EndLagNow.org Forums / EndLagNow.org Forums / ELN News and Announcements / Killer NIC by Bigfoot Networks announced!!!! / Latest Posts

RE: Killer NIC by Bigfoot Networks announced!!!!

exglade — Fri, 01 Sep 2006 22:59:04 GMT

May I ask some question?
How KILLER effects my internet speed?
What really effects my internet speed?

RE: Killer NIC by Bigfoot Networks announced!!!!

mlampkin — Thu, 24 Aug 2006 14:28:46 GMT

Yup... I've already gone thru the various docs on the site including the white paper...

I will admit to being a bit of a skeptic... at least on the cost / benefit site as a gaming product...

I'm also a bit worried about virus / malware scanners... hooking into the network stack is an old malware trick that used to be used by ( among other things ) some p2p setups... to secretively track a user's network behavior and forward it to third parties... so I understand that is NOT what your software is doing but you did mention 'hooking' into the net stack to provide the by-pass... and the issue is concerns about mis-identification of the Killer NIC software / mods by scanners...

The truth is that I do love my games... but my day job for the past couple of decades is doing software consulting with a focus on networked systems ( including games )... and the thing I would be looking at the Killer NIC for would be on the FNApps etc. side... and potential of using it as a fast 'pass thru' and offload for custom protocols that generated heavy loads e.g. having it do handling of custom security protocols... since most add on cards out there are tied to specific 'standard' schemes and not adaptable to such situations...

Michael

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Thu, 24 Aug 2006 11:27:24 GMT

mlampkin,

You seem to have a good nack for understanding a lot of the issues involved! Have you read the white paper yet?

http://www.bigfootnetworks.com/PDFs/KillerNic_LLR_White_Paper.pdf

As for your comments about other Gig NICs, you've hit the nail on the head: they are all optimized for throughput (large pkt sizes, etc. etc.): We're literally the opposite.

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

mlampkin — Wed, 23 Aug 2006 23:56:11 GMT

Drifting on and off topic but... I felt some of this required a response...

John Schultz (8/21/2006)

mlampkin (8/19/2006)

* the docs mention increases for games that use non-blocking sockets using selects... and how this is a CPU waster... yes - if someone were naive enough to write a program that used non-blocking sockets, calls to poll / select with a 0 timeout... and um - the networking / animation rendering engine / etc. all running in a single thread / process... but people don't do that because its really the worse possible design... typically there is a main process / thread... one or more network handler threads... a primary rendering thread ( for static / known events )... and another thread / queue for network cmd and / or user intiated modifications to the display... etc. etc. etc.

Perhaps I am misunderstanding what you are saying here... but...

I too could could write a program to put a socket in non-blocking mode and instead of using a call to the select ( or poll ) function - put it in a loop that e.g. continuously tried to read write from the socket... and if I jammed to that socket at capacity for the interface then the performance would ( probably ) be better... since at that high ( and constant ) of a level of throughput almost every read operation will in reality be able to grab data..

The thing is though that you cannot design a networked system and make any assumptions about traffic level consistency... and in fact you can be certain that the traffic rate ( within limits of course ) will be inconsistent... and then performing e.g. read operations without determining if there is actually data ready first - turns a large(r) number of your read ops into nothing but eating CPU cycles... cycles which could be better utilized refreshing / manipulating the graphical layer...

Now on a positive note... any network program utilizing the type of set up that ( I think ) you are describing... would probably benefit the most from my understanding of the Killer NIC... since all the zero return reads etc. are short circuited ( in a good way ) on their way to query for / retrieve / write data which would result in far fewer consumed cycles...

On single processor systems, multi-threaded network programming results in a performance loss. Thread context switching is expensive, and the extra layer of data copying and marshalling further reduces efficiency. As multi-core CPU's become more common, creating threaded network code for game servers will afford improved performance as long as Mutex (Semaphore, CriticalSection, etc.) overhead and data copying does not overshadow CPU parallelism. Dedicated game servers for MMOG's using multi-cpu systems can take advantage of Overlapped I/O and IOCP (Win32), which is efficiently threaded internally.

I'm not certain I would call thread context switching expensive... switching out the program counter, stack pointer and a few other registers ( not ALL of them ) is actually a very fast set of operations... now if we were talking about context switching between processes - yes - that can be expensive - but we're talking about threads...

