Tech BusinessChips & Semiconductors
Chiplet matching method targets yield in advanced packaging
Researchers from Arizona State University and Ampere Computing have proposed DT-MATCH, a method for matching chiplets with defects so that more functional systems can be assembled. The work targets yield problems in heterogeneous integration and advanced packaging.

Key takeaways
- DT-MATCH aims to maximise functional systems from defective but repairable chiplets.
- It uses defect and topology information to guide chiplet matching.
- One approach uses integer linear programming, while the other uses a faster randomized search.
- Tests on industrial designs found better solution quality from the integer linear programming approach.
In this article
What was announced
According to Semiconductor Engineering, researchers at Arizona State University and Ampere Computing have proposed a chiplet matching method called DT-MATCH. The work, reported on 17 November 2026, aims to improve yield in advanced packaging by making better use of chiplets that have defects but remain repairable. The paper itself is dated 14 November 2026.
The chiplet yield problem

The paper addresses heterogeneous integration, where chiplet-based system-on-chip designs are built with advanced packaging. It says this approach enables new designs but also introduces yield challenges. In the paper, yield is treated as the number of functional systems that can be produced.
Manufacturing flaws and die-to-die interconnect limits can stop an assembly from working. The researchers highlight lane-level defects, scarce spares, and strict structural compatibility rules as key obstacles. The authors describe the chiplets as defective but repairable. These issues make it hard to assemble functional systems from chiplets that are not perfect but can still be used.
What DT-MATCH does
The proposed framework is called DT-MATCH. It is designed to maximise the number of functional systems under realistic assembly constraints. The method uses defect information and topology information when matching chiplets. It is defect-aware because it considers where defects occur, and topology-aware because it considers how chiplets can be connected.
A preprocessing stage reduces the problem size before the main optimisation begins. In this context, topology refers to the structural relationships that determine which chiplets can be assembled together. The paper is by Xuanyi Tan, Sanmitra Banerjee, Dhruv Thapar, Sreejit Chakravarty and Krishnendu Chakrabarty. It is dated 14 November 2026 and is listed as arXiv:2611.19544.
Two matching approaches
DT-MATCH provides two complementary approaches. The first uses integer linear programming, a mathematical optimisation method, with a batching step that approximates the exact formulation. This route is intended to give higher quality matches.
The second uses a greedy randomized adaptive search procedure. This route provides substantial acceleration but does not guarantee optimal yield. Both approaches rely on the same preprocessing stage to make the problem smaller.
Results and implications
The researchers tested the methods on industrial designs. They report that the integer linear programming approach achieved better solution quality. The authors say this better quality was expected.
The greedy randomized adaptive search procedure led to lower runtimes, but it does not guarantee optimal yield. The excerpt cuts off before describing the full tradeoff, does not name the industrial designs, and gives no detailed numerical results. The work matters because chiplet-based designs need careful assembly to reach good yield. If chiplets with defects can be matched correctly, more functional systems may be built from the same material.
Sources
- Defect-Aware Chiplet Matching For Heterogeneous Integration Yield (ASU, Ampere)Semiconductor Engineering
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.



