Eliyan Breaks Chiplet Memory Wall With Standard Packaging
Eliyan recently taped out its NuLink die-to-die PHY IP on TSMC N3, achieving 64 Gbps per bump using standard packaging, Eliyan CEO Ramin Farjadrad told EE Times. This is equivalent to 4.55 Tbps per millimeter of bandwidth at less than half a picojoule per bit. Die-to-die performance like this is usually seen only in advanced packaging designs using costly and complicated silicon interposers, he said, whereas Eliyan’s PHY technology enables multi-chiplet designs with advanced-packaging–like performance on standard organic substrates.

Achieving HBM3-like memory bandwidth on standard packaging could have implications for future chiplet-based AI accelerators, particularly those designed for generative AI inference, in which memory capacity and bandwidth are critical. Eliyan’s technology also allows more HBM stacks to surround each compute chiplet via extended reach and daisy-chaining.
To read the full article, click here
Related Chiplet
- FlexGen Multi-Die Smart Network-on-Chip (NoC) IP
- Ncore Multi-Die Interconnect IP
- Integrated voltage regulator (IVR) chiplet
- High-performance connectivity chiplets
- eFPGA Chiplet
Related News
- Syenta Raises A$8.8M to Break AI’s Memory Wall with Next-Gen Chip Packaging Technology
- Breaking the Memory Wall: How d-Matrix Is Redefining AI Inference with Chiplets
- Eliyan Showcases Next-Generation Chiplet Interconnect and Memory Innovations at OCP Global Summit 2025
- PGC Integrates 2.5D/3D Advanced Packaging Technology to Break the “Memory Wall” and Accelerate AI/HPC ASIC Innovation
Latest News
- Polaris Electro-Optics Secures $50M Series B to Scale Next-Gen Optical Engine Platform for AI Infrastructure
- Arteris Accelerates Industry Transition to Multi-Die Chiplet-Based Architecture
- Fujitsu launches made-in-Japan next-generation CPU FUJITSU-MONAKA and Fujitsu MONAKA Server for sovereign AI infrastructure
- Signaloid joins Open Chiplet Atlas Alliance and Announces Plans to Make Its UxHw ASICs Available via OCA Ecosystem
- Lorentz Solution Jointly Presents with NVIDIA on Automated Large-Scale Chip-Level 3D EM Extraction and Sign-off of Ultra-High-Speed Silicon Photonics and IC/3DIC Designs at 2026 TSMC OIP