This roundup is here to give some answers to the dubious question "Which is the best network card".  What you'll find may surprise you.  Or maybe it'll substantiate what you already believed.

I decided to compare only network cards that were capable of 10/100 Mbps.  The reason I chose to do this is because the price of network cards has dropped so low that there isn't really any point to purchasing a card that can only run at 10 Mbps.  For a few bucks more, you'll have a card that is theoretically 10 times faster.

Background:
So you see the numbers 10 Mbps and 100 Mbps, but do you really know what these numbers mean? Basically it means the card will run up to 10 or 100 Megabits per second.  What the heck is a megabit?  Now for a little math.  A bit is 1/8th of a byte.  Or, a byte is 8 bits. 

To transform megabits per second to megabytes per second, all you have to do is divide by 8.  Here is the speed that a 10 Mbps and a 100 Mbps network card should theoretically reach:
 

But alas, you will never actually transfer a file at those blazing speeds.  This is because your network has additional traffic depending on the network protocol (TCP/IP, Netbeui, IPX, etc.).  There is also a certain amount of negotiation / acknowledgment that must occur between different network cards.  Let's eavesdrop on two network cards that transferring files to each other.

NIC 1: Pay attention, a piece of a file is coming your way!
NIC 2: Okay, send it.
NIC 1: Sending.
NIC 2: Okay, got it.
NIC 1: Next piece of data coming through!
NIC 2: Go for it.

The contestants:

I picked the network cards that are currently most popular and also widely accessible to mainstream consumers.  (Translation: I went to Best Buy, CompUSA, Fry's, etc.).  I also hit several other computer shops looking for NIC's.  The network card that took the longest to find (surprisingly) was the generic network card.  There weren't any computer shows nearby so I finally found one from a small mom and pop shop around the corner. I also tested a couple of NIC's that I already own.

Click to Enlarge

The price that I paid is for reference only.  For the most part, I paid too much.  If you are planning to drive down to the Best Buy and pick up a card, you'll probably pay around what I paid.  However, if you do a bit of searching on the Internet or look for sales, you should be able to get a better price.  In the end, it depends on the convenience of getting the card right away versus having to wait for a package to ship if you bought it on the Internet.

Here are the cards tested:
 

Methodology:

The first thing you should know about network cards is that they are rarely the limiting factor.  The biggest hurdle in your system is your hard disk.  Even though network cards can be saturated by  newer high-end hard disks, most regular hard disks can't, especially when transferring several small files.  The seek and read times of the hard disks are just too slow.  In order to get around this hurdle, I moved to the other white meat: RAM.  RAM is thousands of times faster than hard disks, so this is how I configured my test systems.

Server:  AMD K6-III 333 MHz

This was the system that had the network card that would never change.  This server was running Windows 2000 Advanced Server with IIS 5.0 FTP server.  I created a 200 megabyte RAM disk on the server using RAMdiskNT.  Let me tell you, this is a cool product.  It's shareware and is limited to 100 uses without registering.  If you choose to keep using the product, the cost is $35.  This program allows you to create a disk drive from RAM.  You can specify the size of the RAMdrive using the included applet.  I didn't use the RAMdrive that comes with Windows because that program only allows up to 32 megs of RAM to be used as a RAMdrive.  By using RAMdiskNT, I could be sure that the hard drive would not be the bottleneck.  The server did not have any other application running on it and was disconnected from the Internet so there was no Internet traffic.  (Sorry guys, yes, that's why this site was down for a while - I sacrificed it for NIC testing!)

The network card in the server is a Linksys Etherfast 10/100 v.1.0 using the drivers that came native with Windows 2000.  You may wonder if this card might be a limiting factor in my testing.  Let me tell you, I'm confident it was not.  During the course of testing, I found that when a network card  is receiving data at full speed, there was often 100% cpu utilization.  However, the network card that is sending the data usually uses much less cpu time.  While sending files, the Linksys card in the server never exceeded 40% cpu utilization which tells me that there was still plenty of juice left in the card and that it wasn't the limiting reagent.

