• Skip to main content
  • Skip to search
  • Skip to footer
Cadence Home
  • This search text may be transcribed, used, stored, or accessed by our third-party service providers per our Cookie Policy and Privacy Policy.

  1. Blogs
  2. Digital Design
  3. Accelerating Scalable and SI-Aware 2.5D Interconnect Routing…
Vinod Khera
Vinod Khera

Community Member

Blog Activity
Options
  • Subscribe by email
  • More
  • Cancel
CDNS - RequestDemo

Have a question? Need more information?

Contact Us
OmniRoute
Advanced Interconnect Router
Integrity 3D-IC Platform
3DIC
Inerconnect

Accelerating Scalable and SI-Aware 2.5D Interconnect Routing with OmniRoute

17 Nov 2026 • 9 minute read

In advanced multi-die architectures, a key part of system performance is efficient die-to-die communication. This normally happens through a silicon or RDL interposer layer containing high speed signals. As these architectures scale, interposer routing is increasingly becoming a system-level bottleneck, due to massive interconnect counts, evolving protocols such as UCIe and HBM, and shrinking signal integrity (SI) margins. This is especially true in AI and HPC systems, where interposers are crucial as AI accelerators depend on multi-HBM interposer architectures to deliver bandwidth and performance. Design teams are under tremendous pressure to meet shorter time-to-market while handling high-density routing, low-margin SI requirements, and protocol update demands. 

Taken together, these trends have pushed interposer routing beyond the limits of conventional approaches. That is where the Cadence OmniRoute Advanced Interconnect Router comes in. It is designed specifically to address this shift, enabling automated die-to-die routing for modern 2.5D interposers, silicon bridges, and broader multi-die integration flows, including emerging 3D and 3.5D SoC designs.

This blog focuses on the routing and implementation challenges most visible in 2.5D interposer design, while also recognizing that many of the same scale, SI, and automation pressures extend into 3D-IC integration. It also proposes a unique flow in which the OmniRoute interconnect router can work together with Cadence Cerebrus Explorer, explore the routing design space, and can come up with a SI-aware die-to-die routing solution to accelerate convergence compared to manual tuning.

Why Die-to-Die Routing Has Become So Hard

As systems scale, interposers grow in both size and complexity. Large interposers now integrate multiple chiplets and memory stacks, dramatically increasing the number of interconnects that must be routed. The complexity of interposer routing today is not driven by a single factor; it is the convergence of scale, protocol-driven constraints, and shrinking SI margins.

Modern interposers can span multiple reticle sizes and incorporate numerous dies, leading to routing requirements that can reach tens of thousands or even millions of nets. Evolving protocols also play a critical role in increasing routing congestion.

UCIe® (Universal Chiplet Interconnect Express) has introduced increasingly tight routing requirements for each generation. Moving from UCIe 1 to UCIe 2 increases lane counts and shrinks bump pitch. Designers must meet stricter requirements for layer ordering, length matching, impedance control, and shielding, while SI margins narrow and leave less room for error at high data rates.

HBM evolution is widening the routing problem, from HBM2 through HBM3 and HBM4, and with future HBM generations continuing the trend toward higher bandwidth and denser connectivity, each generation brings higher data rates, denser bump patterns, and wider buses. HBM stacks have grown from four layers to eight, twelve, and beyond, significantly increasing base die bump density and driving more routing congestion while shrinking SI margins.

Signal integrity considerations are becoming inseparable from routing itself. High-speed interfaces require aggressive shielding strategies, often involving both same-layer and interlayer shielding structures, with high shielding ratios applied to critical nets.
As frequencies increase, SI margins shrink, and even small routing variations can impact performance.

These transformations have made interconnect routing a system-level bottleneck, requiring a purpose-built solution that can handle massive scale, evolving standards, and tight signal integrity constraints without slowing design schedules. Cadence OmniRoute Advanced Interconnect Router is designed specifically to address this shift, enabling automated and SI-aware die-to-die routing for 2.5/3D/3.5D interposers, silicon bridges, and broader multi-die integration flows, including emerging 3D-IC designs.

