[fRingE]

Dedicated NIC vs. Onboard LAN Test Guide

Test Steup:

[AMD] Phenom II X4 955 @ 3.2Ghz
[Biostar] TA890FXE
[CORSAIR] 850W
[AMD] 4890 @850/975
[OCZ] 4GB 1333MHZ 9-9-9-20 2T DDR3
[WD] 320GB Caviar Blue
[BFN] Killer 2100
[Netgear] WGR614 Wireless-G 54Mbps Router
[Realtek] Onboard LAN
[Belkin] CAT5e Cabling

Download, Upload and Ping Test:
Both tests were done by Speedtest.net from Randor, Ohio

Onboard LAN:

Killer 2100:

Test 1

Test 2

Ping Tests in Game:

Battlefield Bad Company 2: (Overclock.net's Server)

Onboard: 93ms-104ms

Killer 2100: 85ms-103ms

Call of Duty Black Ops: (Montreal, CA Server)

Onboard: 50ms-52ms

Killer 2100: 50ms-52ms

Call of Duty 4 Modern Warfare: (Chicago Server I Think ^_^)

Onboard: 42ms-48ms

Killer 2100: 42ms-48ms

GuildWars: (American Server North Kryta Province)

Onboard: 159ms-179ms (Also some henchman would skip like if there was network lag -_-)

Killer 2100: 67ms-94ms

Conclusion:

I am really not all that impressed with either. I get network lag skips in Guild Wars without the Killer 2100 and with the Killer 2100 I get less Upload Speed. I don't know if its my routers fault or not because its some cheap Netgear Wireless G. But I am using CAT5e Cabling so I am good on that end.

 

[justarealguy]

Too many variables to really determine what is truly at fault here. The beauty of having a dedicated NIC is that you can have multiple connections running at a time and the highest priority is given to time-sensitive data.

 

[fRingE]

Quote:

Originally Posted by justarealguy Too many variables to really determine what is truly at fault here. The beauty of having a dedicated NIC is that you can have multiple connections running at a time and the highest priority is given to time-sensitive data.

I understand but I thought it would be a fun thing to post

Hope you enjoyed the read.

 

[Razinhail]

I've always wondered about this. Thanks for the testing.

 

[Versa]

Interesting, be nice if you had Intel's onboard as well.

Again to many factors, but nice test

 

[Dirtyworks]

I believe tests like these are akin to doing single threaded tests on a multi-core system. This is barely stressing the components.. I mean, you could get these results with 802.11g wireless.

 

[fRingE]

Quote:

Originally Posted by Versa Interesting, be nice if you had Intel's onboard as well. Again to many factors, but nice test

Ya if I had a decent intel motherboard I would do a test. But I have always wondered if the Dedicated CPU on the Killer 2100 would do a better job of offloading packets than the systems CPU, and thats why I have done these tests.

Now on a older machine like a single or dual core this card would benefit a lot more because the single core/dual core design would be bottlenecked some trying to offload game packets and process the game at the same time. This card would favor in that area because it can offload all of the game packets on its onboard 400MHZ CPU without using any system resources.

 

[beers]

As stated, there are a bunch of factors out of your control with this test.

Quote:

Now on a older machine like a single or dual core this card would benefit a lot more because the single core/dual core design would be bottlenecked some trying to offload game packets and process the game at the same time.

Probably not as bad as you would think. My integrated NIC uses about 17-19% CPU when transferring and receiving ~500 mbps each at the same time. A gaming workload would be a small fraction of this. Cash from the NIC would be better spent on other components if your computer was indeed being affected with regards to CPU cycles specific to network transfers.

 

[fRingE]

Quote:

Originally Posted by beers As stated, there are a bunch of factors out of your control with this test. Probably not as bad as you would think. My integrated NIC uses about 17-19% CPU when transferring and receiving ~500 mbps each at the same time. A gaming workload would be a small fraction of this. Cash from the NIC would be better spent on other components if your computer was indeed being affected with regards to CPU cycles specific to network transfers.

Like I stated before I understand. I just wanted to give some in site to people wondering what results would be. This isn't close to being a perfect benchmark or a comparison run but it does compare some aspects versus both of the network interfaces.