A Historical Look at the VAX: The Economics of Microprocessors [Part I]

This document, split into two parts, analyzes the historical context and economic pressures that led Digital Equipment Corporation (DEC) to transition from its VAX architecture to the Alpha processor.

Part I: The Economics of Microprocessors This section highlights that the high cost of chip design and fabrication ($50M example) necessitates high unit volumes to achieve competitive per-chip costs. It illustrates this by comparing MicroVAX II volumes (65,000 units) to IBM PC volumes (14,000,000 units) from 1985-1987, showing how x86 architectures gained a massive cost and performance advantage due to scale. High volume allows more investment in optimization and leads to lower average costs. The article explains how traditional minicomputer companies struggled to adapt to this new economic model, facing internal conflicts over strategies like designing their own chips (CISC, ECL, CMOS), or buying microprocessors from others. Owning a fabrication plant (fab) was expensive and required high volume to amortize costs, but not owning one meant losing the ability to highly tune designs for peak performance.

Part II: DEC, NVAX, Alpha and Competitors This section focuses on the competitive landscape DEC faced in the early 1990s. It compares DEC's late-stage VAX CPU (NVAX/NVAX+) with its Alpha 21064, which were developed concurrently with similar resources. Despite the NVAX team's greater experience, the Alpha's simpler RISC (Reduced Instruction Set Computer) architecture allowed for more aggressive microarchitectural implementations, resulting in 3-5 times better performance than NVAX. The document shows that by 1992, the VAX was significantly outperformed by other RISC architectures (IBM, HP, MIPS, Sun) in technical applications and was quickly approached in integer performance by less expensive x86 chips (like the 486DX2).

Ultimately, the document argues that sustaining a competitive VAX architecture (e.g., as a superscalar or out-of-order design) against the rapid advancements of RISC and x86 was not economically or timely viable due to the VAX's complex CISC (Complex Instruction Set Computer) instruction set. Robert Supnik's foreword explains that DEC's Executive Committee concluded that only a new RISC architecture (Alpha) could ensure long-term competitiveness while maintaining strong software compatibility, leading to the massive engineering effort behind Alpha AXP.