Shop by Category | Brand | Order Tracking | Customer Care | Contact Us | Wishlist | My Account |
All
  • All Departments
  • Digital Download
Cart
 LATEST TOPICS |  FORUMS » NETWORKING » KILLERNIC REVIEW...
Subject: KillerNIC Review
Share this:  
PREV   |   NEXT   1 2 3 4  
Author Date Posted Tools
DaleF ♥ NCIX Mar 18, 2007 09:25 PM Reply | Bookmark
Like | Top | Bottom

Here's my long awaited review of the KillerNIC.

This message was modified by the poster at 03 20, 2007 08:13 PM

Topic URL: http://forums.ncix.com/forums/topic.php?id=1304406

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:25 PM Reply | Bookmark
Like | Top | Bottom

Here's my test setup. The computer that I did most of my testing on is based off of an Athlon64 4400+ clocked at 2.6GHz on an Asus A8N32-SLI motherboard with 2GB Mushkin Redline XP DDR500 running at 260MHz and 3-3-2-7 timings. Other hardware includes dual EVGA 7900GTOs flashed to GTXs and clocked at 725/825 in SLI, 4x320GB Seagate 7200.10 hard drives (plus a fifth that I bought during testing, connected via eSATA) and a Maxtor MaxLine III 300GB hard drive. Optical is provided by a DVR-110D and DVR-111D with an Antec TrueControl 2.0 550W PSU powering it all. Everything is in a P180 case from Antec, and cooled with Antec Tricool 120mm fans (save the CPU heatsink, an XP-120, which has a SilenX fan on it).
For network hardware, I'm using switched gigabit over Cat5e. I'm on Shaw Extreme-I (SB5101 modem) with a Netgear FWAG114 router connected to an 8-port gigabit switch. From there, I connected to on of the gigabit ports on a 24+2 100/1000 switch and then to the NIC tested.
I ran tests on three NICs. The KillerNIC, Intel Pro/1000MT and my onboard PCIe connected Marvel NIC.
I ran all of my tests in a fresh install of XP Pro with only the needed drivers and software installed (three installs total, one for each card tested). I disable the network interfaces that I wasn't going to be using for each test before I ran it.
I installed Fraps, WireShark (+WinPcap) plus the games, and that's all I can think of for software.
Games tested: Guild Wars, Counter Strike: Source, UT2004, Counter Strike 1.6, FEAR (after I finally got it working), WoW and BF2. I was going to try Rainbow Six: Vegas, but I could never get it to let me online.

This message was modified by the poster at 03 18, 2007 09:30 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:25 PM Reply | Bookmark
Like | Top | Bottom

Settings used on the KillerNIC during testing.

This message was modified by the poster at 03 20, 2007 08:14 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:25 PM Reply | Bookmark
Like | Top | Bottom

Guild Wars

Average ping with the KillerNIC (in game mode) was 179ms.
Average ping with the KillerNIC (in app mode) was 127ms.
Average ping with my Intel Pro/1000MT was 126ms.
Average ping with onboard Marvel 88E8053 PCIe connected NIC was 198ms.
However, it should be noted that the difference between the KillerNIC and the Intel isn't statistically significant.
These numbers differ from the first test I ran (and posted) because I did another two sets of tests after receiving tweaked settings for the Killer.
The Killer was configured to default settings (the tweaks made no difference when I reran the tests).

This message was modified by the poster at 03 18, 2007 09:32 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:25 PM Reply | Bookmark
Like | Top | Bottom

I tested by running through a mission (The Great Northern Wall) on my own, and recorded the mission through each time, making sure to keep my mouse over the connection indicator at the bottom left of the screen as much as possible so as to get the ping recorded. I made sure that I didn't change servers as well. Where possible, I took a sample number as close to every sixty seconds as possible.
In addition, I monitored CPU usage and no run seemed to use more or less CPU power than the others. Framerate likewise didn't change, as even with the game at max settings (at 1680x1050) with 16X TRSSAA and 16X AF enabled from the NVIDIA control panel, the game stayed pegged at 60FPS.
I suspect that App mode functioned better than Game mode because the KillerNIC is meant to accelerate UDP, but WireShark revealed that Guild Wars uses TCP.

