D2D PHY
A D2D PHY (Die-to-Die Physical Layer) is a high-speed physical communication layer that enables data exchange between multiple semiconductor dies integrated within the same package. D2D PHYs are a key building block in modern chiplet-based architectures, 2.5D/3D integration, and advanced heterogeneous computing systems.
The D2D PHY represents the lowest hardware communication layer of a die-to-die interconnect. It manages electrical signaling between dies, including data transport, clocking, synchronization, signal integrity, power management, and calibration mechanisms.
With the rapid growth of artificial intelligence, high-performance computing (HPC), and modular semiconductor architectures, die-to-die interconnects have become essential for overcoming the limitations of traditional monolithic SoC designs.
Background and Evolution
For decades, semiconductor devices were primarily designed as monolithic chips where all functional blocks were integrated on a single die. However, increasing design complexity, rising advanced-node manufacturing costs, and reticle size limitations accelerated the adoption of multi-die architectures.
Chiplet-based designs provide several advantages:
- Improved manufacturing yield
- Better cost optimization
- Reuse of existing IP blocks
- Faster time-to-market
- Ability to combine multiple process technologies within a single package
In these architectures, D2D PHYs provide the high-speed communication infrastructure connecting the various dies together.
D2D PHY Architecture
A D2D PHY typically consists of several major functional blocks.
Transmitters and Receivers
The PHY integrates:
- Transmitters (TX)
- Receivers (RX)
- Output drivers
- Equalization circuits
- Termination circuits
These blocks ensure reliable data transmission over very short distances inside the package.
Clocking and Synchronization
D2D interfaces may use various synchronization mechanisms:
- Source-synchronous clocking
- Embedded clocking
- Clock forwarding
- Strobe forwarding
- DLLs and PLLs
Because die-to-die links operate over extremely short channels, they can support very high operating frequencies with low latency.
Training and Calibration
Modern D2D PHYs include sophisticated calibration and training features such as:
- Lane deskew
- Eye training
- Lane repair
- Thermal compensation
- PVT (Process, Voltage, Temperature) compensation
These mechanisms help maintain reliable communication under varying operating conditions.
Power Management
Energy efficiency is a critical requirement for D2D PHYs, particularly in AI and HPC applications.
Common power optimization techniques include:
- Dynamic voltage scaling
- Power gating
- Low-power idle states
- Adaptive link management
Packaging Technologies
D2D PHYs are closely associated with advanced semiconductor packaging technologies.
2.5D Integration
In 2.5D architectures, multiple dies are interconnected through a silicon interposer.
Typical applications include:
- GPUs with HBM memory
- AI accelerators
- High-performance FPGAs
3D IC
3D IC technologies vertically stack multiple dies using:
- Through-Silicon Vias (TSVs)
- Hybrid bonding
Benefits include:
- Extremely high bandwidth
- Reduced latency
- Higher integration density
Organic Substrate Packaging
Some D2D PHY implementations operate directly on organic substrates without silicon interposers.
This approach is generally more cost-effective but introduces greater signal integrity challenges.
Major D2D Standards
Several industry standards have emerged to standardize chiplet interconnect ecosystems.
UCIe
Universal Chiplet Interconnect Express (UCIe) Consortium is currently the leading open standard for die-to-die communication.
The specification defines:
- PHY layer requirements
- Protocol stack
- Software interoperability
- Multi-vendor compatibility
UCIe supports multiple packaging approaches, including:
- Organic substrates
- Silicon interposers
- Advanced packaging technologies
BoW
BoW (Bunch of Wires), developed by the Open Compute Project, provides a simplified low-power die-to-die interface approach.
BoW focuses on:
- Low complexity
- Reduced power consumption
- Short-reach communication
Proprietary D2D PHYs
Many semiconductor vendors also develop proprietary D2D PHY technologies optimized for:
- Maximum bandwidth
- Low latency
- High density
- Superior energy efficiency
Featured Content
- Modeling, Optimizing and Exploring Multi-Die FPGA Routing Architectures
- A 32 Gb/s 0.41 pJ/bit Single-Ended Transmitter with TX-Based-Only Adaptive Crosstalk Cancellation for Ultra-Short-Reach Wireline Applications
- JCET Opens New High-Density 3D System Integration Facility
- Kandou AI Demonstrates Revolutionary 260 Gbps Tigerwing™ Chip-to-Chip Interface in Silicon at 2026 TSMC Europe Technology Symposium
- InPsytech Highlights UCIe Innovation at COMPUTEX with UCIe Live Demo and Ultra-high speed ONFI 6400 Development
- The Evolution Of UCIe
- Photonics: A Foundational Scaling Layer for AI-Era Computing
- Ayar Labs Joins NVIDIA NVLink™ Fusion Ecosystem to Bring Co-Packaged Optics to Rack-Scale AI Infrastructure
- Lightmatter Joins NVIDIA NVLink Fusion and Powers Next-Generation AI Infrastructure with Photonic Interconnects
- Sivers & GlobalFoundries Advance AI Data Center Optical Solutions
- Transient Multiscale Workflow for Thermal Analysis of 3DHI Chip Stack
- From horsepower to high-performance compute: automotive chiplets take the leap towards autonomous edge computing
- Dispersion-Engineered Terahertz Silicon Interconnects Enabling Terabit-Scale Data Links
- Wiwynn and Ecosystem Partners to Showcase Co-Packaged Optics Innovations at Computex 2026
- CEA-Leti Presents Die-to-Wafer Hybrid Bonding At 1 μm Pitch, Removing Bottleneck for AI Hardware