Microprocessor 1971 Step 16 Yield Cost Economics General Purpose vs Fixed Function

# Price the Difference: Why a Reusable Processor Beats a Fixed-Function Chip Set on Cost, Not Just Flexibility

## 1. Step 9's Tradeoff, Converted Into a Cost Model

Step 9 of this series frames the 4004's central tradeoff as flexibility against raw speed — a fixed-function chip set executes any one job faster than a general interpreter executes the same job. That framing leaves out the variable that actually decided Busicom's and Intel's commercial choice: non-recurring engineering cost, amortized across however many chips of a given design are eventually sold.

A fixed-function chip set's design cost — the engineering hours spent laying out a circuit that does exactly one calculator's arithmetic and nothing else — can only be recovered by selling that one calculator's volume, because the chip has no other customer. A general-purpose processor's design cost is paid once and then spread across every unrelated customer who buys the same part for a different application, because Step 1's architectural separation of behavior from silicon means the same CPU serves a calculator, a terminal, or anything else a customer's own program ROM defines. The reusable design's cost per unit falls as volume grows across customers, not just within one customer's own production run.

Design Cost Amortized Across One Customer, Versus Many the variable Step 9's speed-versus-flexibility framing left out cost / unit cumulative volume fixed-function: one customer's own volume only general-purpose: every customer's combined volume same design cost, divided by a denominator that grows across unrelated customers

## 2. Yield as a Second, Separate Economic Variable

Design-cost amortization is not the only economic difference. A general-purpose processor also benefits from yield learning that a one-off fixed-function chip never accumulates: because the same 4004 design is fabricated repeatedly, in volume, for many customers, Intel's process engineers refine the same mask set's yield over many production runs, something a bespoke chip made for a single customer's one-time order never gets the chance to do before that order is complete. Step 8's instruction-cycle-versus-discrete-logic comparison already shows the 4004 winning on system assembly cost; this step adds that the 4004's own fabrication cost per good die was also falling over time in a way no single-customer fixed-function design could replicate, because fixed-function designs rarely survive long enough in production to climb the same learning curve.

Yield Learning Needs Repeated Production Runs a fixed-function chip rarely stays in production long enough to climb this curve ONE-CUSTOMER FIXED CHIP one order, one production run yield fixed at whatever the first runs achieve no second customer to fund further runs MULTI-CUSTOMER GENERAL CHIP many orders, many production runs yield improves run over run on the same mask set every new customer funds more learning the reusable design's economics compound; the fixed-function design's do not

## 3. Why This Reframes, Without Replacing, Step 9's Conclusion

Step 9 is still correct within its own scope: a fixed-function circuit executing exactly the job it was built for will usually run that one job faster than a general interpreter running the same job through stored instructions. What this step adds is the reason that speed advantage did not win the commercial argument — Busicom's actual decision, described in Step 11, was driven by a renegotiated price, and the economics behind that price are design-cost amortization and yield learning, both structurally favoring the reusable design once more than one customer exists. Raw speed was never the only variable in the comparison Step 9 sets up; cost, compounding across customers neither Busicom nor Intel had yet signed when the 4004 was designed, decided the outcome just as much.

Take microprocessor 1971 step 16 yield cost economics general purpose vs fixed function further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app