Of course... some OLD threading impls were notoriously bad... since threads were actually disguised processes... but I'm concluding we aren't talking about those... lol

Now going crazy with anything is bad... and if someone were to write a program / game and micro-manage everything into tiny little tasks each with their own thread... it would be a bad thing... but I believe I mentioned 4 threads... one which is really just used for starting up / cleaning up ( the main )... graphic display, keyboard io and network io... and just for clarification their priorites would be set - in order - to very low, normal, high and high... and the latter two should be floating so they could be dynamically lowered / raised when required to limit keyboard & network flooding from overly affecting the display FPS...

On the topic of multiple CPUs... mutexes and what not... and worries about the associated overhead overshadowing parallelism... I don't think that will be an issue since the vast majority of CPUs provide a command(s) that perform ( atomic ) compare / exchange and that is what we use for creating mutexes and such at the kernel level... the ops are very fast and trust me - its not going to be an issue even when talking about synch'ing multiple CPUs... ;-)

Related to threading proper... to effectively utilize multiple CPUs you're going to want your code to use multiple processes or threads... with typically a preference to the latter since there is much less overhead in the context switching / data sharing / etc. ... otherwise you're going to be wasting a CPU...

As for OICP... AIO... the majority of 'edge triggered' IO ops... these in general actually do very well without multiple CPUs - though of course extra CPUs almost never hurt performance... and can significantly decrease CPU utilization in comparison to using things like non-blocking io with ( or without for that matter ) select / poll calls... the reason is that most of these implementations actually by-pass many layers when sending notifications / what not up to the application layer in comparison to what is required for diving 'downward' with reactive type ops...

Harlan, have you tested with Intel CSA motherboards? For those not familiar, CSA MB NIC's do not sit on the PCI bus, but instead use the CSA pathway, allowing for improved performance. Interestingly, latency in one of the CSA benchmarks I reviewed was not improved over PCI cards.

Related to this and Harlan's comments / response... its part of the reason I was curious and inquired about MTU and packet sizes used during testing... since some of the other gigabit capable cards out there really appear to be optimized as high speed data streamers ( doing best with big MTUs etc. )... and maybe not really that desirable for a gaming system...

Anyway...

Apologies for being long winded... I was confused about some of the comments made... figured that they may have been written because of a mis-understanding of what I was trying to state in portions of my previous message... and just wanted to make certain I clarified things...

Michael

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Tue, 22 Aug 2006 05:55:30 GMT

Thanks for the pointer! Pretty much everything intel makes has great throughput: they focus on large packet sizes and effecient throughput (rather than latency)!

That was always the tradeoff with my Designs at intel as well.

RE: Killer NIC by Bigfoot Networks announced!!!!

John Schultz — Mon, 21 Aug 2006 23:57:31 GMT

CSA benchmarks faster for throughput: http://www.gamepc.com/labs/view_content.asp?id=st61g4&page=4&cookie%5Ftest=1

(My main dev box is an ASUS P4C800-E (Gigabit CSA)).

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Mon, 21 Aug 2006 19:41:44 GMT

Great points John. That coincides well with what we have seen. (of course Killer helps quite a bit even with threaded designs for reasons already mentioned).

When I worked at Intel, we designed a proprietary bus in the past that didn't sit on the PCI bus... but in the end it still has to arbitrate with either the southbridge or the northbridge. (in my design it hung off the southbridge! which is bascially still PCI).

I've not yet personally tested with Intel CSA (we are trying to get one in the Game Testing Lab soon)... but it still has to interface to the ICH (essentially a Northbridge) which makes it little different from just PCI-E, which we have tested and found the same results you mention: no real benefits.... the reason is that PCI/PCI-E and CSA aren't really bottlenecks for Online Gaming (they are all sub 10us latency adders)... which is noise in the grander scheme.

Killer, of course, is designed to change the game (so to speak), by getting rid of the real bottleneck (in many games) which is Sockets overhead and strangeness (and possibly ineffecient network coding design).

We clean up the mess, in essense: so PCI/PCI-E nor CSA really matter.

I have some friends who helped with CSA, by the way, and I'll ask them what the real motivation was... usually it's done so that they can do better "Bundling" (e.g. proprietary bundling) with system makers.

Tytus (also known as Harlan, for those of you who don't know)

RE: Killer NIC by Bigfoot Networks announced!!!!

John Schultz — Mon, 21 Aug 2006 17:29:14 GMT

mlampkin (8/19/2006)

If one uses non-blocking sockets, select() is not needed. I benchmarked blocking sockets with select() and non-blocking sockets (checking error codes, bytes read, etc.) and found non-blocking sockets without select to be faster (Win32). However, when reading the socket once or twice per game frame, it does not make a significant difference.

