd probably already have one of these in my system as the differences were obvious and substantial. For CS or the other games i play it wouldn't make enough difference to me to justify the price as it stands.
 
J-Mag said:
Usually I do the same. However, if there was only minimal ping loss while doing so I think I might change my mind.
Click to expand...

There is a way to achive this and you don't even need an overpriced linux-networking-stack-on-a-PCI-card for it.

Give cFos Speed a try. It's a traffic shaping application (it prioritizes certain packets to maximize either bandwidth or ping with several concurrent connections). This application actually works - measurably - so you don't have to rely on "gamers" babbling about being "ahead of the server" and similar esoteric ramblings.

IIRC it's 30 days free trial and then it's €9 (~ $11.50) and it's actually worth the money. Gaming while DL'ing becomes viable and if you're running P2P applications that hog your bandwidth and connections, it completely eliminates the connection slowdowns when surfing the web. This is, by the way, not simple QoS.
 
NvidiotPrime said:
My main disappointment was all the hype and rave ratings on HardOCP about 2-3 months ago.

I wish Kyle and Co. would have tested this thing in their lab. Had they gone through a complete install and test then you would not be open to my criticism. But, sadly, that was not done so all we can do is take you at your word. If Kyle does actually test and install one of these then I will accept any rating they give it at that point. Until then this looks like a sad case of advertising $$ getting a rating from H|OCP. Maybe that is what this site has come to. Who knows....
Click to expand...


Please point out all the "hype and rave rating on HardOCP."

To quote for this article.

Getting the Killer NIC back into our own offices here recently proved to be a pleasurable experience. After the Killer's release a couple of months ago, there were some initial problems with proper installations and system errors. Today, from our experiences, those bugs seem to be well ironed out. It is worth mentioning that Bigfoot maintains an active support forum to help out those guys that need it.
Click to expand...

I am sorry you have had issues, but do not lie and blame HardOCP for your problems. If you are buying totally new hardware as an early adopter, you should be prepared for issues. I have no problem with BFN having ads on our site, and I think they have been a fine company to work with.

A quick trip to BFN's forums would have shown you the install issues people were having. Please do not blame HardOCP for your lack of research.
 
NvidiotPrime said:
Glad to hear kinks are worked out. My main disappointment was all the hype and rave ratings on HardOCP about 2-3 months ago.
Click to expand...

I'm pretty sure we all pretty much blasted the KillerNIC from day one. Mainly because of the price. Many of us are still skeptical and are playing the price-drop/waiting for PCIe game. I'll get a KillerNIC when it drops below $150 and is on PCIe. I have a feeling it will be another year or two. So, by then, hopefully I'll be looking at 10GB NICs.
 
One of the more notable things he conveyed to me was that his “machine was ahead of what was on the server.” He explained this was allowing him to take shots and get cover before others had time to react to his presence.
Click to expand...

I don't get this, how can your machine possibly be ahead of what is on the server? I mean if the server hasn't sent you the data yet then how is your PC going to know what's going on?

I'm sure the Killer NIC makes a difference, just not a $250 difference that's all.

I guess my biggest issue with the article is that it is completely subjective with subjective results. I'm not saying you need to whip out a 10 page Excel spreadsheet to satisfy people but a simple double blind test with some information about the exact configuration of the Killer and non-Killer machine along with framerate and maybe some serever side ping data.

Everyone keeps asking "what good willa bunch of numbers do you?" The numbers tell me if a product is actually making a difference. Sure it may "feel faster" but if I canget the same results with some just telling me that I have a Killer NIC in my machine, then why sped the $250? I don't mind spedning money on my PC, I just want to make sure that it's on something that is actually going to make a verifiable difference, not just something that "I think" is helping.
 
its intresting reading this thread, make me kinda want one!

jsut remove that hs and fix the price a bit and its golden...
 
tomstomper.... how does your card compare with an Intel MT1000?? Have you guys done a comparison??
 
Maybe we dont need ping graphs, or framerate graphs, they would certainly make a difference though if we could clearly see for ourselves. See if pings or frames were more consistant, or often erratic. A lot of people are asking for them, and with good reason. The very *least* the article should include the hardware setup, software and driver information.

And in efforts of being thorough, should have included some sort of test alongside an add-in NIC. Possibly have been tested on a lower end system too in order to see if theres a larger benefit from the Killer card in an older computer. Not to mention that would have allowed for a different onboard solution to be tested. Testing two different computers? Who would have thought of it..

Maybe Killer would have done better on a less than top-of-the-line system. The differences might be more pronounced, and would likely not match the experiences of the ideal system, maybe they would make enough difference for the low end system to warrant using in that application instead?