This message was modified by the poster at 03 18, 2007 09:32 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

Counter Strike: Source

Average ping with the KillerNIC (in game mode) was 41ms (51ms).
Average ping with the KillerNIC (in app mode) was 50ms (55ms).
Average ping with my Intel Pro/1000MT was 43ms.
Average ping with onboard Marvel PCIe connected NIC was 88ms.
This test had slightly different results than my previous one. Originally, the Killer performed slightly worse than the Intel card, until Tom sent me some tweaks to the drivers, and those made quite a difference. I've included the results at stock in parenthesis. The KillerNIC does average a statistically significant 2ms better ping in my CS:S tests.

This message was modified by the poster at 03 18, 2007 09:34 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

For CS:S, I stayed on servers in Seattle as much as possible (I'm on Shaw, and Seattle is usually only a couple of fast hops), especially the server Death_Hawk setup (when there were people on it). I had Fraps take a screenshot every 60 seconds with net_graph 3 turned on. I didn't tweak any of the connection settings on CS:S, so they were all stock. I tried to play in full servers, and kept coming back to the same server to complete the set of tests, which usually meant playing a few rounds, leaving the game, coming back in and playing some more.
Framerates and CPU usage were similar for all three NICs, as my computer can easily run CS:S at max detail without much of an issue.

This message was modified by the poster at 03 18, 2007 09:33 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

Counter Strike 1.6

Average ping with the KillerNIC (in game mode) was 65ms.
Average ping with the KillerNIC (in app mode) was 69ms.
Average ping with my Intel Pro/1000MT was 66ms.
Average ping with onboard Marvel 88E8053 PCIe connected NIC was 72ms.
Curiously enough, CS 1.6 didn't seem to have much of a spread in pings, even versus the onboard. Maybe it’s just not that lag sensitive, as it is a rather old game, tracing its roots back to the days when dial-up was still big. Either way, I can't say that this test goes to either the Killer or the Intel NIC.

This message was modified by the poster at 03 18, 2007 09:35 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

Testing is similar with CS:S, except that I don't know CS 1.6 well, so I manually did the ping. No tweaks were done to any of the game’s settings. I ran the game at maximum detail settings at 1680x1050 resolution.
Framerate and CPU usage both didn't show any change, but then, CS 1.6 isn't very computer intensive.

This message was modified by the poster at 03 18, 2007 09:36 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

UT2004

Average ping with the KillerNIC (in game mode) was 128ms.
Average ping with the KillerNIC (in app mode) was 142ms.
Average ping with my Intel Pro/1000MT was 133ms.
Average ping with onboard Marvel 88E8053 PCIe connected NIC was 159ms.
Here the KillerNIC was actually around 4% faster than my Intel NIC. However, looking at how close the pings were, I think that maybe my computer wasn't having the best day with UT2004 as the onboard isn't much higher. Still, my statistics say that the Killer was faster.

This message was modified by the poster at 03 18, 2007 09:38 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

A favorite of mine, I played UT2004 quite a bit. Once again, I played on servers that are within a hop or two of (or on) Shaw Business Internet's (AKA Big Pipe's) backbone. I played Assault and Onslaught maps (my favourite game types) as well as some invasion and good old CTF. Settings were maxed.

This message was modified by the poster at 03 18, 2007 09:39 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

FEAR

Average ping with the KillerNIC (in game mode) was 58ms.
Average ping with the KillerNIC (in app mode) was 67ms.
Average ping with my Intel Pro/1000MT was 65ms.
Average ping with onboard Marvel PCIe connected NIC was 78ms.
What can I say? FEAR seems to like the KillerNIC. It was 12% faster than the Intel NIC, and this showed to be statistically significant.

