【Semiconductor】15. What Is Legacy Technology

— A Record of Failures from the Era When Semiconductors Were Ruled by Physics

topics: [“Semiconductor”, “Technology History”, “DRAM”, “Reliability”]


🧭 What Is Legacy Technology?

Legacy Technology is not
an archive created to nostalgically look back on old semiconductor technologies.

It is a structured record of
real failures and recoveries that occurred in an era
when semiconductor devices were directly and unavoidably governed by physical constraints.

This was before problems could be
“hidden” by software, firmware,
or system-level compensation.

Moments when process integration, cell structure, and device physics
directly determined yield, reliability, and business decisions
are preserved here.

📌 The complete archive covered by this article is available at:
https://samizo-aitl.github.io/Edusemi-Plus/archive/legacy/


🕰 Historical Background

— The Era When DRAM Was “Pure Physics”

Until the mid-1990s, Japan stood at the forefront of DRAM technology worldwide.

Multiple domestic manufacturers simultaneously achieved:

The core axes of competition were not limited to density or cost,
but focused on:

At that time, DRAM cell design was treated fundamentally
as an analog, physics-driven problem.


🏗 Design Culture and Assumptions of the Time

The design culture of that era was built on clear assumptions:

While this culture produced extremely robust memory devices,
it relied heavily on one critical premise:

Physical margins must never be violated

As long as this premise held,
reliability was something that could be guaranteed by design.


⚠️ The Turning Point

(Late 1990s to Early 2000s)

As scaling entered the 0.25 µm generation and beyond,
this premise began to collapse.

🔄 What Changed

Item Reality
🧮 Cell capacitance Structural margins collapsed
💧 Leakage / Disturb Became the norm, not exceptions
⏳ Retention Became the dominant constraint
💰 Market pressure Speed and cost prioritized over physical certainty

At the same time, DRAM prices entered a long-term decline,
forcing decisions to push products into mass production
before failure mechanisms were fully understood
.

Failures such as Pause, Disturb, and Retention loss
were no longer hypotheses.

They became real problems that surfaced:


🔍 Why Read Legacy Technology Now?

The technologies discussed here are
already more than 20 years old.

However,
the structure of failure has not aged.

The same causal structures repeatedly appear
in modern semiconductor systems:

What has changed is only the
scale, vocabulary, and abstraction layers.


🎯 Scope of This Archive

This archive focuses on the
intersection of physics, manufacturing, and decision-making.

The following are intentionally excluded:


🧭 How to Read This Archive

Each case is organized in the following order:

  1. 🏗 Process / Structure
  2. 🔎 Observed failure modes
  3. 🧠 Physical root causes
  4. 🧪 Manifestation in tests / binning
  5. 🔧 Yield recovery or strategic decisions

This is not a retrospective explanatory order.
It reflects the actual sequence in which problems appeared and were resolved in the field.


🧱 Positioning

There is only one reason this archive exists.

🧱 Physics does not disappear as technology advances.

It merely becomes easier to ignore.

This archive records the moments
when physics could no longer be ignored.



⏭ Next

From the next article onward,
we will enter the actual anomalies observed in 0.25 µm generation DRAM.

First, we will address only
what was happening.

The causes and interpretations will follow later.