On single processor systems, multi-threaded network programming results in a performance loss. Thread context switching is expensive, and the extra layer of data copying and marshalling further reduces efficiency. As multi-core CPU's become more common, creating threaded network code for game servers will afford improved performance as long as Mutex (Semaphore, CriticalSection, etc.) overhead and data copying does not overshadow CPU parallelism. Dedicated game servers for MMOG's using multi-cpu systems can take advantage of Overlapped I/O and IOCP (Win32), which is efficiently threaded internally.

Harlan, have you tested with Intel CSA motherboards? For those not familiar, CSA MB NIC's do not sit on the PCI bus, but instead use the CSA pathway, allowing for improved performance. Interestingly, latency in one of the CSA benchmarks I reviewed was not improved over PCI cards.

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Sun, 20 Aug 2006 21:31:20 GMT

mlampkin, thanks for your questions... I think you'll find our FAQ discusses a lot of things: but I agree some of your questions are not addressed. The main reason is we are trying to both satisfy information hungry folks such as yourself while keeping things relatively simple for everyone else. I'm going to try to respond to some of your questions below. Thanks for the interest. And I know we will be releasing more and more information as time goes on.

http://www.bigfootnetworks.com/FrequentlyAskedQuestions.aspx

mlampkin (8/19/2006)

* the Yukon chipset is being used to show a FPS increase when using the Killer NIC... but at least some of the drivers for that Yukon cards are well known CPU cycle wasters... so why in the world was the Yukon used as the only reference in that section when there were 3 other NICs with more efficient drivers mentioned later...?

This was an "early look" at performance analysis. You can expect more extensive information soon. We chose the Yukon for this test as it performed the best on the calls per second metric against many of the other NICs we've tested against.

* speaking of cards and drivers in general... no mention of specific driver versions used for the various 3rd party nics is made... or of the configuration of those cards in general ( many of the gbit cards have quite a few 'extra' options )... or configuration of any intermediaries if present in the test network... continuing below...

You can assume all the latest drivers, and no tweaks whatsoever. This includes no tweaks away from default for our card either that keeps it fair.

* many gigabit cards require a high / jumbo MTU to see their full benefit and are optimized at the higher MTU level... for many in fact - their speed when operating with an MTU at / near the 1500 size is actually lower than 100Mbit cards... but the docs don't mention the MTU used for testing... and it also doesn't mention what MTU size the Killer NIC is optimized at...

* there is also ( again ) a lack of information on the UDP packet sizes used during testing... we get the situation where if using a lower MTU size and packets that exceed that size... then of course the card which does onboard re-assembly is going to get an definite advantage... where as increasing MTU or decreasing packet size would probably equalize / flip the results the other way...

The only test we published was a UDP TX Blasting test (that another board member pointed out had some issues)... but for that test 8192 packet sizes were chosen.
On the gaming 'early look' performance, of course, whatever packet sizes the games were using.

Our card is designed to accelerate gaming: so we have not focused on throuhput nor large packet sizes (as gaes don't use either).

Glad you mentioned that... MORE games than you might imagine use such methods. But in fact, we can help games that use multiple threads as well. We've optimized every possible game type we could figure out (without source code).. and analyzing how the use network sockets.... it was a FUN job to do: and we believe people will be VERY pleased with the results.

* it sounds like ( its hard to tell ) that part of what can be loaded into the Killer NIC is really just a refined QoS type system... where priorities ( and subsequent processing / placement into the network queue ) is based on packet ids... but if this is the case - doesn't it mean that you have full cooperation of the game manufacturers to implement the 'rules' to be loaded i.e. so you know what ids are used for ping packet(s)... what ids are used as graphic triggers... for voice... so you'll be tied into games that support this card... or losing the potential benefit...?

No cooperatin of game makers is needed for any of Killers features. QoS is used (we call it GameFirst) and this feature is really the smallest feature of the whole NIC!
See Here for latest feature information http://www.bigfootnetworks.com/killerfeatures.aspx

* the talk of 'bypassing the TCP stack' isn't anything new... while I haven't personally written anything like that which is tied to hardware / a specific NIC - I have written code to sit at the kernel level and intercept data and send / queue it directly to the application layer... but using this type of set-up required either the application knowing a set of ( new ) custom calls to retrieve / send data OR dropping in a modified OS network stack ( e.g. replacing the Window's stack )... this comes back to ( as far as I can tell ) the fact that it appears cooperation from the game developers and programming against the Killer NIC's API would be required to get the benefits... ?

