technologybriefs
9:56in productionCh. 1 · The First Family/ 9:56 · ceiling 15 min
Systems

IBM System/360

Compatibility was engineered — not promised — and paid for in transistors, not silicon.

The IBM System/360 established architectural compatibility as a design principle — not a marketing claim. It delivered real portability across a wide performance range using microcode and custom transistor packaging. Its limitations — no ICs, partial model incompatibility, rare max-memory deployments — are as instructive as its successes.

Chapters & takeaways5
  1. 1:07
    The First Family

    It was the first computer system family — launched on April 7, 1964.

  2. 2:02
    Architecture Over Hardware

    It separated architecture from implementation to enable price-tiered compatibility.

  3. 3:20
    Microcode and Transistors

    It used microcode and Solid Logic Technology — not integrated circuits — to execute a unified instruction set.

  4. 4:45
    Scale Without Uniformity

    Performance ranged from 34,500 to 16.6 million instructions per second; memory from 8 KB to 8 MB — though 256–1024 KB was typical.

  5. 5:50
    One Code, All Sizes

    It enabled upgrades without rewriting applications — the first system designed for both commercial and scientific use across all sizes.

Worth your time?

Yes. Study the whole thing.

4.5/ 5
What works
  • application portability across models
  • performance scaling from 34.5k to 16.6M IPS
  • unified instruction set with 8-bit byte addressing
What does not
  • use integrated circuits
  • achieve full model compatibility
  • deploy 8 MB memory routinely
Study it if
  • system architects
  • computer historians
  • legacy infrastructure engineers
Skip it if
  • cloud developers
  • AI practitioners
  • embedded systems designers
The written brief1 min read

What it is and the problem it solves

The IBM System/360 is a family of mainframe computers announced on April 7, 1964. It solved the problem of vendor lock-in and software rewrite costs by enabling companies to upgrade hardware without rewriting applications.

How it works

It separated architecture from implementation, enabling compatible models at different price points. It used microcode (except in Model 44 and the most expensive systems) to implement an instruction set with 8-bit byte addressing and multiple arithmetic formats. It introduced Solid Logic Technology — discrete transistor-based circuit cards — deliberately excluding integrated circuits as immature.

What works

Application compatibility worked across most models. Performance scaled predictably: Model 30 delivered up to 34,500 instructions per second; Model 91 delivered up to 16.6 million. Memory configurations ranged from 8 KB to 8 MB, with common large installations using 512–1024 KB.

What does not

It did not use integrated circuits. It did not deliver full compatibility across all models: Model 44 was only partially compatible. It did not achieve uniform memory scaling: 8 MB was a theoretical maximum, rarely deployed; typical large installations used 256–1024 KB.

What it changes

It changed how computer families were conceived: no longer bespoke machines, but tiered, upgradeable systems sharing software. It established that application portability across performance tiers could be engineered — not assumed — through disciplined architecture/implementation separation.

Is it worth your time

Yes, if you work with legacy system design, compatibility trade-offs, or hardware abstraction. Its mechanism — architectural standardisation without silicon integration — remains a masterclass in pragmatic scalability. But its relevance is historical, not operational.

Same field · Systems4 of 61
Up next in Technology

IEEE 802.11

IEEE · 9:09

802.11 doesn’t deliver speed—it delivers a contract for sharing air, written in CSMA/CA and enforced by band splits.

9:09