Client: Intel Celeron 366@550Mhz

CPU: Intel Celeron 366@550Mhz
Motherboard: Tyan Tiger 100
RAM: 256 Kingston PC100 CAS 2

Actually, this machine is a dual cpu machine, but since I ran Windows 98 SE on it, only one of the cpu's was actually being used.  I chose to use Windows 98 SE on the client machine because it's the operating system most people are using these days.  Windows 2000 is very popular with the computer savvy group of techies, but most consumers and gamers still use Windows 98 SE.

This client also had a RAMdrive of 200 megabytes using the Windows 98 version of the aforementioned software: RAMdisk98 (Registration is $25 versus RAMdiskNT's price of $35). 

The only software running on the client computer was WS FTP, a file transfer program.  After I had configured the client computer without any network adapter or network drivers, I made a clone of the hard disk using Norton Ghost.  After each network card was finished testing, I restored the fresh image of the hard disk to the client .  This was to insure that the network drivers of one network card would not interfere with subsequent network cards.  Time consuming, but necessary.

The server and client were connected to a Allied Telesyn LANedge 8 port 10/100 Switch which was isolated from the rest of the network and disconnected from the Internet.

The test consisted of transferring a 200 megabyte file from the server to the client - from the RAMdrive of the server to the RAMdrive of the client.  The 200 megabyte file is made of MP3's of Britney Spears songs zipped into one large file.  (You got a problem with this?)  This was to make a 200 megabyte file that is nearly incompressible.

For each network card, I installed the card using the drivers included in the box, regardless if Windows 98 SE already came with native drivers.  I then ran 3 practice runs to get the server and clients ready.  The 200 megabyte test file was then transferred 20 in each direction (server to client, client to server).  WS FTP keeps track of how long it takes to transfer the file which I wrote down.  The fastest two 2 times and slowest 2 times were thrown out and the rest were then averaged.

With 20 transfers in each direction, this testing took the better part of a week.

Results:

I calculated the speed of each network card by dividing the size of the files by the time it took to transfer the file in seconds then multiplied the whole thing by 8 (to convert bytes to bits).  The speeds you see here are the actual speeds the file was transferred, minus the overhead and extraneous network jabber.  Simply stated, the speeds represent actual data transfer and does not include network overhead.

Here are the speeds in Mbits/second.
 

The first thing I must say is that these speeds are pretty fast.  Remember, this is the speed of the file transfer without the overhead.  Add in the overhead and you're pretty close to the theoretical 100Mbits/sec.  The winner in this round is the Netgear FA312TX.  You can see that the scores are very close to each other except for the D-link DFE 530TX+ and the Trendnet - Realtek RLT8139.  These two cards bring up the tail end of this list.  Another thing to notice is that Intel and 3com aren't at the top of the list.  However, since the speed differences are so small, this isn't really all that important.

To get a better idea of how these numbers translate into working numbers, here is the same data presented as megabytes/second.
 

These numbers look even more impressive as megabytes/second.  Remember, these numbers were obtained in close-to-ideal conditions. Real world numbers probably won't look like this at all.  Like I said before, the network card is rarely the bottleneck.  I have several hard disks that can't even sustain  5 megabyte per second transfers.  All the numbers look pretty good except for the D-Link and Trendnet.  One thing you should note is that even the slowest card, the Trendnet - Realtek card was able to manage a speed of 3.08 megabytes per second.  In comparison, this is at least 3 times faster than a 10Mbit network card can muster.  There is a huge difference between 100Mbit and 10 Mbit and unless you have really good reasons, you should get 100Mbit equipment.