Nope. No integration with the games is required: we hook into the MS stack. But you are correct, TCP Stack bypassing has been done before in Corporate Server NICs that cost upwards of $800-$5000 and don't do much for games (most games today use UDP). So we've done a BIG thing here creating a sub $300 card that can rival those guys for functionality...

I hope this helps!

And again, thanks for the questions!

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

mlampkin — Sat, 19 Aug 2006 18:03:01 GMT

Things that are missing / raise questions in the docs etc. on the Killer NIC website:

* the Yukon chipset is being used to show a FPS increase when using the Killer NIC... but at least some of the drivers for that Yukon cards are well known CPU cycle wasters... so why in the world was the Yukon used as the only reference in that section when there were 3 other NICs with more efficient drivers mentioned later...?

* speaking of cards and drivers in general... no mention of specific driver versions used for the various 3rd party nics is made... or of the configuration of those cards in general ( many of the gbit cards have quite a few 'extra' options )... or configuration of any intermediaries if present in the test network... continuing below...

* many gigabit cards require a high / jumbo MTU to see their full benefit and are optimized at the higher MTU level... for many in fact - their speed when operating with an MTU at / near the 1500 size is actually lower than 100Mbit cards... but the docs don't mention the MTU used for testing... and it also doesn't mention what MTU size the Killer NIC is optimized at...

* there is also ( again ) a lack of information on the UDP packet sizes used during testing... we get the situation where if using a lower MTU size and packets that exceed that size... then of course the card which does onboard re-assembly is going to get an definite advantage... where as increasing MTU or decreasing packet size would probably equalize / flip the results the other way...

* the docs mention increases for games that use non-blocking sockets using selects... and how this is a CPU waster... yes - if someone were naive enough to write a program that used non-blocking sockets, calls to poll / select with a 0 timeout... and um - the networking / animation rendering engine / etc. all running in a single thread / process... but people don't do that because its really the worse possible design... typically there is a main process / thread... one or more network handler threads... a primary rendering thread ( for static / known events )... and another thread / queue for network cmd and / or user intiated modifications to the display... etc. etc. etc.

* it sounds like ( its hard to tell ) that part of what can be loaded into the Killer NIC is really just a refined QoS type system... where priorities ( and subsequent processing / placement into the network queue ) is based on packet ids... but if this is the case - doesn't it mean that you have full cooperation of the game manufacturers to implement the 'rules' to be loaded i.e. so you know what ids are used for ping packet(s)... what ids are used as graphic triggers... for voice... so you'll be tied into games that support this card... or losing the potential benefit...?

* the talk of 'bypassing the TCP stack' isn't anything new... while I haven't personally written anything like that which is tied to hardware / a specific NIC - I have written code to sit at the kernel level and intercept data and send / queue it directly to the application layer... but using this type of set-up required either the application knowing a set of ( new ) custom calls to retrieve / send data OR dropping in a modified OS network stack ( e.g. replacing the Window's stack )... this comes back to ( as far as I can tell ) the fact that it appears cooperation from the game developers and programming against the Killer NIC's API would be required to get the benefits... ?

So I guess the point of all this... I really find the supplied docs too 'light' on details...

But if I were to make a guess...

It seems like you guys are really trying to do the equivalent of what embedding OpenGL etc. functionality / commands on a video card does... but of course providing an API for network access instead of graphic manipulation... BUT if that is the case then I would think success would be getting the API out to the world so people could work on / with it... otherwise the card and all the ( potential ) extra functionality will wasted and not worth the expense to the user(s)...

Anyway... more information... updates to the online docs... and so forth... would be appreciated... if for no other reason than simple curiousity...

Michael

RE: Killer NIC by Bigfoot Networks announced!!!!

Elios — Sat, 12 Aug 2006 09:49:04 GMT

alright lets say i make a lan CS:S server lan pings from in side that server with other PCs on my network are in the 10-12ms range soo over my LAN pings are nic and low i get very high framerates @1600x1200 maxed settings 100avg FPS

now on my fav CS:S server i get 70ms pings from inside the server now i live with 2 other ppl and thay use the net here too if thay use BT that goes up to 100ms big deal

now i did the math for were this server is it would take light ~20ms(oneway 40round trip) to get there
soo theres 30ms to 60ms getting added in some were and id bet its due to network traffic on the LAN and or the the cable modem being half duplex and the hops it has to take to get there

so unless this thing can some how send packets faster then the speed of light i dont see it doing much

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Sat, 12 Aug 2006 08:51:34 GMT

Elios (8/12/2006)

ok lets try this ONE more time

its been dug up that you guys won 4 million it start this thing up
WHAT has been done with that

Eilos, this comment makes no sense. Nobody won $4Million, if they had, it'd been all over the news!!!
We won the right to allow U.T. to invest $100,000
http://www.mccombs.utexas.edu/news/pressreleases/mootcorp2005_2.asp

Later, we raised over $4Million from investors... as in a real business.. and the investment dollars are part of the business (e.g. research, development, working captial...etc.) I hope this puts it to rest.

