Compute Express Link (CXL) FPGA IP

Overview

Industry's first FPGA-based hardened CXL IP solution for Type 1, 2, and 3 devices.

First FPGA to pass CXL Consortium Compliance Program (up to 32 GT/s speed).

CXL Protocol Hard IP Functionality Accelerates a Wide Range of Data-Centric Workloads

  • Accelerated Computing: CXL seamlessly connects FPGAs to accelerators and GPUs, enhancing performance in compute-intensive workloads like AI, machine learning, data analytics, and scientific simulations.
  • Memory Expansion: CXL enables FPGA-based systems to extend their memory capacity by connecting to CXL-compatible memory devices, improving performance in memory-bound applications like memory databases and large-scale data processing.
  • Data Processing: FPGAs with CXL support are used in data processing workloads, including real-time data analytics, financial modeling, and high-frequency trading, where low latency and high bandwidth are critical.

Key Features

  • Fully integrated CXL IP solution with PCIe 5.0 x16 (32 GT/s) performance.
  • CXL 1.1 compliant IP that combines both hard and soft resources:
    • R-Tile Hard IP (PHY IP) reduces logic resourcing and simplifies timing closure.
    • Soft wrapper for added design flexibility and easy integration.
  • Validated with 4th and 5th Gen Intel® Xeon® Scalable Processors.
  • Compute Express Link Consortium compatible with CXL certification software (CXL_CV_app).
  • CXL protocol support: CXL.io, CXL.mem, CXL.cache.
  • Device type support: Type 1, 2, and 3.
  • CXL 1.1. IP enhanced with 2.0 features and functionality.
  • Flexible device-attached memory subsystem architecture:
    • Allows FPGA developer to mix and manage different memory types for maximum scalability and application fit.

Block Diagram

Compute Express Link (CXL) FPGA IP Block Diagram

Technical Specifications

Short description
Compute Express Link (CXL) FPGA IP
Vendor
Vendor Name
×
Semiconductor IP