The other aspect of these network cards that I examined was the CPU utilization when downloading at full speed.  Like I said previously, these network cards utilize different amount of cpu time when downloading versus uploading.  When upload files, the network cards rarely broke past 50% utilization.  However, when downloading files, the cpu utilization jumped tremendously.  I don't know why this is the case.  I have a theory that the network card that is downloading files has to process packet information which is slower than simply sending files when uploading.  The numbers here are the cpu utilization when the network card is receiving files.  I couldn't find a good way to measure the cpu utilization when the network card was sending files because the graph for cpu utilization would fluctuate so much that I couldn't get a single number pinned down.
 

Strangely enough, the network card that has the worst performance had the lowest cpu utilization.  This is because the network card wasn't able to work fast enough to stress the cpu.  The rest of the cards all used 100 cpu utilization when downloading at full speed, except the Intel Pro/100+ Management Adapter.  This network card only used 77% cpu utilization at full speed.  This would be very useful for a computer that needed to multitask heavily while downloading files from a file server.  However, this Intel card is also the most expensive card in the roundup.  In this particular case, you get what you pay for.

I know I will get the question: "On my computer, when I transfer files, I NEVER hit 100% cpu utilization.  You must be messing something up!"  The reason is simple.  On my network, I never hit 100% either. In a real working network there is always something that will be the bottleneck, and that something is rarely the network card.  In these tests, we were able to max out the network cards because we have them in an ideal situation where we have eliminated the hard disk bottleneck and all other traffic on the network.  Actually, it turns out that by decreasing the network speed slightly (introduce network traffic, allow hard disk access instead of RAM access) the cpu drops utilization a lot.  The highest I've seen cpu utilization run on a typical network is about 50%.  The cpu utilization is not linear, but instead increases significantly at higher transfer speeds.

My impression is that all the network cards did fairly well.  The transfer speeds were fairly consistent among most of the network cards although there were some that didn't do as well. 

One thing we showed is the big names (i.e. Intel and 3com) don't own the title to the fastest network cards.  They were solid performers, but they aren't the fastest.  Netgear was impressive with the network speeds and placed all of their cards in the top half of the pack.  The Linksys card also did very well.

Real problems with Realtek

Much like the video card business, oftentimes one company will create a chip or controller and then other companies will create a product based upon that controller.  In the network card business,  Realtek is the big fish when it comes to making network card controllers.  In fact, they have about a 70% market share. 

For 10/100 Mbps network cards, the Realtek 8139 controller is practically everywhere.  They are usually sold as generic cards, but also appear in several name brand network cards.  If you have a motherboard with an integrated 10/100 network card, chances are that it's based upon the Realtek 8139.   Here are the cards in the roundup that definitely use the Realtek 8139 controller:

• D-Link DFE 530TX+ 
• SMC EZ Card 10/100 PCI (SMC1211 Series)
• Trendnet - Realtek RTL8139

As you can see, there is a huge difference between these 3 cards.  In the best case (SMC EZ Card 10/100 PCI - SMC1211 Series), the card is on par with the other high performing network cards.  In the worst case (Trendnet - Realtek RTL8139), the card was more than 2.5x slower than the competition.  This is extremely strange for network cards based upon the same Ethernet controller.

In order to see where the problem was, I tried many different drivers for each card.  I tried the boxed drivers, updated drivers from the manufacturer's website, and the generic drivers from Realtek.  I even tried using drivers from one company for the other company's card.  Nothing made an improvement in the results and some of the drivers caused a further degradation in performance.

When I physically examined the the Realtek based cards, it was apparent that each manufacturer designed the network card differently.  The Trendnet card was by far the smallest network card I have ever seen with a PCB dimension of about 4 inches by 1 inch.  That is really small.  The D-link and SMC cards were twice as big as the Trendnet card.  All I'm saying is that there are design differences even though they use the same Realtek controller.

(Brian anticipating next question):  But I use a network card which uses the Realtek 8139 and I don't notice any speed problems like your article shows!  What's the deal?

I believe that there are three possibilities to why the Realtek based cards have such disparate numbers.

