World’s First 3D PCIe 6.0 PHY 8-Lane Test Chip Validates High-Speed Connectivity for Multi-Die Designs

Madhumita Sanyal, Manmeet Walia

Aug 19, 2026 / 2 min read

Subscribe to Our Blog
Thanks for subscribing to the blog! You’ll receive your welcome email shortly.

Introduction

Today, AI, high-performance computing (HPC), data center switching, and storage systems are pushing more data across more interfaces, through more heterogeneous dies, also called chiplets, at faster rates than traditional monolithic SoCs can comfortably support. Synopsys’ world-first 3D PCIe 6.0 test-chip demonstration shows how 64.0 GT/s PCIe connectivity, delivering up to 128.0 GB/s through an 8-lane configuration with PAM4 signaling, can be implemented and validated beyond a traditional 2D design and into a stacked-die, face-to-face architecture.

3D PCIe 6.0 test-chip demonstration

3D PCIe 6.0 test-chip demonstration

3D multi-die design offers a path to higher bandwidth, tighter coupling, and improved efficiency by reducing interconnect distance between dies. In 2.5D packaging, dies are placed side by side and connected through interposers; in 3D packaging, dies are stacked vertically to shorten interconnect distance and enable tighter coupling between heterogeneous functions.

Figure 1. Transmitted Eye at 64 GT/s

Transmitted Eye at 64 GT/s

Figure 2. Receiver Eyes at 64 GT/s Exceeding PCIe 6.0 BER Requirements by Multiple Orders of Magnitude

Receiver Eyes at 64 GT/s Exceeding PCIe 6.0 BER Requirements by Multiple Orders of Magnitude

The test chip for the 5-nm PCIe 6.0 PHY, with design updates for 3DIC technology, is built on an existing PCIe 6.0 test-chip implementation. The excellent silicon results confirm that high-speed PCIe 6.0 IP can be adapted, packaged, brought up, and measured in a 3D environment, an increasingly necessary requirement as customers move from monolithic SoCs to multi-die designs. This enables homogeneous and heterogeneous die integration for high performance, high compute density, energy efficiency, low latency, higher integration, smaller form factors, better power and signal integrity, and lower power consumption compared with current state-of-the-art packaging approaches. The achievement is not just the PCIe 6.0 eye or loopback result in isolation, but the connection between 64 GT/s PCIe behavior, 3D packaging, validation infrastructure, and customer system requirements for AI, HPC, data center switching, and storage. Synopsys’ PCIe 6.0 at 64 GT/s, along with 3D-ready implementation expertise, helps customers design AI accelerators, HPC processors, data center switches, SmartNICs, DPUs, SSD controllers, and CXL-enabled systems.

PCIe 6.0 Implementation Challenges in 3D Packaging

This 3D implementation introduced several challenges, including 3D PHY architectural analysis, TSV implementation and dummy die strategies to optimize performance, accurate modeling of interactions between the top and bottom dies, full-stack modeling and verification of inductor implementation, and minimizing added TSVs while meeting PCIe 6.0 performance goals.

This milestone reinforces Synopsys’ long PCIe legacy, including more than two decades of expertise, silicon-proven PHYs, digital controllers, IDE security IP, verification IP, and extensive third-party interoperability testing, covering hundreds of PCIe 6.x implementations and approximately 4,000 customer tape-outs across seven generations of PCI Express. The results demonstrate a path from standards leadership to mature PCIe IP, 3D-aware implementation, and silicon success. They also connect the proven PCIe ecosystem with emerging multi-die design architectures and give customers designing the next generation of AI infrastructure, HPC platforms, storage systems, and data center fabrics a clearer path for high-performance connectivity into 3D packaging without sacrificing the validation discipline, interoperability experience, and IP completeness that modern systems require.

As a trusted partner, Synopsys is driving the industry's transformation to multi-die designs with a comprehensive and scalable solution for fast heterogeneous integration. The solution, including EDA and IP products, enables early architecture exploration, rapid software development and system validation, efficient multi-die/advanced package co-design and optimization with multiphysics analysis, robust die-to-die and chip-to-chip connectivity, and improved manufacturing and reliability.

Continue Reading

ASK
BETA
Ask BETA This experience is in beta mode. Please double check responses for accuracy.

End Chat

Closing this window clears your chat history and ends your session. Are you sure you want to end this chat?