Running Ping command to your gateway is not going to tell you the overhead your networks stack creates, or how your game would respond. Ping is just telling you how much latency is on your wires... not how much is introduced by your NIC, your operating system's UDP stack, the game itself...etc.

i realy dont see how this card works

I'm here to help... ask any question (without getting personal attacks, etc.) and I'm happy to answer if I can.

RE: Killer NIC by Bigfoot Networks announced!!!!

jeepdave — Sat, 12 Aug 2006 08:48:57 GMT

Elios (8/12/2006)

ok lets try this ONE more time

its been dug up that you guys won 4 million it start this thing up
WHAT has been done with that

"dug up"? Dude, it's posted on their website under 'corporate news'. Doesn't take Magnum PI to find that

Elios (8/12/2006)

and as the post that was deleted said YES i read it

like Magoo i have the means to buy this thing if i wanted its just i can see 300 bucks spent else were that would do that same thing like say a nice Cisco router over a 70 buck Linksys from bestbuy

That's cool, although you might wanna check www.pricewatch.com, they usually have lower prices then best buy. Sounds like you like tech toys as much as I do.

Elios (8/12/2006)

to the tech guy
excatly HOW does this thing make MMOs less laggy WoWs lag is 99.9% server side

and i have been doing some tests on my own network ping times i have here are <1ms to my gateway pings from my gateway via its software show 40-100ms pings to the same places

seems 99% of lag i see is the out side net work
i have a very beefy PC i was messing with CS:S lastnight and i get 100fps with a low of 70 with totaly maxed setting on a 20 man server

i realy dont see how this card works

"EXACTLY" how it works is at: http://www.killernic.com/KillerNic/PDFs/KillerNic_LLR_White_Paper.pdf. It's a little above my technical vocabulary, but maybe you understand it.

I'm off to play some WoW, PEACE OUT!

RE: Killer NIC by Bigfoot Networks announced!!!!

Elios — Sat, 12 Aug 2006 07:25:02 GMT

ok lets try this ONE more time

its been dug up that you guys won 4 million it start this thing up
WHAT has been done with that

and as the post that was deleted said YES i read it

like Magoo i have the means to buy this thing if i wanted its just i can see 300 bucks spent else were that would do that same thing like say a nice Cisco router over a 70 buck Linksys from bestbuy

to the tech guy
excatly HOW does this thing make MMOs less laggy WoWs lag is 99.9% server side

and i have been doing some tests on my own network ping times i have here are <1ms to my gateway pings from my gateway via its software show 40-100ms pings to the same places

seems 99% of lag i see is the out side net work
i have a very beefy PC i was messing with CS:S lastnight and i get 100fps with a low of 70 with totaly maxed setting on a 20 man server

i realy dont see how this card works

RE: Killer NIC by Bigfoot Networks announced!!!!

jeepdave — Sat, 12 Aug 2006 07:01:03 GMT

Tytus (8/11/2006)

Techie,

It's expensive for a lot of reasons:
The main reason is that designing chips/ building chips and boards, etc. is just plain expensive.

Tytus

Is that why I just shelled out $600 for my Nvidia 7950? Almost beat out my $1000 Lacie Terabyte external firewire drive....of course, I'm a gadget freak.

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Fri, 11 Aug 2006 12:03:44 GMT

Techie,

It's expensive for a lot of reasons:
The main reason is that designing chips/ building chips and boards, etc. is just plain expensive.

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

techie81 — Fri, 11 Aug 2006 09:49:19 GMT

I just want to know why this piece of hardware is so expensive.

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Thu, 10 Aug 2006 20:13:23 GMT

magoo,

most of the folks in here applaud the effort of fighting lag: even if they can't afford the product.

I hope you can do that at least.

And I also hope that when more specs are released, you'll consider Killer (as you do other proven products).

It's not easy to get a new product to market... literally we're working 16hour days to make this happen.

When it's out there: it's going to wake up a lot of people: and hopefully get them thinking about Lag.

