Bigfoot Netoworks' Killer NIC Sneak Peek

so it been over 20 days since HardOCP did the preview on the KillerNIC in which it praised (“hyped”) the product and told its readers that the advantages are there and that the difference are considerable from what they saw at KillerNIC’s offices.

HardOCP received good money (obviously) for this, which in all fairness is normal and totally acceptable and the ads run on the frontpage for this product.

What I do not find acceptable is that we do not get a review! No offence it does not take over 2 weeks to test a network card if we test FPS and latency improvements, for the applications on the network card I would like to read a separate review anyway.

Somehow I feel cheated, especially reading and seeing the benchmark results from another reviewer which show that the products marginally increased FPS and does give minimum ping improvements (considering 95% of latency lies beyond your ADSL modem… DOHHH)

I remember when HardOCP had a legal battle against a certain vaporware console manufacturer and another time changed their review system and decided not to use synthetic benchmarks because of issues with these and people cheating.

Common guys, I want the old days back. I want a fair review, even when you run ads and have to make a living. The community needs sites like this.
 
I heard a point made about the card being a vector for a possible back door by the hardware company who made it, possibly by some FCC mandate or whatever. I was surprised to hear this comment fromt he person because they seemed to me to be a logical, thinking, down to earth person.

The cheating point makes alot of sense too, but I'm thinking there will be very little chance of this because the seriously tallented people are less abundant than the cheaters, and less likely to get their hands on something like this and release some mod that the average cheater could actually accomplish on their own.

I'm still on the whole "why int he name of god did they use PCI?!?!?!!?" kick. I don;t care if 133MB/s is adequate for Gbit ethernet.... the PCI bus is shared!!!!!!! Which means large file transfers can result in music playback problems/video playback problems.

I challeneg anyone with a PCI express motherboard to try this:

1) Open device magaer

2) in the "view" menu, choose "devices by connection"

3) look through the items under "PCI bus" for any PCI-express root hub devices

most of your onboard devices are there, connected through PCI express. This is cheaper AND faster. Notice if you have an Nforce chipset, your nvidia network is on it's own bus. Need I say more?

My criticism isn't limited to bigfoot, creative is just as bad. PCI needs to be phased out and no new mainstream devices should be released under PCI unless they are released side by side with a PCI express part.
 
mhhh...
I have 2 cpu's in my computer, one of which has a 100% load while gaming and the other about 2% to 5% depending whats happening....

why offload when 1 CPU can handle all this Windows crap and any other application that are running in the background. I went dual core for 1 reason... offloading of non game code to the other core. ;)

all reviews I will read better use dual cores unlike on KillerNics offical page...

oh man....
 
If this improves my average FPS by 30%, i'll get it :)

I have a 7950 GX2 on a dual core opteron system so my expectations are high :D
 
LordBritish said:
If this improves my average FPS by 30%, i'll get it :)

I have a 7950 GX2 on a dual core opteron system so my expectations are high :D
Click to expand...


but it wont youll be lucky to get +10% out of it from what i have seen that over all avg. is more like +4% with the low being -12%
 
reyalp said:
This is not correct.

Using linux 'ping' as an example, since the source is avialable:
Before sending the packet it calls gettimeofday, and records that value in the packet. It sends the packet, and when it gets the reply back, it compares the gettimeofday value stored in the packet with the current time. In other words, it includes all software and hardware latency, including getting from the network driver to a userland program.

Feel free to inspect ftp://ftp.inr.ac.ru/ip-routing/iputils-current.tar.gz to confirm this for yourself.

Bigfoot claims that UDP has a lot more overhead than ICMP, but you can verify this is false for yourself. linux traceroute sends UDP by default. It gets an ICMP TTL exceeded message back. Thus it should give you a good indication of the latency associated with sending UDP. Switching it to ICMP and comparing the result is left as an exercise to the reader :)
Click to expand...

My aplologies about mis-describing the ping command functionality.
What I had in mind was sort of what you said (in spirit), where the hardware ping and software ping are black boxed. (specifically that the software latency isn't overt, and the 'actual' benefits of killer nic are real, but hard to *isolate* and measure.)

Lt = Lh+Ls.

Ping tells us Lt. (total latency)
We cant change Lh. (hardware outside latency)
We can manipulate Ls (software latency)

To audit my statement:

"because ping will just tell you how long it took to get a reply [Lt]. it won't tell you how long it took before your system bothered to send out the ping request [Ls]. [because we don't know the porportion of Ls and Lh inside of Lt, you can't tell how much time is wasted in Ls]"

Packed prioritization will still have the effect I described, since it will bring down the average Ls, and as a result bring down the average Lt.

