The Legend of the Novell NE2000 [video]

> Some call it a legend, and others consider it the greatest network card ever made. The story of the NE2000 begins with National Semiconductor's DP839x chipset, and a company called Novell. ## Starting with "NE2000 Compatible" If you touched computers from the late 80s through the 90s, you almost certainly heard of the NE2000. The first time I remember seeing the term was in the pages of *Computer Shopper* magazine. On weekends, my dad and I would go to the local bookstore, and I always looked forward to getting the latest issue. Once I got it home, I might spend hours poring over the various ads and articles. And one phrase I always remembered seeing was "NE2000 compatible." Not just there, either—in manuals and even operating systems as well. But at the time, I had no real idea what it meant. I didn't actually know what an NE2000 was, only that it was some mysterious network card that everything seemed to want to be compatible with. But looking at it now, it seems just like any other pre-10BASE-T network card. So what was so special about it? Well, it all started with these chips: National Semiconductor's DP8390, 8391, and 8392. ## The Early Evolution of Ethernet Controllers We've covered a lot of Ethernet history on this channel, but we haven't really touched on the evolution of the original Ethernet controllers. The first commercial Ethernet products weren't aimed at desktop PCs—they were for systems like minicomputers, and they were huge. These boards collectively made up the Ethernet card for the DEC PDP-11: - A link module, connecting to the AUI transceiver - A port module, connecting to DEC's Unibus architecture Obviously, these boards were enormous, packed with lots of discrete logic and central controller chips. Shrinking these down to a size that could fit into the original IBM PC would require consolidating most of that. And this was the result: the **Seeq 8001**, the industry's first Ethernet chipset, replacing over 60 discrete components with a single package. It was designed by Dado Banatau and released in 1982. But it wasn't until Bob Metcalfe and Ron Crane's 3Com came along that it was integrated into a card for the IBM PC. 3Com introduced the original EtherLink, part number **3C500**. This was the world's first Ethernet controller for the IBM PC. While it was by no means a small card, compared to DEC's boards, it was positively compact. This card combined the Seeq chipset with 3Com's own ASIC, allowing the first IBM PCs to connect to Ethernet networks. Other chip manufacturers quickly followed in Seeq's footsteps: Fujitsu, Intel, and later AMD, all released their own Ethernet chipsets. But as you might imagine, being the first PC network card, the 3C500 wasn't very fast. It couldn't actually reach the 10 megabit per second limit of Ethernet at the time. Though that wasn't a huge problem, because when you're used to kilobit-per-second transfer rates, even approaching 1 megabit per second was revolutionary. The Intel 82586 and later the AMD AM7990 LANCE were more capable and more integrated than the Seeq chip, and they supported **bus mastering**, meaning they could move data directly from the controller into host memory without requiring the host CPU's involvement. These chips were popular in workstations of the day, which were designed and built with onboard Ethernet. But PC network cards using these chips were very expensive. The XOS 205T, for example, retailed for over 900 dollars—about 2700 dollars in today's 2026 money. So what Ethernet really needed was a cheaper, higher-performance chipset to help bring it into the mainstream. ## Cheaper Ethernet: 10BASE2 The original Ethernet standard used thick coaxial cable, paired with an external transceiver, connecting to the network card through an AUI interface. The IEEE standardized this as 10BASE5. Because it was expensive and difficult to handle, the IEEE committee almost immediately began work on standardizing a new physical layer called 10BASE2—or as it was affectionately known, "Cheapernet." Manufacturers from every level of the computing ecosystem wanted a piece of it, and one such company was National Semiconductor. ## National Semiconductor: The Birth of a Chipset National was one of the largest semiconductor manufacturers in the early 80s, the first of its kind to reach 1 billion dollars in annual sales. Based in Santa Clara, California, the company was already well known for chips like the 8250 UART—a groundbreaking design used in the original IBM PC—and they immediately set their sights on Ethernet. Ramanatha Balakrishnan, known to everyone as Balu, joined the company in 1978, working in the interface design group under Charles Carusillo. Balakrishnan was a member of the IEEE 802.3 committee and had been developing a new transceiver and serial network interface for CheaperNet, called the **DP8391 and 8392**, respectively. The beauty of these new chips was that they achieved a high degree of integration, meaning they needed very few external components to function. The coaxial transceiver, for instance, only needed a single resistor to achieve its required functionality. Balakrishnan's chips handled communication on the physical Ethernet interface, but integrating with the host would require a third chip. That was the **DP8390**. It was architecturally designed by Herb Schneider, who joined National in 1981, working as an applications engineer in Mike Evans' group. Rather than directly bus-mastering host memory like the AMD and Intel chips, the 8390's design used its own local buffer memory to receive and transmit multiple Ethernet frames without any intervention from the host. The 8390 supported both 8-bit and 16-bit modes, and required very little glue logic—that is, the supporting chips between the Ethernet controller and the host system. All of this meant that the complete chipset interface, including 4 KB of local buffer memory, took up only about 15 square inches of board space. In volume production, it cost OEMs around 100 dollars—far less than the cost of implementing something like the Intel 82586. The Intel 82586 was unquestionably technically superior, but it was also far more complex and expensive. And for one company, this new chipset couldn't have come at a better time. ## Enter Novell If there's one name that defined PC networking in the 80s, it was Novell. Originally called Novell Data Systems, the company was founded in Utah in 1980, and it actually started out in hardware. They designed and built complete computer systems, and even a printer. However, they struggled to find success in a crowded market; their headcount grew to 150 employees, but as difficulties mounted, it shrank to just 10. It was around this time that four friends from Brigham Young University came together to form SuperSet Software. And they began providing Novell with...... Source: The Legend of the Novell NE2000 - YouTube (https://www.youtube.com/watch?v=nNXzQ7V1S_k)