In the meantime, do fight lag however you can!

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

magoo — Thu, 10 Aug 2006 14:41:35 GMT

Well I tell you what.....if I'm going to spend 300 bucks......I'll put it into any one of several components in my computer with known advantages and reproducable benchmarks.

I still think a Physx card is a losing proposition and something as nebulous as 3 milliseconds, that has multiple variables which can manipulate the result is really hard to grasp.

Your nic looks nice. Thats as far as I'll go. But then again, so did the Phantom Console and the acclaimed Phantom wireless gaming keyboard/mouse......neither one made it to retail.....millions of dollars later.

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Thu, 10 Aug 2006 10:13:26 GMT

magoo,

thanks for your post! I'll answer what I can.

In short, TODAY (or tomorrow)! I am planning on posting very soon the 'History of Killer'...

I will show our early board designs all the way to the latest design.

I hope it pleases!

As for VALUE of how it works: the performance will speak for itself. And gamers that want performance, will love it.

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

magoo — Thu, 10 Aug 2006 09:05:26 GMT

When can we expect to "see" the first real product.

Pardon my skepticism but this product seems to have little value to all but a very small number of people. I am a very avid computer enthusiast, I game on-line, and I could care less about "lag", its like wind in golf.....you just have to know how to play it. I am VERY hard pressed to think, wait -- conceptualize how anyone can perceive a 3 millisecond delay in transmission.

You're going to have to prove this to me....or it goes in the bucket with the Phantom Gaming System.

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Tue, 01 Aug 2006 13:25:52 GMT

Thanks DanDruff!!!

You ain't seen nuthin yet!!!!!!!!

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

DanDruff — Sun, 30 Jul 2006 19:16:46 GMT

Not that I understood "everything" on the white paper...the benchmark data looks very impressive!!!

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Thu, 27 Jul 2006 12:22:51 GMT

For all of you who had questions about : HOW DOES IT WORK!!!!

The answer is now available in a white paper!

Detail on how the Killer NIC www.killernic.com work are now available in a white Paper.

Download it now!

http://www.bigfootnetworks.com/PDFs/KillerNic_LLR_White_Paper.pdf

And discuss it here.... that's why I'm here!!!

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

Stinger — Tue, 18 Jul 2006 20:11:13 GMT

Thanks. I would love to upgrade to dsl or cable but it is going to be atleast a year or two before dsl comes to the area I live. I doubt if we will ever see cable.

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Tue, 18 Jul 2006 19:38:03 GMT

GREAT question!

short answer: sorry you are stuck with satellite!!! upgrading to dsl/cable will be your best upgrade EVER!

that being said, Killer NIC will help some, making your connection a tiny bit more stable: and with FNA, i expect new utilities to come out that might help more.

for the present, Killer is designed best for dsl/cable/t1/t3/etc

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

Stinger — Tue, 18 Jul 2006 18:33:17 GMT

I connect using satellite internet and my ping in games is always high. Everything I have read about satellite says that this is pretty much the norm and it's a waste of time gaming with this type of a connection. Will your card be able to help me or is it something that will only work with the good connections?

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Mon, 17 Jul 2006 09:09:59 GMT

I don't think you will be convinced at this point, that's okay! Your not supposed to be. We haven't released product, performance results, or anything like it as of yet!

But I do agree international latency is a problem: international packet loss is a bigger problem actually.

If the latency is known and steady, any game designer can work around that problem: an unstead packet lossy international connection is another matter altogether!

As for the LPM definition, let's discuss that here:
http://www.endlagnow.org/ELNForums/Topic172-9-1.aspx

At any rate, we have a long list of reviewers and folks who want their hands on our card!!! I'm afraid you'll have to wait like everone else for the release.!!!

On a purely technical basis however, there is very little marketing that I do on a day to day basis! Believe me when I say, there is an incredible amount of innovation and engineering effort going into this product. Not all of which will be evident to the average person or engineer at first (FNA, for example will require some additional effort on the behalf of our users to develop some amazing things that run in Killer [like VOIP, pkt sniffers, filters, firewalls, etc]... all of which are totally possible with FNA technology.)

That's pretty darn amazing by itself!

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

Vince — Mon, 17 Jul 2006 08:36:10 GMT

Hi Tytus,

Thanks for taking the time to read and respond to my post, I appreciate you staying around to discuss this with us.

I've just read your information on LPM, and I'm astounded by your definitions and it's relation to your product. You're essentially saying that there are 3 main causes of lag, correct?