This message was modified by the poster at 03 18, 2007 09:42 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

Once again, I stayed on servers as close to me (hop wise and latency wise) as possible to minimize their impact in my testing. Once again, I played FEAR with the settings maxed (and even had soft shadows turned on). With these settings, it was the only time that framerates also seemed to be better, averaging 1FPS difference between the Killer and Intel NIC and onboard. However, neither dedicated NIC seemed to do a better job of this than the other. I suspect that this is due to the game becoming CPU limited at that detail level, and hence any lessening of CPU usage by other things (like onboard networking) helped.

This message was modified by the poster at 03 18, 2007 09:43 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:26 PM Reply | Bookmark
Like | Top | Bottom

World of Warcraft

Average ping with the KillerNIC (in game mode) was 108ms.
Average ping with the KillerNIC (in app mode) was 113ms.
Average ping with my Intel Pro/1000MT was 105ms.
Average ping with onboard Marvel PCIe connected NIC was 127ms.
With WoW, I don’t actually have the game, so I installed it and had a friend who plays WoW lots, but I did supervise. The statistical difference in this one was enough to say that WoW didn’t run better on either card.

This message was modified by the poster at 03 18, 2007 09:44 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:27 PM Reply | Bookmark
Like | Top | Bottom

The game was set to maximum in game settings, and framerate was monitored using Fraps, which was also used to periodically capture the screen with the ping displayed.
In addition, I asked my friend if the game seemed smoother to him on any of the NICs, which he said all seemed faster, probably due to my computer being significantly faster than his own.

This message was modified by the poster at 03 18, 2007 09:44 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:27 PM Reply | Bookmark
Like | Top | Bottom

Battlefield 2

Average ping with the KillerNIC (in game mode) was 109ms.
Average ping with the KillerNIC (in app mode) was 112ms.
Average ping with my Intel Pro/1000MT was 100ms.
Average ping with onboard Marvel PCIe connected NIC was 127ms.
The Intel NIC was definitely faster in this test.

This message was modified by the poster at 03 18, 2007 09:45 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:27 PM Reply | Bookmark
Like | Top | Bottom

Once again, the game was set to maximum settings (and modded to run widescreen resolution). I once again chose servers to be as close to home as possible (network wise).
NO difference in framerate or CPU use was apparent.

This message was modified by the poster at 03 18, 2007 09:45 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:27 PM Reply | Bookmark
Like | Top | Bottom



This message was modified by the poster at 03 18, 2007 09:53 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:27 PM Reply | Bookmark
Like | Top | Bottom

Problems

Unfortunately, testing was not without its share of problems. The first KillerNIC refused to work in my motherboard. After the second KillerNIC arrived, I still had some issues with it dropping the net connection (but it only seemed to do this when I was running in application mode in Windows and not in any games). This, however, was quite infrequent and hard to pin down.
The other issue I ran into was that with the second KillerNIC in my computer, I got a BSOD during the setup of Vista (Ultimate 64b), and was also unable to get Vista to boot without a BSOD after I’d removed the card for installation.
In addition (this isn’t a Killer related problem), I was unable to test the NIC versus an Intel Pro/1000PT PCIe network card as I had originally planned due to circumstances.

This message was modified by the poster at 03 18, 2007 09:49 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:27 PM Reply | Bookmark
Like | Top | Bottom

Bigfoot

My experiences with the company as a whole were quite positive. They treated me well. In addition, their tech support was top notch. They actually called me to troubleshoot the issue that I was having with the first NIC.
They also FedExed out the replacement priority so that I could start the testing.

This message was modified by the poster at 03 18, 2007 09:50 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:27 PM Reply | Bookmark
Like | Top | Bottom

Conclusion

Long conclusion is coming soon.
Summarized: The KillerNIC did work as advertised in some games, most notably FEAR. However, the differences between a good Intel NIC were not very large (except in FEAR).

