Local Area Network Technology: Part 3 - Network Interface Cards (NICs)

by Jeff Carrell, Electronic Communications Chairman [reprinted from IEEE - Fort Worth Section Signals June/July 1997]

Continuing our discussion of OSI Physical Layer Electronics (Part 1 was cabling, part 2 discussed hubs) this article addresses the "host" device connection point, or Network Interface Cards (NICs). Occasionally, the terms LAN Adapter Card or network adapter card are used, but the preferred term is NIC.

In order to connect a device to the network, it must contain a network interface. For the most part, we call any device that has a direct network connection a "host." (The use of host in this particular context does not refer to the mainframe or minicomputers, which have been traditionally called host computers.)

Today, we have many different types of hosts: PCs, Macintosh's, UNIX systems, notebook computers, minicomputers, mainframes, printers, communications systems, routers, etc. All of these devices, no matter their function or physical design, will have a NIC installed.

The I/O connector of the host can be one of many different types of buses, for example: ISA, EISA, MCA, PCI, PC Card, or proprietary. The different bus types also have different I/O data paths (i.e., how many bits the bus moves at a time). Traditionally, the ISA bus is an 8 or 16 bit bus, EISA is 32 bits, MCA is 16 or 32 bits, and PCI is 32 bits.

Ethernet and token ring NIC options

In the realm of Ethernet NICs, there are many different options to choose from today. And, in addition to connector styles, there is the speed issue to consider. Is the network running 10Mb/s, 100Mb/s or even (soon) 1000Mb/s? Also, what about the upgrade path? A lot of networks today run at 10Mb/s, but the cost of a 10/100 NIC is usually only a little bit more. That being the case, it is best to go ahead and install 10/100 NICs now, to facilitate a later upgrade or the swapping of hosts for servicing/redeployment. Since Ethernet has been around for longer than forever (in networking time anyway), there are three different connector/cable types to be supported. The original AUI or DB15 connector for Thicknet cable, the BNC coax connector for Thinnet and the RJ45 supporting UTP/STP modern day cable plants. NICs can come equipped with: any one of the three specific connectors, two connectors (usually the AUI and one other), or a combo card of all three. Obviously, an Ethernet NIC that is a dual speed with combo connectors can support virtually any network and cable plant. Of course, the more options the NIC has to offer, the higher the price.

Token Ring NICs have similar issues. There are two speeds of Token Rings - 4Mb/s and 16Mb/s. Although the original Token Ring NICs were 4Mb/s only, the current day Token Ring NICs are usually dual 4/16. Likewise, early day Token Ring cables terminated at the NIC via a DB9 connector. However with the Cat5 structured cable systems of the 90's, vendors have been putting RJ45 connectors on most Token Ring NICs, and some come with both connector styles.


Each NIC has a unique hardware level address, called the MAC address. This address is "burned-in" by the vendor, and usually cannot be changed. The MAC address is a 12 digit, hexadecimal number divided into two components, a prefix (first 6 digits) and suffix. The prefix number is assigned by the IEEE. During the manufacturing process the vendor takes their specific prefix and adds a suffix - sequentially. This allows for all NICs to have an individual 12 digit address. If a vendor runs out of suffix addresses, they apply for a new prefix and keep on going. (Imagine the problems if two or more NICs on the same network had the same MAC address.)

Installing NICs can be easy or very troublesome depending on the host platform type, bus type, operating system and network drivers. Some points to ease the pain: buy only name brand NICs (no "clones"); verify the parameters of the host before buying NICs (are they all compatible with each other); once you find a NIC that works in your environment, stay with that brand; and be prepared during the first installation (have all drivers, documentation and technical support numbers handy).

NIC is a very integral component in the network. I don't recommend saving a few dollars and installing lessor quality NICs - I have seen too many intermittent problems with them. When NICs go bad, they can cause all kinds of problems many of which are hard to troubleshoot. There are many diagnostic products available and it always helps to have accurate documentation of the network and the host types, NIC types, and driver software configurations available. Often, the root problem with a NIC is a simple software driver configuration error.


Copyright © 1997 Jeffrey L. Carrell All Rights Reserved