OmniRoute Overview: Purpose-Built for Interposer Routing

OmniRoute interconnect router is Cadence’s new router for die-to-die routing needs. It is designed from the ground up for auto-routing on silicon interposers and silicon bridges. The physical implementation of the interposer involves routing between dies (e.g., between an HBM and an ASIC) or between any two dies. Space congestion and a limited number of routing layers create significant challenges. Furthermore, inter-die interconnects typically need to traverse longer distances than on-die interconnects. In most designs with access to multiple HBM stacks, 45 degree-angled routing is a key requirement as well. Traditional manual routing consumes a significant amount of time to address these challenges.

OmniRoute Advanced Interconnect Router

The auto-router brings together several capabilities that the complexity of modern interposer routing demands. It can automatically recognize UCIe and HBM protocol bump maps without manual pre-processing and provides fast automation required for modern interposer routing. Usually, for a single UCIe or HBM, it just runs for 30 seconds or one minute, very fast.

Unified Routing Cockpit

OmniRoute interconnect router integrates with the Cadence Integrity 3D-IC Platform, bringing routing, extraction, and SI analysis into a unified environment. This allows package design teams to evaluate insertion loss, return loss, crosstalk, and eye diagrams while optimizing interposer designs. The result is a more interactive engineering flow, where fast turnaround supports rapid iteration and quicker convergence on a manufacturable, SI-clean implementation.

Unified Routing

Cadence also has Automatic Substrate Router (ASR) technology that addresses package-level and RDL routing requirements across the broader system. By enabling package routing and silicon routing to be optimized within the same implementation framework, the unified routing cockpit helps eliminate the inefficiencies of managing these activities as separate design tasks.

Together, OmniRoute and ASR create a highly automated implementation flow that spans both routing domains, particularly for advanced packages with embedded silicon bridges. Engineers no longer need to manually coordinate silicon bridge routing and package routing as separate activities. Instead, both operate within a unified constraint framework, allowing silicon bridges, package routing, connectivity intent, and implementation results to be managed from a single cockpit. The result is faster convergence, greater consistency, and significantly lower integration overhead across the complete system. We will discuss more about this in the next blog post about silicon bridges.

In essence, it is built to solve a very specific problem: how to make interposer routing scalable, repeatable, and fast enough to keep pace with modern system design, which involves integrating a large number of chiplets and embedded bridges.

A New Routing Approach: Constraint-Driven Automation

One of the most significant shifts OmniRoute interconnect router introduces is the move from manual routing effort to constraint-driven automation. The workflow is intentionally simple and built around two steps.

Step 1: Read route constraints.

Step 2: Run OmniRoute with a single command.

From there, the OmniRoute interconnect router handles bump ordering, layer assignment, shielding generation, and shielding tie-off automatically. If a parameter needs to change, the designer updates a YAML constraints file and reruns. With routing completed in seconds to minutes, the iteration loop is fast enough to become part of a normal design session rather than an overnight batch job.

Advanced Capabilities and Integrated Workflow

The OmniRoute interconnect router supports a broad range of routing capabilities and topology options beyond basic auto-routing.

Routing Topology Control

Multi-zone routing allows designers to define different widths, spacings, and shielding for different regions of the same interface, such as a tighter pitch near the bumps and a wider pitch in the open-trunk

 region between dies. The OmniRoute interconnect router handles zone transitions cleanly, which is particularly useful for HBM designs. Bus guide is another mechanism for routing topology control. Designers can create bus guides interactively in the GUI or through TCL scripting, giving precise directional control over routing paths when needed.

The OmniRoute interconnect router provides multiple routing modes for sideband signals in UCIe designs. By default, all nets are routed together without distinguishing sideband signals. In Hug mode, sideband signals are routed adjacent to the edge of mainband signals. In Embed mode, sideband signals are routed together with mainband signals, while in Skip mode, sideband signals are not routed.

