HomeChiplets & UCIeDay 6
DAY 6 · PHASE 2 — ADVANCED PACKAGING

2.5D Packaging
Silicon Interposers & CoWoS

By EcrioniX · Updated July 2026

Phase 1 covered the interconnect standard. Phase 2 opens with the physical packaging technology that actually makes UCIe's Advanced Package numbers achievable. Day 6 explains what a silicon interposer really is, how TSMC's three CoWoS variants differ, and how Nvidia's Blackwell GPU uses the newest one to connect two dies at 10 TB/s.

2.5D — Side by Side, Not Stacked

"2.5D" describes a specific arrangement: multiple chiplets sit side by side on top of a shared interposer, which itself sits on the package substrate. The interposer is the thing doing the hard interconnect work between chiplets — the chiplets themselves aren't stacked directly on top of each other, which is what "3D" packaging means instead (Day 7). Think of 2.5D as chiplets on a shared, high-density wiring board that's itself part of the package, rather than a discrete PCB.

2.5D Packaging — Cross-Section Compute Chiplet HBM Stack micro-bumps (fine pitch) Silicon Interposer (passive, mature node) RDL — fine-pitch lateral routing between chiplets C4 bumps (coarser pitch) — TSVs carry signal through Package Substrate (to board/BGA)

What a Silicon Interposer Actually Is

A silicon interposer is deliberately unglamorous: it's a piece of silicon fabricated on a mature, older process node (there's no benefit to using an expensive leading-edge node for something that does no computation), and it is entirely passive — no transistors doing logic, no active circuitry. Its only job is interconnect, achieved through two mechanisms:

TSV (Through-Silicon Via)A vertical electrical connection drilled and filled through the body of the interposer, carrying signals from the coarser package-side bumps up to the fine-pitch chiplet side
RDL (Redistribution Layer)A fine-pitch metal wiring layer on top of the interposer that routes laterally between chiplets — this is what actually carries the die-to-die UCIe link from one chiplet to its neighbor

Put simply: TSVs go through the interposer (vertical), RDL goes across the interposer (lateral). Together they let a chiplet's fine-pitch bumps connect both to its neighboring chiplet (via RDL) and down to the coarser package substrate (via TSV), without the chiplet itself ever needing bumps at two different pitches.

TSMC CoWoS — Three Variants, Three Tradeoffs

CoWoS (Chip-on-Wafer-on-Substrate) is TSMC's flagship 2.5D packaging technology, introduced in 2012 and now the dominant choice for AI and HPC silicon. It comes in three distinct variants that are not just configuration options — they trade off cost, yield, and performance differently:

VariantInterposer TypeKey TradeoffTypical Use
CoWoS-SSingle silicon interposer + TSVsBest raw performance, but yield drops as interposer area growsNvidia H100, AMD MI300 — current AI/HPC mainstream
CoWoS-ROrganic RDL interposer (InFO fan-out, no silicon)Lower cost, no TSV silicon needed, less bandwidth-criticalNetworking products
CoWoS-LLocal Silicon Interconnect (LSI) bridges + RDL interposerPreserves TSV-level performance while stitching past single-interposer size/yield limitsNvidia Blackwell B100/B200

Variant details per TSMC technical disclosures and industry packaging coverage.

CoWoS-L — Beating the Reticle Limit Twice Over

CoWoS-L is the most architecturally interesting of the three. Instead of one large monolithic silicon interposer (which, like a die, suffers yield problems as its area grows — the same yield-cliff logic from Day 1), CoWoS-L uses small local silicon interconnect (LSI) bridges embedded into a larger RDL interposer, stitching multiple segments together. This lets CoWoS-L build effective interposer areas over 3,000 mm² — far beyond what a single silicon interposer or even the reticle limit itself would allow — while still getting TSV-grade electrical performance exactly where the LSI bridges sit, right under the chiplet-to-chiplet connections that need it most.

This is the same idea from Day 1 applied one level up: just as splitting one large die into several chiplets sidesteps the reticle limit and the yield cliff, stitching several small silicon bridges into a larger RDL interposer sidesteps the same yield problem at the packaging level.

Case Study — Nvidia Blackwell B100/B200

Nvidia's B100 and B200 GPUs were the first processors to use CoWoS-L in production. Blackwell connects two large compute chiplets using local silicon interconnect bridges and an RDL interposer, achieving 10 TB/s of chip-to-chip bandwidth between the two dies and packing 208 billion transistors into one package — more than 2.5× the transistor count of the prior Hopper generation's roughly 80 billion.

This is the practical payoff of everything covered so far in this course: two reticle-limited dies, joined by die-to-die interconnect running over a CoWoS-L 2.5D package, behaving as a single GPU to the software stack above them.

The Honest Challenge — Thermal Mismatch

2.5D packaging isn't free of real engineering pain. Nvidia's early Blackwell production ramp ran into coefficient of thermal expansion (CTE) mismatch — the GPU chiplets, the silicon/RDL interposer, the LSI bridges, and the package substrate are different materials that expand and contract at different rates during manufacturing and thermal cycling. That mismatch caused warping severe enough to affect yield and reliability during the initial ramp. Advanced packaging solves the reticle limit and yield-cliff problems from Day 1 — but it introduces new thermomechanical problems that a monolithic die never had to deal with.

🎯 Day 6 Key Takeaways

Frequently Asked Questions

What does 2.5D packaging mean?
2.5D packaging places multiple chiplets side by side on top of a shared interposer, which handles the fine-pitch interconnect between them, rather than stacking chiplets directly on top of one another (which is 3D packaging, covered on Day 7). The interposer sits between the chiplets and the package substrate.
What is a silicon interposer?
A silicon interposer is a passive piece of silicon, fabricated on a mature process node, that provides no active computation of its own. It exists purely to carry interconnect: through-silicon vias (TSVs) route signals vertically through the interposer, and a redistribution layer (RDL) on top provides fine-pitch lateral wiring between chiplets.
What is the difference between CoWoS-S, CoWoS-R, and CoWoS-L?
CoWoS-S uses a single silicon interposer with TSVs for direct high-speed signal transmission, but yield drops as interposer size grows. CoWoS-R replaces the silicon interposer with an organic RDL interposer built using InFO fan-out technology, reducing cost for less bandwidth-critical products. CoWoS-L combines local silicon interconnect (LSI) bridges with an RDL interposer, preserving TSV-level performance while stitching multiple segments together to exceed the size a single silicon interposer could yield well.
What packaging does Nvidia's Blackwell GPU use?
Nvidia's Blackwell B100 and B200 GPUs were the first processors to use TSMC's CoWoS-L packaging, connecting two compute chiplets with local silicon interconnect bridges and an RDL interposer to achieve 10 TB/s of chip-to-chip bandwidth and 208 billion transistors in one package.
What are the practical challenges of 2.5D packaging?
A major challenge is coefficient of thermal expansion (CTE) mismatch between the chiplets, the silicon or RDL interposer, and the package substrate — different materials expand and contract at different rates during manufacturing and operation, which can cause warping severe enough to affect yield and reliability, as seen during Nvidia Blackwell's early CoWoS-L production ramp.