Intel Agilex® 7 F-Series and I-Series FPGA Memory Subsystem IP User Guide

ID 789389
Date 12/04/2023
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

3. Memory Subsystem IP Architecture and Feature Description

The memory subsystem IP includes a configurable, hardened protocol stack for external memory interfaces that allows you to configure from 1 to 8 EMIF instances, all implemented based on Intel Agilex EMIF IP core designs. The goal of the memory subsystem is to provide efficient ways to instantiate EMIF designs and provide application-level optimizations.

Memory subsystem build flow automatically identifies and connects subsystem internal building blocks to form the base subsystem and likewise for extended/vertical subsystem. For example, a memory base subsystem would automatically identify and connect the specific calibration block to the external memory instance(s) according to your specified subsystem placement of the physical external memory instance(s). The subsystem infrastructure comprises foundational IPs and standard protocol busses over an AXI interface.

The following figure shows the subsystem IP block diagram, with important blocks and their connections:

Figure 4. Memory Subsystem IP Block Diagram

Each memory subsystem can contain single or multiple instances of memory IPs. You can connect subsystems with multiple IPs through the Platform Designer Interconnect. The subsystems can contain 0 or more lookup IPs and single or multiple user interfaces (AXI-MM). Each interface has dedicated clock and reset domains.

Although each usage has unique requirements, the subsystems share scalability, composeability, and a certain degree of infrastructure.