Auto PG Routing with Innovus Flash PG

Flash PG is an existing mature technology in Innovus Implementation System.

 It can be leveraged to enable automated PG routing for interposer bumps. It uses the PG Structure Description Language (PSDL) to capture the design intent along with the relation between the patterns and layers in a single PSDL file.

The PG engineer directly uses this file with the help of the route_pg command to generate the PG structure, and this process saves a significant amount of time. The PSDL file can be used to map the PG structure specifications in the design.

Interposer Auto PDNTraditionally, creating PG stripes relied on TCL-based scripts or manual editing, which were difficult to reuse because they depended on hard-coded coordinates and had to be modified whenever the bump map changed.

In this flow, PG is not treated as an afterthought; it is essential for robust PI and supports SI by providing the return path required to tie off shielding nets to the PG grid.

With the Flash PG-based auto PDN solution, Cadence has upgraded the PSDL syntax to support interposer bumps, including the ability to recognize bump cells, cluster them based on user-defined rules, and create stripes over the bumps using a follow-bump feature.

As a result, designers only need to define the routing rules up front, and Flash PG can automatically recognize the bumps and generate the PG stripes according to the specified routing pattern, making PG routing much easier, more reusable, and better aligned with automated signal routing in the broader OmniRoute-driven interposer flow.

AI-Driven SI Optimized Interposer Route using Cadence Cerebrus Explorer

Routing quality for interposer routing is generally checked post-routing by running extraction and simulation. This can take a few iterations before the route quality is optimized. Routing optimization involves defining parameters, running routing, extracting models, and simulating SI. This iterative process is time-consuming when done manually. Cadence Cerebrus Intelligent Chip Explorer automates this flow by driving auto-routing, model extraction, and SI simulation.

3DIC_Cadence

It learns from previous iterations to refine parameter combinations, converging on optimal results much faster than manual tuning. For example, on the case with five metal layers on a 65nm interposer, manual routing took 3–4 weeks to reach optimal results with manual iterations, whereas Cadence Cerebrus Explorer was able to exhaustively search and complete the process in just three days while iterating through 80 scenarios.

What This Means for Customers

For design teams, these capabilities translate into tangible benefits. Iteration cycles become dramatically shorter, allowing teams to explore more design options in less time. Eliminating manual routing effort improves productivity by a huge amount and reduces variability across designs. At the same time, integrated SI analysis increases confidence that the final implementation will meet performance requirements. Results show an 8-10X faster convergence rate compared to manual tuning.

Equally important, the solution scales with future design needs. As interposers continue to grow in size and complexity, and as standards like UCIe and HBM evolve, the Cadence OmniRoute Advanced Interconnect Router provides a foundation that can adapt to these changes without requiring a complete overhaul of the routing methodology.

Learn More | Request a Demo

If your team is exploring 2.5D, 3D, or 3.5D integration, request a design assessment or demo to see how the Cadence OmniRoute Advanced Interconnect Router in the Cadence Integrity 3D-IC Platform can streamline interposer routing, improve signal integrity analysis, and accelerate time-to-results.

  • Explore the Integrity 3D-IC Platform to see how unified design, analysis, and signoff can accelerate 3D-IC implementation.
  • Explore the Cadence Multi-Die 3D-IC Solution to learn how Cadence supports next-generation multi-die system design.
  • Talk to a 3D-IC Specialist to discuss your interposer routing challenges and roadmap.

In the next blog in this series, we'll take a closer look at silicon bridge design and explore how constraint-driven routing applies in embedded bridge contexts.


CDNS - RequestDemo

Try Cadence Software for your next design!

Free Trials

© 2026 Cadence Design Systems, Inc. All Rights Reserved.

  • Terms of Use
  • Privacy
  • Cookie Policy
  • US Trademarks
  • Do Not Sell or Share My Personal Information