Comparing killer nic, and a regular nic, if we assume killer nic has 0 software latency for a prioritized packet, then we can take the ping difference and speculate about what Ls is on a regular nic. But without some experimentation it's hard to say what Ls is. Suffice to say that it isn't actually 0 in either case, and is farther from 0 on a regular nic than the killer nic. Isolating Ls on each nic would let you get an accurate comparison (which is the ~same Ls delta as the Lt delta on the same network).

The statement immediately above is against what my last sentence states in my original post. I screwed up on that one. Sorry.
Sorry, I was a bit too careless in my original response. I spent too much effort emphasizing how we can't get direct information on the software latency, and I just got carried away. Ping is a good comparison for killer and standard nics. Thank you for drawing my attention to it.

-scheherazade
 
scheherazade said:
My aplologies about mis-describing the ping command functionality.
What I had in mind was sort of what you said (in spirit), where the hardware ping and software ping are black boxed. (specifically that the software latency isn't overt, and the 'actual' benefits of killer nic are real, but hard to *isolate* and measure.)
Click to expand...
Thanks for the clarification.

My original point many pages back was that any improvements in software latency will show up in total latency, unless they are insignificantly small. Furthermore, total latency gives you an absolute upper bound of software latency. No improvement in software latency can possibly exceed the total :)

On my home network (which is all outdated 100mbit stuff, with one of the boxes involved being a 66mhz FSB celeron), total round trip latency for a 1000 byte packet runs about 0.5ms, and only rises slightly if I use a significant portion of the links capacity (e.g. 1000 of those packets per second). This tells us something about the possible gains from reducing software latency, especially in games, which use a much lower volume of smaller packets. It also tells us something about the honesty of a person trying to sell you reduced software latency as a must-have feature. The other features they are trying to sell you may have value, but it should make you stop and think.
 
germanjulian said:
so it been over 20 days since HardOCP did the preview on the KillerNIC in which it praised (“hyped”) the product and told its readers that the advantages are there and that the difference are considerable from what they saw at KillerNIC’s offices.

HardOCP received good money (obviously) for this, which in all fairness is normal and totally acceptable and the ads run on the frontpage for this product.

What I do not find acceptable is that we do not get a review! No offence it does not take over 2 weeks to test a network card if we test FPS and latency improvements, for the applications on the network card I would like to read a separate review anyway.

Somehow I feel cheated, especially reading and seeing the benchmark results from another reviewer which show that the products marginally increased FPS and does give minimum ping improvements (considering 95% of latency lies beyond your ADSL modem… DOHHH)

I remember when HardOCP had a legal battle against a certain vaporware console manufacturer and another time changed their review system and decided not to use synthetic benchmarks because of issues with these and people cheating.

Common guys, I want the old days back. I want a fair review, even when you run ads and have to make a living. The community needs sites like this.
Click to expand...

they might have NDA's with the company till a certain date or something that prevents them from releasing certain information... who knows, i am sure we will get a review soon enough
 
I'd like to see other cards that offload networking loads from the system CPU run against the killer. Like this D-Link 560T: http://www.dlink.com/products/?pid=406&sec=0

Especially since I already have a DGL-4300 to take care of QoS for my game traffic.

I mean if an $80 card can offload even just a few of the more CPU intensive tasks, give me true Gbe connectivity via the PCIe bus, and work hand in hand with a wireless router that can serve all the computers on my LAN why would I spend $280 on a NIC? Just because it has some RAM and a mini-linux install?

While I think its cool and all, and the idea of uTorrent making a client that can offload the connecitivty load to the card so the cpu is left only to md5 hashing is sexy indeed, I am typically in the $500 upgrade crowd and if theres another viable also-good solution I'd like to know!
 
germanjulian said:
...
What I do not find acceptable is that we do not get a review! No offence it does not take over 2 weeks to test a network card if we test FPS and latency improvements, for the applications on the network card I would like to read a separate review anyway.
...
Click to expand...

I don't think they were planning on just reviewing it, from the last paragraph of the preview:

Kyle said:
Assuredly you will be seeing reviews forthcoming that showcase the Killer NIC. Many of those will be jam-packed with synthetic graphs showing this and that. As a pure gaming NIC, the true talent of the Killer NIC should be to give a better gaming experience for those skilled enough to take advantage of the ping reduction. This is certainly not going to be shown on any graph. That is why we would like to reach out to the gaming community in Austin, TX and have some of the best gamers there join us locally to give gaming on the Killer NIC a go. Then of course, give a regular NIC a go as well and tell us if you see a difference. Please mail me at Kyle @ HardOCP.com if you would like to see if you are skilled enough to give the Killer NIC a try.
Click to expand...