This message was modified by the poster at 03 18, 2007 09:53 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:28 PM Reply | Bookmark
Like | Top | Bottom

More of the conclusion. Things I'd do differently.
If I had the time, I'd try to come up with something that synchronizes keystrokes and mouse movements with the video and logs them so I could see if the KillerNIC actually does make things run smoother. Unfortunately, this was a completely subjective test on my part, and already having a fast computer and internet connection, I couldn't see any difference. But it could likely make a difference for someone on a slower connection or with a less powerful computer.

This message was modified by the poster at 03 20, 2007 08:04 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:28 PM Reply | Bookmark
Like | Top | Bottom

Quick math note: I rounded the average values to the nearest integer. Most of them didn't come out nearly that nice.

This message was modified by the poster at 03 20, 2007 08:05 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:28 PM Reply | Bookmark
Like | Top | Bottom

Note: I didn't find the differences in this test statistically significant.

This message was modified by the poster at 03 20, 2007 08:06 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:28 PM Reply | Bookmark
Like | Top | Bottom



This message was modified by the poster at 03 20, 2007 08:07 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:28 PM Reply | Bookmark
Like | Top | Bottom



This message was modified by the poster at 03 20, 2007 08:09 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:29 PM Reply | Bookmark
Like | Top | Bottom



This message was modified by the poster at 03 20, 2007 08:10 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:29 PM Reply | Bookmark
Like | Top | Bottom



This message was modified by the poster at 03 20, 2007 08:10 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:29 PM Reply | Bookmark
Like | Top | Bottom

Note: with WoW, the two cards weren't statistically different.

This message was modified by the poster at 03 20, 2007 08:10 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
DaleF ♥ NCIX Mar 18, 2007 09:29 PM Reply | Bookmark
Like | Top | Bottom



This message was modified by the poster at 03 20, 2007 08:11 PM

Tip Posting | My Postings (14745) | My Reviews (17) | My Profile | View Source
PREV   |   NEXT   1 2 3 4  




Contentname:Forums_Bottom
Call Queue
Now Serving
Closed
Avg. Response Time
Closed
Request Call Back
(Web Orders Only)

Recent Experts
1. Greg * NCIX.com (33129)
2. Saberon (15209)
3. Spork™ (13958)
4. Pawtucket (9408)
5. MasterFuji (913)
All Time Experts
1. Mr. Friendly™ (51387)
2. The Wizard (39824)
3. death_hawk™ (35144)
4. Not An Expert (26343)
5. Lance W (25246)
 
Your Order Company Info Memberships Services Hotlinks
Order Status
RMAs (Merchandise returns)
F.A.Q.
Contact Us / Store Locator
Careers
Terms & Conditions
Privacy Policy
NCIX Rewards
Premier Partner Reseller Program
VIP Memberships
Advantage Membership
Affiliate Program
Super Shipping Saver Club
Express Coverage
Express RMA
Express Shipping Service
Express Exchange and Care Coverage
Gift Cards & Balance Checker
Advanced Store Price Match
Black Friday Sale
Boxing Week Sale
NCIX Newsletters
NCIX Forums
Folding @ Home Team
The Banner Vault
International Sites:  Canada
Get the best discount computers online for online computer shopping and discount PC computer components as well as notebook computers, laptops and Canada custom computers for the best deals on computers.
NCIX.com provides Canada the best priced computers in the country for computer components and computer parts and computer accessories with the best online computer store in Canada for your personal and business computer needs.
At NCIX you can build your own computer online as well as search for the cheapest and best video cards, computer processors and CPUs, motherboards computer memory RAM, ssd and hard drives with the fastest shipping in Canada and the US.
So for discount computers online NCIX provides the cheapest digital cameras, printers, motherboards, flash memory, PC speakers and computer electronics with the best Canadian computer and Canadian online computer shopping website in the world.

NCIX.com is powered by Intel® Xeon® Quad Core Processors
Copyright© 2014 NCIX.com/Netlink Computer Inc. All rights reserved. S / Server ID:74 20141022234712  360 ms