Researchers Unveil DT-MATCH Chiplet Matching Framework

The defect- and topology-aware framework targets higher yields of functional systems built from repairable chiplets in advanced packaging.

Updated on Nov. 17, 2026 in Tech

Isometric editorial illustration of identical square chiplet tiles in a loose grid on a slate surface, a few cracked, with a thin copper strip linking two tiles.
Arizona State University and Ampere Computing researchers published DT-MATCH, a framework for matching repairable chiplets to raise yields of functional systems in advanced packaging. AI Illustration. Upload story photo >

Researchers at Arizona State University and Ampere Computing published DT-MATCH, a defect- and topology-aware chiplet matching framework that maximizes the number of functional SoCs under realistic assembly constraints. The paper, listed as arXiv:2611.19544, was authored by Xuanyi Tan, Sanmitra Banerjee, Dhruv Thapar, Sreejit Chakravarty, and Krishnendu Chakrabarty.

Why it matters

Heterogeneous integration and advanced packaging introduce yield challenges from manufacturing defects and die-to-die interconnect constraints. Lane-level defects, limited spares, and strict structural compatibility requirements make it hard to assemble functional systems from defective but repairable chiplets.

The ILP-based approach uses integer linear programming with a batching strategy, while the GRASP-based approach uses a greedy randomized adaptive search procedure, both preceded by a preprocessing pipeline that reduces problem size. The extraction does not report specific runtime or yield figures, and the paper's industrial-design results have not been independently verified.

The players

Arizona State University

A public research university in Tempe, Arizona, whose researchers are the lead team behind the DT-MATCH framework.

Ampere Computing

A semiconductor company designing server processors, which is listed as a co-publisher of the DT-MATCH research.

Krishnendu Chakrabarty

A computer engineering researcher named among the paper's five authors on the DT-MATCH framework.

The details

DT-MATCH uses an integer linear programming approach implemented with a batching strategy to approximate the exact formulation, and a greedy randomized adaptive search procedure that accelerates computation without guaranteeing yield optimality. Both approaches rely on a preprocessing pipeline to reduce problem size before matching chiplets.

Timeline

  1. November 14, 2026: The arXiv paper 2611.19544 describing DT-MATCH was listed.

  2. November 17, 2026: Semiengineering.com published an article on the framework.

The Tech Race

DT-MATCH targets the yield problem created by chiplet-based heterogeneous integration, the packaging approach that multi-die server processors now rely on. It extends that architecture by addressing how to assemble functional systems from defective but repairable dies rather than discarding them.

Higher yields from defective but repairable chiplets could translate into more functional processors per wafer over time, though the extraction does not link this to specific consumer products. The research remains a preprint, so practical adoption in commercial manufacturing has not been confirmed.

The takeaway

Chiplet yield gains depend on matching defective dies to compatible partners rather than discarding them, a shift that could matter as advanced packaging spreads. Readers tracking semiconductor manufacturing should watch whether these preprint results are validated in peer review.

Further reading

For related coverage, explore Semiconductors in Arizona.

More information

Read the DT-MATCH technical paper on arXiv for the full methodology and results.

Source note: This article includes information reported by Semiconductor Engineering.