HBM4 Memory Controller

Overview

Rambus is offering the industry’s first HBM4 Controller IP core designed to support customers with deploying a new generation of HBM memory for cutting-edge AI accelerators, graphics and high-performance computing (HPC) applications.

How the HBM4 Controller Works

The core accepts commands using a simple local interface and translates them to the command sequences required by HBM4 devices. It also performs all initialization, refresh and power-down functions. The core queues up multiple commands in the command queue. This enables optimal bandwidth utilization for both short transfers to highly random address locations as well as longer transfers to contiguous address space. The command queue is also used to opportunistically perform look-ahead activates, precharges and auto-precharges, further improving overall throughput. The Reorder functionality is fully integrated into the controller command queue increasing throughput and minimizing gate count.

The controller can be delivered standalone or integrated with the customer’s choice of PCIe 6.2.1 PIPE compliant SerDes.

Key Features

  • Supports HBM4 memory devices
  • Supports all standard HBM4 channel densities (up to 32 Gb)
  • Supports up to 10 Gbps/pin
  • Refresh Management (RFM) support
  • Maximize memory bandwidth and minimizes latency via Look Ahead command processing
  • Integrated Reorder functionality
  • Achieves high clock rates with minimal routing constraints
  • Self-refresh and Power-down Low Power Modes
  • Support for HBM4 RAS features
  • Built-in hardware-level performance Activity Monitor
  • DFI compatible
  • End-to-end data parity
  • Support AXI or native interface to user logic

Block Diagram

HBM4 Memory Controller Block Diagram

Deliverables

  • Controller (Source Code)
  • Testbench (Source Code)
  • Complete Documentation
  • Expert Technical Support
  • Maintenance Updates

Technical Specifications

Foundry, Node
Any
Availability
Now
×
Semiconductor IP