HBM3 PHY IP at 7/6nm

Overview

Features a state-of-the-art mixed-signal architecture that addresses the challenges of DRAM integration in high-performance and low-power environments. This architecture enables OPHYs to overcome issues with long-term impedance drift and clock phase drift, enabling impedance and clock phase updates without interrupting data traffic. Programmable timing at the OPHY boundary combines flexibility with analog precision, resulting in low read/write latency between the ORBIT Memory Controller (OMC) and the DRAM.

OPHYs are designed with subsystem and system-level considerations in mind. Built-in power management logic and advanced PLL design allow aggressive power state management and optimal system power usage. Tight integration with the ORBIT Memory Subsystem enables ActiveQoS bandwidth and latency control for maximum performance of the SoC memory subsystem. At the system level, OPHYs have been designed to minimize package substrate layer and PCB layer requirements, enabling usage in cost-sensitive applications.

Key Features

  • Compliant with PHY standards
    • JEDEC compliant LPDDR5X/5/4X/4, DDR5, GDDR6, HBM3 support
    • DFI Interface Compliant
  • ​Flexible Configuration
    • LPDDR54: 8-/16-/32-bit data width per channel
    • GDDR6: 16-bit data width per channel; pseudo-channel mode
    • Supports multiple DFICLK: CK: WCK ratio
    • Multiple DFICLK: CK: WCK ratios
    • Up to 4 ranks with Tx and Rx channel equalization
  • ​Maximum Data Rates
    • Up to 8533 Mbps data rate for LPDDR5x
    • Up to 16 Gbps data rate for GDDR6
  • ​Programmable State Machine (PSM)
    • Proprietary microcontroller and custom ISA enable customizable DFT features and multiple LPDDR standard support efficiently while reducing the area
  • ​Multiple FSPs and the LP States
    • Supports up to 4 frequency set points (FSPs)
    • Supports multiple low power states for system power optimization

Benefits

  • Configurability with Flexible Applications
    • Configurable channel and floor-plan allow connection to different DRAM package types and lane ordering
    • Minimal package substrate/PCB layer requirements enables PHY usage in low-cost applications
  • ​Performance
    • PSM enables accelerated firmware-based training
    • Ultra-fast fractional training
    • Programmable PHY boundary timing provides low read/write latency
  • Fast switching between FSPs
  • ​Capacity
    • Channel equalization and fast timing adjustment circuits enable 4 rank support to maximize capacity
  • ​Power
    • Power-saving modes with a variety of exit times
    • Multiple voltage domains to optimize voltage versus frequency

Block Diagram

HBM3 PHY IP at 7/6nm Block Diagram

Applications

  • Provides an efficient solution for applications requiring high memory bandwidth, particularly in the fields of artificial Intelligence (AI), Machine Learning (ML), and general-purpose graphics processing units (GPGPU).

Deliverables

  • Hard & Soft IP
    • GDSII, LEF, LVS, timing models, etc.
    • Verilog behavior models and encrypted RTL
    • Synthesis and STA constraints
    • Example test benches
  • Documentation
    • PHY Technical Reference Manual
    • Implementation, package, and PCB design guidelines
    • Test and characterization guidelines
    • Physical verification reports

Technical Specifications

Foundry, Node
7nm, 6nm
Maturity
Silicon-proven
Availability
Now
×
Semiconductor IP