Client Lag - Essentially when your PC is overloaded, and cannot render a game at over 45FPS.
Your product could do nothing to help this. Any PC that is overloaded to the point of lagging when playing a game online is clearly under the specifications to play that game, and needs certain components upgraded. The absolutely miniscule CPU performance increase your hardware may offer them will probably not even give them a 1FPS increase.

Think back to the days of 56K modems. There were generally 2 flavours, controller-less (Soft Modems) and proper Hardware modems. I had the former, and noticed at least a 40% increase in CPU usage on a very very old Cyrix 166MHz CPU. Those were back in the days when hardware was less advanced, and more expensive. I don't believe NIC's ever suffered from the same problem, nor with the common imlementation of pro-2GHz CPU's and increasing cache sizes, that this is even an issue that does exist anymore.

For your product to be able to combat PC overload, it would somehow have to act as either a GPU or a CPU - which would not make it an NIC. Even if you somehow offloaded all network processing to your card, I'd put money on you not being able to tell the difference.

Network Lag - Basically when your ping is less than 100ms. Apart from the fact that network interrupts could cause momentary lag, or perhaps ping spikes depending on network conditions. But again, your card can in no way decrease international latency - it's international latency for a reason; the fact it has to cross several thousand miles of pipe/cable.

If your product can significantly decrease international latency times (I'd bet a million dollars it can't, if I had it, heh) then you will have created the Holy Grail of networking. Does anyone believe in The Holy Grail?

Server Lag - I'm a bit confused by your definition of server lag, but then I'm not a network engineer - you are/were. I'll take your word for it

I'm trying really hard to be optimistic of your product. Not because I want to buy it and try it, but because there's a chance I could be completely wrong and this could be an amazing product for people. Just with my limited technical background, and the input of several other technical people I know personally, it just doesn't seem like a reality to me. I'll be honest, it seems like a very clever marketing scam. If you, however, want to send me a trial card I'll test it for you and eat my words if I'm wrong

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Mon, 17 Jul 2006 07:42:40 GMT

Vince, thanks for your passionate post.

This is EndLagNow.ORG, and that kind of passion is appreciated here!!!

You obviously care about ping and lag and all things online gaming... and for that I applaud you... I hope you'll stick around and continue to be involved in ELN.

In the meantime, I can answer a couple of your questions:
1.) Bigfoot Networks will release more and more details of our technology as we go forward. The big question is: How? and the reality is we can't say too soon, or we'd create our own competition.

I do encourage you to read a little bit about LPM (Lags-per-minute): I think you will find it interesting, that while ping is a contributor to Lag, it is not the only contributor.

http://www.bigfootnetworks.com/LPM.aspx

2.) as for watching every packet to determine it's type, etc. Trust me, this has been done for over 12 years...very effectively. Literally this is what we mean when we say 'corporate server network acceleration' technology (a technology I've been involved with for 10 of those years, including architect of several such devices).

3.) Killer NIC may not be for you. That is okay! Killer NIC is designed for people who want an edge... who want to win... and who love technology. I encourage you to read a little more about the FNA technology we are proposing... FNapps could revolutionize what happens in your computer. and this is only the beginning.

Enjoyed your post: I wish you were this passionate with Game Developers, Broadband companies, Router companies, etc. etc. If so, we might all make serious inroads on Lag!

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

Vince — Mon, 17 Jul 2006 03:08:38 GMT

Disclaimer: I am not a 'network engineer' or anything of the sort. I know bits and bobs about networking, and the Internet in general, having worked for ISP's over the past few years.

To be completely honest with you, this card sounds like a complete scam. Let's consider something here... if you play a game on a LAN what is your ping? Usually <1-3ms right? In a good environment, with decent cabling and a fairly average router, you'd probably find that your average ping is <1ms. So can someone please explain how this card will ever possibly benefit gamers and give you a 'lower ping' as stated? How is it possible to have lower than 1ms ping?

Before anyone retorts with how it will improve your Internet/WAN gaming ping, then please also explain how it could ever achieve this? To understand why it simply can't, you need to understand what latency is in the first place. MANY MANY factors affect latency, most of which have nothing to do with the components in your PC. Let's keep it simple and just say that your ping will primarily be affected by the location of the server you are connecting to. For example, I live in the UK. If I connect to any US server I'll get a ping on average of 100ms~. This is simply due to International Latency. We could go into much more depth and say that my ISP's peering is responsible too, but that's an in depth technical discussion and I'm perhaps not qualified enough for that, though I understand the logistics.