That said, it has been about 6 weeks now. Any updates Kyle? I'm a bit curious how this thing turned out.
 
note: not bitching about price here, but I have to comment on it.

$280 is ridiculous. Not that it's overpriced, but it's like buying a car with spinners stock. Who really needs to run bittorrent on that 95% of empty NPU power? You buy a computer to run that, not a NIC.

Just make a version with 1/4 or 1/8th the processing power and ram, and sell it for $100 or something. Then I'd consider it. And screw the USB too (unless adding all the hardware, software, etc needed for the USB adds less than $5 or so to the card).

I want a nic to connect to the internet, not to have a mini linux box in my PCI slot.
 
Russ said:
I want a nic to connect to the internet, not to have a mini linux box in my PCI slot.
Click to expand...

wait surely you already have a NIC to connect to the Internet otherwise you would not post here. :p and again 90% of latency comes from router to router connections on the internet.

Go to Poland... the have the fastest and best backbone in their country... (build with ADVA optical DWDM technology and ethernet over optic... great stuff)

:p if only the US network was anywhere close to europe
 
South Korea had some awesome networks also.

OnTopic

I'd kind of like to have JUST the mini-Linux-in-a-slot just to tinker with. I don't need the NIC, per se, but there is definitely tinker-factor in this device.
 
Yea, I did it, I bought one.............. YEOW MING!!!!!

sadly while it worked well for 2 days (and was impressive I might add), it is now borked by some auto upgrade THEY decided I needed.

I now have to do a format to get it back up by this weekend, customer service? LOLOLOLOL, pls don't ask.

That's why that format. You buy this thang you had BETTER talk to a few peeps that own it first..... and oh yea, check the return policy (buyer bewarze!!!!)
 
good you say you liked it.

nice we got a review too... oh well income is more important i suppose then a review
 
EODetroit said:
That's pretty sweet, I look forward to the results. Once again Austin seems like a cool place to live. College town, lots of IT companies, not as disgustingly conservative like the rest of the state... hah. Anyways, I'm seriously considering this thing, as I can see multiple reasons to, unlike the PPU devices which I'd get zero benefit from. And yeah, for perspective, I bought two I-Rams w/ 4 Gig each and Raid-0-ed them to play WoW off of. So this is my kind of eccentric overpriced device. I justify it with the word "hobby".
Click to expand...

Um, its typicaly conservatives the pour venture capital into small companies and fuel inovation. Socialism results in stagnation and lack of inovation.

Just my .02
 
jacuzz1 said:
Um, its typicaly conservatives the pour venture capital into small companies and fuel inovation. Socialism results in stagnation and lack of inovation.

Just my .02
Click to expand...


The politics forum is elsewhere....
 
germanjulian said:
wait surely you already have a NIC to connect to the Internet otherwise you would not post here. :p and again 90% of latency comes from router to router connections on the internet.

Go to Poland... the have the fastest and best backbone in their country... (build with ADVA optical DWDM technology and ethernet over optic... great stuff)

:p if only the US network was anywhere close to europe
Click to expand...


When Germany and Poland have to wire 9.1 million square kilometers, then we can compare national infrastructure. (Gerrmany is 357 thousand square km where poland is 316 thousand square km.)

Fiberoptic isnt cheap...
 
400MHz processor, 64MB of memory....this NIC's specs are disturbingly similar to the computer I bought in '99. Crazy. I want one. :D
 
Lol,


yah, it kinda is, except that PC from 1999 didn't fit onto a single slot PCI card with passive cooling.. and Integrate seamlessly with your much more powerful CPU... :)

Seriously though, i'm here if anyone has questions about how Killer works or what it does. Management at Bigfoot decided that they would let me (an engineer) be around to help answer questions as needed. (and I guess listen to customers too, lol)

TS

~~~~
Tom S.
Bigfoot Networks Engineering
 
tomstomper said:
Lol,


yah, it kinda is, except that PC from 1999 didn't fit onto a single slot PCI card with passive cooling.. and Integrate seamlessly with your much more powerful CPU... :)

Seriously though, i'm here if anyone has questions about how Killer works or what it does. Management at Bigfoot decided that they would let me (an engineer) be around to help answer questions as needed. (and I guess listen to customers too, lol)

TS

~~~~
Tom S.
Bigfoot Networks Engineering
Click to expand...
Welcome! Thanks for your willingness to supply this service to us, and to Bigfoot for making you available.
 
You must log in or register to reply here.
Back
Top