Those sorts of tests have already been performed with graphics cards over the years, thats something we take for granted. But when have these types of tests been performed with network solutions specificaly involving high end cards and low end solutions, each in different configurations? Dosnt it stand to reason that a slower machine will perform slower with an onboard solution and thus adding an external one into the mix might have more pronounced results?

It seems that testing and reporting proceedures these days involve slapping a card into a machine, fiddling about with it for an hour or two and then writing an arbitrary opinion without the slightest bit of data to back it up with. Did you guys collect ANY data at all? After reading that article i cant be certain as to how thorough the testing proceedures were.

I might be alone when i say this, but that lack of accuracy neither helps us as readers, nor Bigfoot as a product sales team. Half the reason that the results were inconclusive is because you cant expect any meaningful results if you havnt exerted extensive control over the test environment, its variables, or followed a comprehensive testing plan that included more than a single comparison.
 
why didnt you guys go out and purchase one just like the average joe computer geek would do and run it in one of your own rigs? also what about people running sli and a sound card? seems to me people with that config are going to be hard pressed to find room to install this thing.
 
I thinks it's obvious by the numerous, well described comments (minus a few trolls) that [H]'s review here is lacking the information many of their readers need to make a objective purchase decision here.

The sad thing is that the theme we are getting from Kyle is:
"So what exactly is your problem with this? "
"Again, this was not the focus of of our article, but the data is there with other sites should you require it. "
"There are already several sites with that data should you wish to use Google"
"I am sure you will have no trouble finding a myriad of other sites that can give you exactly what you are looking for. Your readership will be missed. We appreciate the support you have given in the past."
"Should you like a bunch of graphs and data, there are plenty of other sites that cater to this."

Is anyone else getting the "Tough sh!t, go somewhere else" theme here?

Maybe instead of dismissing valid and repeated suggestions and sending us to other sites, you might want to consider that (shudder), you produced a shitty review and maybe next time you might want to change a few things?

But then again we wouldn't want our opinions to get the way of your methodology.

Where is Steve on this? I would really like to hear how he weighs in on this.

And yes, I know I'm a pain in the ass.
 
QuakerOatz said:
Is anyone else getting the "Tough sh!t, go somewhere else" theme here?

Maybe instead of dismissing valid and repeated suggestions and sending us to other sites, you might want to consider that (shudder), you produced a shitty review and maybe next time you might want to change a few things?

But then again we wouldn't want our opinions to get the way of your methodology.

Where is Steve on this? I would really like to hear how he weighs in on this.

And yes, I know I'm a pain in the ass.
Click to expand...

Did you happen to read the description at the top of each page? I am unsure why your expectations of the content would be as you described.

Gaming with the Killer NIC
You have seen the benchmarks and read the ping data. Now find out what the Killer NIC can do for you in real world gaming situations. No canned benchmarks, just feedback from gamers that have used the card.
Click to expand...
 
The opinions of those involved in the testing of the KillerNIC for this article are very much in line with my own opinions. I reviewed the KillerNIC a couple of months ago and found it to show improvements across the board as well. The main issue that I saw, and indeed the main issue that Kyle seems to see with the KillerNIC is its price. There are quite simply better upgrade paths available for the same price. However, there are scenarios in which the KillerNIC can be beneficial.

Sidenote: for some numbers, an alternate testing methodology, and my take on the KillerNIC, check out my review over here

kyle: I am very sorry if I broke a rule by posting that link!
 
shoes said:
Sidenote: for some numbers, an alternate testing methodology, and my take on the KillerNIC, check out my review over here
Click to expand...

While i dont fully agree with the single solution or single machine testing, this was a thousand percent better in terms of a review, opinions, or just information in general. Here we have pictures of the card, the contents of the box, a few graphs.. not to mention system specifications and a description of the testing methodology, a brief rundown on features.. Props for applying some effort to the testing and documentation, shows that you were trying to isolate the results. Id be more inclined to recommend it as a read for anyone who wasnt satisfied with the local coverage of Killer thusfar.

As for what Kyle has responded with on multiple occasions, its been made pretty clear now that if you want a full review of Bigfoot's Killer NIC, you should go elsewhere. Im not sure why thats the case exactly, generally its in ones own best interests to keep their readers both informed and happy, but we shouldnt dwell on that matter as it will only frustrate things further. He has his way of reporting and thats to be respected, im sure it works for some people. On that note however if anyone else wants to post links to more complete reviews that they felt were a good read, perhaps we can still cover all the bases here before the thread loses cohesion.
 
You must log in or register to reply here.
Back
Top