APPLICATIONS / SRAM REPAIR · CAPACITY PLANNING應用 / SRAM 修復 · 先進節點容量規劃

Size your SRAM repair data把修復需求,算清楚

Explore compression and allocated OTP / eFuse capacity for large SRAM repair datasets.從大型 SRAM 到修復資料,掌握壓縮效益與 OTP / eFuse 配置容量。

Capacity overview容量總覽

Updated已更新
Raw repair data壓縮前修復資料↗
16 Mb
16,000,000 bits
2,000,000 B · 1,953.125 KiB
Compressed payload壓縮後有效資料↘
160 kb
160,000 bits
20,000 B · 19.531 KiB
Capacity reduction資料容量減少↓
99 %
15,840,000 bits saved節省 15,840,000 bits
Compression 100 : 1壓縮倍率 100 : 1

How compression changes capacity壓縮倍率如何改變需求

Compare allocated capacity across compression ratios比較不同倍率下,兩種技術所需的配置容量

OTPeFuse● Current ratio● 目前倍率
↳

16 Gb SRAM at 1/1000 raw repair data and 100:1 compression requires a 160 kb payload.16 Gb SRAM 在 1/1000 修復資料比例、100:1 壓縮假設下,需 160 kb 有效修復資料。

ALLOCATION COMPARISON配置容量比較

One payload, two allocations相同資料,兩種配置方式

Adjust overhead, reserve and allocation blocks independently to calculate the required bits.分別調整額外開銷、預留空間與配置粒度,查看實際需要配置多少 bits。

OTP and eFuse capacity breakdownOTP 與 eFuse 位元數比較明細
Capacity breakdown容量明細 OTP eFuse
Both technologies currently have the same allocated capacity.目前兩種技術的配置容量相同。
ƒMethod and assumptions計算方法與假設+

Data requirement資料需求

SRAM bits = ceil(input capacity × bits per unit)SRAM bits = ⌈輸入容量 × 單位位元數⌉

Raw repair bits = ceil(SRAM bits × repair fraction)原始修復 bits = ⌈SRAM bits × 原始修復資料比例⌉

Payload bits = ceil(raw repair bits / compression ratio)有效資料 bits = ⌈原始修復 bits / 壓縮倍率⌉

Capacity reduction = 1 − payload bits / raw repair bits資料容量減少比例 = 1 − 有效資料 bits / 原始修復 bits

Allocated storage實際配置

Overhead bits = ceil(payload bits × overhead %)開銷 bits = ⌈有效資料 bits × 開銷 %⌉

Reserve bits = ceil((payload bits + overhead bits) × reserve %)預留 bits = ⌈(有效資料 bits + 開銷 bits)× 預留 %⌉

Required bits = payload bits + overhead bits + reserve bits需求 bits = 有效資料 bits + 開銷 bits + 預留 bits

Allocated bits = ceil(required bits / block bits) × block bits配置 bits = ⌈需求 bits / 粒度 bits⌉ × 粒度 bits

Reading this estimate如何閱讀這份試算

Each stage rounds up to whole bits. Bytes round up as ceil(bits / 8). kb / Mb / Gb are decimal; Kib / Mib / Gib are binary, with 1 Gib = 2³⁰ bits. Lowercase b means bits; uppercase B means bytes.⌈ ⌉ 表示向上取整。各階段取整至完整位元;B 顯示容納位元所需的完整位元組(⌈bits / 8⌉)。kb / Mb / Gb 採十進位;Kib / Mib / Gib 採二進位,1 Gib = 2³⁰ bits。小寫 b 是位元、大寫 B 是位元組。

The defaults of 16 Gb, 1/1000 and 100:1 are adjustable planning assumptions, not IP specifications. Both technologies start with 0% overhead, 0% reserve and 1-bit blocks to show payload demand. Overhead can approximate ECC, indexes and control data; use actual encoding and macro capacities for implementation.預設 16 Gb、1/1000 與 100:1 為可調的規劃情境;OTP / eFuse 的開銷、預留均為 0%,粒度為 1 bit,用來呈現純資料需求,並非特定 IP 規格。開銷可概估 ECC、索引與控制資訊;正式配置請依實際編碼及巨集容量設定。

This model compares storage capacity. It excludes bitcell area, peripheral circuits, reliability, power and cost, so results do not establish area advantage, repair success rate or technology superiority. The 16 Gb example is not a validated area crossover.本模型比較儲存容量,未納入每位元面積、周邊電路、可靠度、功耗或成本,因此容量結果不代表面積優勢、修復成功率或特定技術的優劣。16 Gb 是範例容量,並非已驗證的面積交叉點。

Calculations run in your browser. Save A / B below to retain scenarios, or download JSON to restore them elsewhere. CSV preserves a readable result; JSON restores inputs.參數在瀏覽器中運算。使用下方 A/B 保存情境,或下載 JSON 在其他瀏覽器還原;CSV 保留可讀結果,JSON 用來還原輸入。

SAVE / COMPARE / HAND OFF保存/對照/交接

Compare two planning scenarios讓兩個規劃情境可以重現

Each scenario includes both OTP and eFuse. Save inputs to A or B, then compare or restore them. Calculator edits do not change saved scenarios.每個情境都包含 OTP 與 eFuse。將輸入保存至 A 或 B,再並列比較或載入;修改計算器不會改動已存情境。

Scenario A情境 A

Scenario B情境 B

Saved scenarios stay in this browser when storage is available; they are not sent to a server. Download JSON for a portable copy. Reloading resets the calculator draft; A and B remain saved. Import accepts this tool’s schema 1 files up to 64 KiB.儲存空間可用時,情境會保存在此瀏覽器,不會上傳伺服器;可下載 JSON 攜帶。重新整理會重設計算器草稿,已存的 A、B 仍保留。匯入接受本工具結構版本 1、最大 64 KiB 的 JSON。

FIRST-PRINCIPLES YIELD MODEL · BIRA ECONOMICS第一性原理良率模型 · BIRA 經濟學

SRAM Yield Recovery & BIRA Wafer Economics Simulator SRAM 備援良率挽救與 BIRA 晶圓經濟效益試算器

Model Poisson/Murphy random defect statistics, bank redundancy repair, and dollar value recovery per 300mm wafer. 基於 Poisson 與 Murphy 隨機瑕疵統計模型,估算記憶體區塊備援修復良率增益與 300mm 晶圓產值挽救金額。

65%
0.08 /cm²
未修復原始良率UNREPAIRED BASE YIELD
—
BIRA 修復後良率BIRA REPAIRED YIELD +0%
—
單片晶圓多得好晶片EXTRA DIES / 300MM WAFER
—
每片晶圓挽救產值VALUE RECOVERED / WAFER
—
萬片量產累計產值ANNUAL 10K RUN VALUE
—
良率提升對照 (Unrepaired Base vs BIRA Repaired):YIELD ENHANCEMENT BENCHMARK: 0% ──── 50% ──── 100%
未修復原始良率Baseline Yield
BIRA 備援修復良率BIRA Repaired Yield
—

Review imported scenarios預覽匯入情境

Confirming replaces saved A and B. Current calculator inputs stay unchanged. Imported results are recalculated, and file source metadata is not authenticated.確認後會替換已存的 A、B,目前計算器輸入保持不變。結果會重新計算;檔案來源欄位未經真實性驗證。

Saved A / B · A4 summary已存 A/B · A4 摘要