Are all NIC created equal?

[MultimediaMan]

The results might surprise you...

http://www.dslwebserver.com/nictest.html

Check it out.

[SteveC]

Cool article - thanks for the find!

------------------
Cheers,
Steve

"Life is what we make of it, yet most of us just fake"

[Bohrn]

Great stuff! Thanx. I was contemplating upgrading this NIC I have had since '94! Hehe! And lo-and behold, it is on your poo-poo list! Thanx again

------------------
AsusP3B-F,P3 880 Slot1, 512megsPC133,G40032megSH, Diamond MX300,13.5gigs of HD's,52X CDrom,WinME, PD 6.21 ,Altec Lansing ATP3Subwoofer,
Envision 17",Terrayon Cable modem w/D-Link nic,1 grey cat,1 black cat & 1 calico

[spoogenet]

Interesting, though. Isn't a megabit defined as 10^6 bits, which would mean 10 Mbps is NOT 1.25 MB/s? It would rather be 10,000,000 bits * (1 byte/8 bits) * (1 KB/1024 bytes) * (1 MB/ 1024 KB) = 1.1921 MB/s, no?

Which makes me wonder....are the other numbers valid???

b

[MultimediaMan]

I think you're right, but on a Base 2 Numering system 1000 IS 1024....so it works out to 1.25 MB/Sec...Not that it really matters.

[dZeus]

I doubt that is a good test... I just don't trust it.

I've tested some card myself, and came up with _completely_ different results than the author of that article did.

[RichL]

Thats a good article, thanks for the link.
Personally I've encountered numerous reliability and driver stability issues with Dlink and SMC cards over the years, while 3Coms and Intels never seem to give me any trouble.
http://forums.murc.ws/ubb/Forum3/HTML/003626.html
3Com, Intel or Bay Netgear will all do the job nicely. Personally I wouldnt touch a DLink or SMC card with someone elses 10 foot barge pole,

Ahh, smug mode.. :-)

Rich

[spoogenet]

MultimediaMan: A kilobyte is defined as 2^10 bytes (1024 bytes), while a megabyte is defined as 2^20 bytes (1048576 bytes). A megabit is defined as 10^6 bits, and a byte is defined (for most people) as 8 bits. This often causes confusion among people who do not realize/understand that all numbers in computing are not rooted in base 2. Hard drives, for instance, are sold on the base 10 system. Common FAQ question: I bought a XX GB hard drive, but Windows reports it as being far less than XX GB, is it defective?

By using the incorrect number a 5% error is introduced to the theoretical bandwidth. This of course propagates to the calculations of how the measured bandwidth stacks up to the theoretical bandwidth. I would consider this a significant error.

My issue is mostly one of principle. I cannot trust the validity of a test when the tester exhibits a lack of understanding for the subject that he/she is testing. I especailly cannot trust it when the lack of understanding causes a noticeable change to the accuracy of the results.

b

[dZeus]

hmmmm... that's weird... I just re-run the test I did some time ago, and this time the results are quite different.... hmmm

this time I got about 80% CPU load with a Celeron 333 at 7-8MBps... I swear that last time I tested I got about 40% CPU load with 9MBps... (all from ram to ram)... this is/was on 2 3c905c-tx cards...

[Dr Mordrid]

I've had great luck with the Linksys LNE100TX cards. Cheap, fast and best of all they don't seem to interfere with editing rigs like some others do (SMC, DLink etc.).

Dr. Mordrid

[Jammrock]

I want to see some tests using a 3XP based 3Com NIC. The 3XP will unload most of the network load off the CPU and does it on the NIC. Maybe I can test it once I get one...someday.

Jammrock

[Electric Amish]

Hmmmm... Using the drivers that came with the cards instead of the latest... I don't know...

I've been using Realtek cards forever. Made by Ovislink or DLink. Seem to work fine.

amish

[This message has been edited by Electric Amish (edited 05 April 2001).]

[spazm_1999]

It would had been fun to see a little more 10mbs card in the test. I've got a Realtek 8029(as) (10mbs) and i'm getting a normal transfer rate of 1115K/s on my internal ftp so i guess that card is really good.

Spazm

[Salmonius]

I have used many cards at my previous work to build linux boxes with multiple adapters, and I can say that my favorite were the D-Link 530-TX. The other ones, especially the intel ones, were too much trouble.

There was one thing funny though. One of the 530TX was different than the other ones. Even Linux reported VIA-Rhine instead of RT-1839 as usual for that type of card. I wonder if D-Link changed the 530-TX at some point. I know that now they have the 538.

By the way, I have a 486 with a Matrox Millennium I and a 530-TX (it's got a PCI bus) and it runs Mandrake Linux 7.2 perfectly.

[MultimediaMan]

DZeus: Did you use the same methodology? When it comes to 10/100 transfers, your mass storage devices tend to be the bottleneck.

Remember, he was using a RAMDisk, which takes the HDDs out of the equation. He also specified what type of a file he was transferring (A Zipped folder of MP3s), which is pretty much a worst-case scenario for file transfers.

Also, he also specified which drivers he was using.

Not a bad way to test overall.

I was wondering if he was running the Win98SE TCP/IP Patch, and whether or not he was running with NetBIOS on or off.