So, if we can agree that the ping time is caused by international latency, I'd like to know how a simple 'wonder-NIC' can decrease this.

Let's also take something else into consideration. This card states that it prioritises your gaming packets to decrease ping. To do that, the card would have to promisculously (sp?) sniff all of your network packets and decide which ones are gaming packets, and which ones aren't. This would require a great deal of work on the cards part, especially when transferring large amounts of data across a LAN. How will it know what is gaming traffic? It can't use ports, as ports will vary from game to game and client configuration. It can't possibly know every game on the planet, and recognise it's packets. Does it have some wonder technology that nobody has figured out? In addition to prioritising your gaming packets, let's face it, the other packets have to leave at some stage - thus potentially increasing your ping, so it still makes no sense.

The amount of traffic being used by an online game would NOT cause any noticeable CPU overhead on a typically average system. You really think that transferring up to 10K/s of data will bother your CPU? Not at all. Does downloading something cause your CPU usage to increase?

OK, maybe i'm being too harsh. Maybe this card has indeed made some advances in technology and can improve your pings. I can accept that there are certain things it could do in this aspect, but fundamentally I still stand by the reason that your ping has NOTHING to do with your network card or your PC. There are far too many factors to even consider this as a reality, and to me it sounds like nothing more than a marketing ploy aimed at gamers to sell a product. Unless you can prove otherwise?

At the end of the day if it can improve your ping, the difference will be so negligable that you'd really have to justify paying for this card.

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Fri, 14 Jul 2006 15:08:33 GMT

A 300W Power supply should be super fine.

(It's what we test on)

The 5-10W power spec. is very preliminary, we will have better data very soon... we are being conservative.

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

likeagod — Fri, 14 Jul 2006 12:03:01 GMT

What?! LOL... I won't tell anyone I promise -No one will see the post besides me...

Okay, so now that I've read through the .pdf I have a question for ya... It doesn't require a specific, hard answer but since I do plan on getting one of these wonderful cards (... I still want to know how I can get it as a prize LOL...)

The .pdf speaks of the power draw of the network card as 5-10 watts (that seems like a lot). I'm looking at a system that runs an AMD 64 3800+ with 1 GB DDR400 of Memory, on a ATI Crossfire ready ECS RD480-A939 motherboard using at least one Powercolor X1600 Pro 512 MB graphics card. The system will also run at least 2 Hard drives and 2 DVD drives.

What amount of power should my powersupply pull? I currently have an Antec TrueBlue 480 watt. Will that be sufficient to power this system if I add in the KillerNIC?

[ELN] likeagod - EndLagNow.ORG <---

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Fri, 14 Jul 2006 08:54:56 GMT

I could tell you but then i'd have to kill you!

Seriously, my marketing manager informs me that everything will be disclosed in due time!

The one question I can answer is that we will support VISTA when it is released.

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

likeagod — Fri, 14 Jul 2006 08:18:23 GMT

Tytus,

The other thing that might prove interesting, and I appologize if you've posted this elsewhere and I missed it, would be to see some benchmarks...

likeagod

RE: Killer NIC by Bigfoot Networks announced!!!!

likeagod — Fri, 14 Jul 2006 08:07:53 GMT

Tytus,

Since I'm currently looking at upgrading my system... any price range for this new product? And, will it run under Vista when that is released?

likeagod

RE: Killer NIC by Bigfoot Networks announced!!!!

Discrate — Fri, 14 Jul 2006 00:52:35 GMT

Yay!!!! im going to buy one!

RE: Killer NIC by Bigfoot Networks announced!!!!

likeagod — Thu, 13 Jul 2006 22:14:19 GMT

Tytus,

I had to download the update for Adobe Reader and that did the trick... For some reason I had to open Reader by itself and update it rather than just selecting the file and being asked to auto update Reader...

Now I can look at it and cry at the system requirements LOL

likeagod

RE: Killer NIC by Bigfoot Networks announced!!!!

Tytus — Thu, 13 Jul 2006 15:27:14 GMT

sorry about the .pdf.

I'm trying to make the link right-clickable...

in the meantime, almost all the information is available by clicking on the various tabs on the flash.

the only thing missing is some of the more techie specs.... (10/100/1000, etc.)

Tytus

RE: Killer NIC by Bigfoot Networks announced!!!!

likeagod — Thu, 13 Jul 2006 13:13:09 GMT

ARGH! The .pdf file for the specs crashes my browser and I can't even open it if I directly dowload it!

Must be time to reinstall adobe...

likeagod