David vs. Goliath: The 1980s CPU Battle Between NEC and Intel
How a Japanese manufacturer took Intel’s 8086 & 8088, made them better, and sparked a legal war that would define an industry. (Part 1)
The 1980s witnessed a surprising battle between Intel, one of today’s largest processor manufacturers, and NEC, the Japanese electronics giant. It was one of the most well-known tech sagas of the era, its technical and legal dramas discussed and recounted many times. But I want to tell this story from my own perspective, in two parts.
At the heart of this battle lay Intel’s 16-bit 8086 and 8088 processors, launched in 1978 and 1979. The 8088 was the chip IBM famously used in its first “PC,” the 5150. I’ve covered the features of these two processors in other articles (such as “Legendary CPUs of the home-computing era” and “Understanding ‘Bits’ in CPUs”) so I won’t list them again here. I’ve also written briefly about the “second-source” production model that set the stage for this fight, but it’s worth a recap. I promise I’ll keep it short.
The “Second-Source” Model
The second-source model is something we’re not used to in today’s CPU world. Intel itself hasn’t used it since the 80386 (released in 1985), but it was very common in the 1970s and 80s.
Consider this: Let’s say you have a hit product (like Intel’s 8088) and a massive customer (like IBM) who guarantees they’ll buy millions of units. Things go great the first year, but in the second year, you go bankrupt. You’re ruined, but you’ve also disrupted your customer’s entire supply chain.
To prevent this scenario, large companies (and especially governments) would require you to find a backup manufacturer for your product before they would sign a contract. You’d find a second (or third, or fourth) manufacturer, negotiate a deal, and transfer the intellectual property: the production masks, the microcode, and the rights to produce your chip. As seen in the Intel-AMD example, these second-source companies could even be your fierce competitors.
The 8088 has one of the longest lists of second-source agreements in microprocessor history. A quick search shows Intel licensed it to AMD, NEC, Fujitsu, Harris, OKI, Siemens, Texas Instruments, and Mitsubishi.
The “Upgrade”
Our story begins with Intel asking NEC to be a second-source supplier for the 8086 and 8088. The two companies shook hands. The processor masks and microcode provided by Intel reached Japan (albeit with some delay), and production began.
So far, so good.
But everything changed when NEC decided to use this information not just to produce Intel’s chips, but to launch its own competing product.
While the 8086/8088 were revolutionary, by the early 80s, their weaknesses were well-known. NEC was aware of them, too, and set out to address them using Intel’s own design as a base. Instead of just copying the product, they tried to create something BETTER.
They developed the V20 (to compete with the 8088) and the V30 (to compete with the 8086). Here’s what they did.
1. Production Technology
Today, almost all microprocessors are built with CMOS technology. But when Intel produced the 8086 and 8088, it was still using the older NMOS technology, which consumed far more energy and, consequently, ran much hotter.
NEC opted for the relatively new CMOS process for the V20 and V30. The resulting chips required significantly less power and could operate perfectly without any cooling measures — not even a passive heatsink.
2. A Dual Internal Data Bus
The 8086 and 8088 both had a single internal 16-bit data bus. NEC realized that by adding a second 16-bit bus, the processor could handle internal data exchanges in parallel instead of serial, speeding up operations. That’s exactly what they did.
3. An Extended Instruction Set
After the 8086/8088, Intel didn’t stop. It released the the next two generations of x86 family: 80186/80188 and 80286 (both in 1982). While the 80186/80188 were popular as microcontrollers, they weren’t widely used in PCs and, most manufacturers, led by IBM, eventually jumped from the 8086/8088 straight to the 80286 as seen in IBM AT.
NEC chose to enhance the 8086/8088 instruction set.
- First, they took all the new commands from Intel’s 80186/80188 (for I/O, stack manipulation, etc.) and built them into the V20/V30.
- Then, they added several Real Mode commands from the new 80286.
- But they didn’t stop there. NEC also added instructions for Packed BCD arithmetic and numerous bit/nibble operations, which they felt were missing from the x86 family.
So, the NEC V-series didn’t just have the 8086/8088 instruction set; it had commands from the next two Intel generations.
4. 8080 Hardware Emulation
One of the biggest criticisms of the 8086/8088 was their lack of backward compatibility with the 8-bit 8080/8085 processors. This meant Intel was deliberately turning its back on a vast software portfolio and the popular CP/M operating system. Companies with heavy investments in 8080-based hardware and CP/M software were reluctant to switch to a new architecture.
In my opinion, this is where NEC scored its biggest goal.
The V20/V30 instruction set was designed to be a superset of the Intel 8080. With a single BRKEM command, the chip would switch into a hardware emulation mode for the 8080. The RETEM command would instantly switch it back to x86 mode when issued. This meant a PC with a V20 or V30 processor could run all 8-bit 8080 software and all 16-bit x86 software like MS-DOS.
5. Hardware Effective Address (EA) Generation
Look, I’ll say this with confidence: Gather a hundred people who were active in computing in the 80s and ask them which processor had the best memory addressing. You’ll get answers ranging from the Motorola 6800, 6809 or 68000 to the MOS 6502 or Zilog Z80 but you will not find anyone defending Intel’s segmented memory model.
That’s a topic for another article, but here’s the problem: The 8086/8088 used 16-bit registers to address 20 bits of memory (1MB). This required a complex calculation, and the processor had no dedicated hardware to do it. The calculation had to be run on the general-purpose ALU (Arithmetic Logic Unit), creating a significant bottleneck and wasting many clock cycles.
NEC’s solution? They added dedicated hardware logic, an Effective Address Generator, to the V20 and V30 that could perform this 20-bit address calculation in a single cycle. This alone made many instructions run nearly twice as fast as on the Intel chips.
The Drop-in Replacement
As a result of all these improvements, the NEC V20/V30 had 63,000 transistors, more than double the 29,000 in the 8086/8088.
NEC’s V-series was compatible with the 8086/8088, the 80186, and (to a lesser extent) the real mode of 80286. It even appealed to 8080 users because of the backward compatibility Intel has so annoyingly neglected. It consumed less power, ran cooler, and delivered an average of 30% higher performance at the same clock speed. To top it all off, NEC sold them for less than the Intel chips.
You might think using these new processors with their numerous improved features would require new chipsets, new motherboards, or new RAM. You’d be looking for the catch. So, how much investment was needed to upgrade an 8088-based system to an NEC V20?
None.
The NEC V20 and V30 were pin-compatible with the Intel 8088 and 8086, respectively. They were true “drop-in replacements.” To upgrade your PC, all you had to do was pry the Intel chip from its socket with an IC extractor (or your dirty fingers) and press the new NEC chip into its place. It was a 10-minute job even for an unexperienced user.
(Note: It’s a common misconception that the 8086 and 8088 are pin-compatible but they are not. The 8086 has a 16-bit external data bus, while the 8088 has an 8-bit one. This is why you must replace an 8088 with a V20, and an 8086 with a V30.)
The Threat and the Coming Storm
I’m sure you can guess what happened next.
Intel immediately realized it was in danger of losing the entire PC clone market to NEC. The company did not sit idly by. It embarked on a massive legal battle that would take many years to resolve.
This battle is now taught as a case study in law schools because it has every dramatic element: a judge replaced for conflict of interest; the definitive ruling that hardware requires patents while microcode requires copyright; the “white room” evidence that made one engineer fabulously wealthy in just two weeks; and much more.
But that… is a story for Part 2.
Until next time!