Possibility 1:
Realtek designed a perfectly good controller, but the network card manufacturers decided to include, or not include some of the capabilities of the Realtek chipset.  It is very likely that the faster Realtek based network cards use all the available capacities of the controller to squeeze out the maximum possible speed, while the generic card just used enough of the controller's capabilities to simply function, but not perform extremely well.  This happens in the computer sector all the time.  For example, the BX chipset by Intel is capable of running dual cpu's but many of us have motherboards that only use one cpu, leaving the dual capability of the BX chipset unused.  This also happens in the graphic chipset market where companies will "neuter" their fastest products and then sell them at lower speeds and lower prices. 

Possibility 2:
It may be that the Realtek based cards just doesn't get along well with my hardware, OS, etc.  I was only able to test the network cards on one computer and it's possible that there happens to be a conflict with my Tyan Tiger 100.  Sometimes hardware just doesn't like other pieces of hardware.  If this is the case,  this isn't a very satisfactory answer since as a consumer, I need to have the the product fit my needs, I can't always have the perfect setup for the product to work optimally.  (I'm not going to change my motherboard just to get the network card to function correctly)  In my case, I have a Tyan Tiger 100 which is based on the Intel BX chipset which is a very solid platform.  If I were using a less supported VIA chipset, then the results would be more questionable.

Possibility 3:
The result are correct, however, since they were obtained under near ideal conditions, the difference between the network cards are more pronounced.  Under real world conditions, you may not notice the difference between a 3 MB/sec transfer versus a 8 MB/sec transfer since there are so many other possible confounders like hard disk speed, applications, or network traffic.  Aside from the testing, I also used the network cards for bit on the Internet, playing network games, and transferring files to see how they "felt" in real world conditions.  Quite honestly, I couldn't tell much of a difference between most of the cards just because there are so many other confounding factors that influence network speed than just the network cards.

The truth most likely is a combination of !, 2, and 3.  How much each factor contributes is unknown.

Because of the wide variability in performance, I cannot recommend purchasing any network card based upon the Realtek 8139 controller.  For about the same price or even a few dollars more, you can have a network card that won't make you guess if you have a good incarnation of the Realtek 8139 or if it the card will be compatible with your existing hardware.

Breakdown

I have broken down the tested network cards into 3 recommendation categories for easy comparison.

Group 1. Fast and Inexpensive (Recommended)

• Linksys LNE100TX Fast Ethernet Adapter
• Netgear FA310TX
• Netgear FA311TX
• Netgear FA312TX
• SOHOware Fast - SFA110A

• 3com OfficeConnect NIC 3CSOHO100-TX
• Intel InBusiness 10/100 Network Adapter
• Intel Pro /100+ Management Adapter

• SMC EZ Card 10/100 PCI (SMC1211 Series)
• D-Link DFE 530TX+
• Trendnet - Realtek RTL8139

Conclusions:

Network cards have come along way since their introduction many years ago.  All the network cards in this roundup were plug-and-play and extremely easy to install.  The computer detected all the cards and all you had to do was pop in the right drivers and you were off and running.  It is so great to have jumperless cards where you don't have to mess around with IRQ's and other inane settings.  This is good news for all computer users as this makes networking much more user friendly.

Which card should you choose?  You have the results in front of you and which card you choose will be determined by 1. price 2. driver support for your operating system.  The speeds of most of these cards are so close that you will most likely not notice any difference between any group 1 or 2 cards.  However, you may notice speed differences with certain group 3 cards.

One thing we cannot test in this roundup is reliability.  All I can say is that all the cards worked pretty much as promised and none of them died on me.  Actually, I've never had any computer component die on me.  Regarding computer peripherals that don't have any moving parts, once a product works, it usually keeps on working for a long time.  However, if you must based your purchase on reputation of reliability, then 3com and Intel are your main choices, but keep in mind that Netgear and Linksys have also built their reputation on making very good products.
 

Take Me Home!