UCIe/BoW BlueLynx™ Dual Mode PHY and subsystem IP for chiplet interconnect

Overview

BlueLynx™ is a revolutionary D2D subsystem architecture enabling best in class optimized Power-Performance-Area (PPA) and drastically reduces customer time-to-market.

BlueLynx™ interconnect IP subsystem includes physical (PHY) and optional link layer for chiplet interfaces and supports Universal Chiplet Interconnect Express (UCIe), OpenCompute Project (OCP) Bunch of Wires (BoW), as well as custom applications.
The BlueLynx™ link layers connect to on-die buses/Networks-on-Chip (NoCs) with various standards, including AXI4, AXI5 lite, CHI, ACE, and more.

Customers receive industry-standard ASIC integration views with reference silicon bring up platform and software required for first-pass silicon success in the shortest amount of time. Die-to-Die subsystem solutions have been delivered in over 7 different process nodes to date across multiple semiconductor foundries.

Key Features

  • High data rate of 2–24 Gb/s
  • Extremely low power of < 0.375 pJ/bit @ 2–16 Gb/s 0.5-V VDDQ
  • Scalable IO voltage for optimizing channel performance
  • Very low latency of < 2 ns PHY-to-PHY
  • Support for many serialization and deserialization ratios including 4:1, 8:1, 12:1 and 16:1
  • Support for bonded and independent slices
  • Side band and redundancy support
  • Staggered bump pitch for organic substrate (example macro above)
  • Bump pitch customizable from 45um (advanced) to 180um (standard) package applications
  • Each slice (TX+RX) macro area set by macro bump pitch (multiple slices can be stacked)
  • Support for N/S and E/W Poly Orientations
  • UCIe compatibility mode supports UCIe v1.1 FLIT Data Interface (FDI)
  • Programmable alignment between link layer and PHY clocks to aid timing closure
  • APB pclk pass-through to the link layer clock domain for reset of synchronous flops
  • Integrated, background-calibrated DLL with duty-cycle correction for input clock
  • Firmware-controlled startup and shutdown through control and status registers
  • JTAG and/or APB register access
  • Built in self-test (BIST) and diagnostic logic
  • On Demand eye plotter
  • Integrated microcontroller, PLL

Benefits

  • Flexibly Configurable:
  • BlueLynx™ PHY and link layer architecture is customizable to support a wide range of physical and complexed system bandwidth use cases.
  • Supports many bump pitches, metal stacks, and poly orientations.
  • Best in Class PPA:
  • Achieve superior link power efficiency with ultralow voltage (300mV) support
  • Operational Simplicity:
    • Flexible and robust APB bus and JTAG control interface
  • Future Proof Design:
    • Our partners stay ahead of the competition with speed -of-software BlueLynx™ Generator technology next generation IP delivery. BlueLynx™ integrated microcontroller to support evolving chiplet interface standards.

Block Diagram

UCIe/BoW BlueLynx™ Dual Mode PHY and subsystem IP for chiplet interconnect Block Diagram

Applications

  • Standard Organic and very advanced 2.5D package applications
  • On-die Buses/Networks-on-Chip (NoCs) with various standards, including AXI4, AXI5 lite, CHI, ACE, and more.

Deliverables

  • Hard macro with soft digital RTL
  • IBIS model
  • Integration Support
  • Verilog Behavioral Model
  • Timing Libraries
  • Soft Digital RTL
  • LEF
  • CDL netlist
  • DRC & LVS reports
  • GDSII
  • RTL Timing Constraints
  • DFT Support for Scan

Technical Specifications

Foundry, Node
16nm, 12nm, 5nm, 4nm, 3nm in multiple foundries including TSMC and Samsung
Samsung
Pre-Silicon: 4nm
TSMC
Pre-Silicon: 3nm , 4nm , 5nm , 6nm , 7nm , 10nm , 12nm , 16nm
×
Semiconductor IP