rmNVMe-IP, Random Access plus Multi User NVMe Host

Combine random I/O optimization with concurrent multi user access in a single hardware NVMe host that does not require a CPU or external DDR.

 

rmNVMe-IP is built for systems that need both high random IOPS and multiple data streams at the same time. The core schedules commands and moves data in pure logic, which delivers predictable latency under mixed workloads while keeping interfaces simple for RTL integration. It suits event capture with live analytics, shared scratch storage at the edge, and pipelines that interleave frequent small updates with larger transfers.

 

 

Highlights

  • Hardware NVMe host optimized for random I/O with multi user concurrency
  • No CPU and no external DDR required
  • Configurable buffering to tune for read heavy, write heavy, or mixed access patterns
  • Simple FIFO or AXI-style user interfaces for control and data
  • Compact resource usage compared to CPU based stacks
  • Reference designs available to shorten bring up

 

 

Why Macnica?

  • Architecture reviews and sizing guidance for Agilex platforms
  • SSD topology recommendations for target IOPS, queue depth, and stream count
  • Evaluation adapters and kits to validate quickly
  • Supply chain support from prototype through production

 

Macnica + Toppan

 

 

 

Key Specifications

  • Storage protocol: NVMe to NVMe SSDs
  • Host resources: Pure RTL implementation, no external DDR, no embedded CPU
  • Concurrency: Multiple user interfaces can access the same SSD with controlled arbitration
  • User interfaces: Register control and streaming data paths for random reads and writes
  • Integration: Works with Altera PCIe hard IP on supported boards
  • Collateral: Getting started notes, reference design, and test guidance available

 

Typical Applications

  • Event capture with live analysis on shared NVMe
  • Multi-camera or multi sensor recorders that need simultaneous streams
  • Mixed analytics and logging where random updates compete with sequential transfers
  • Edge scratch storage shared by parallel pipelines

 

 

 

Evaluation Hardware

  • AB20-U2PCI U.2 adapter for up to four U.2 or U.3 NVMe SSDs
  • AB19-M2PCI M.2 adapter for up to four M.2 NVMe SSDs
  • AB17-M2FMC FMC adapter for up to two M.2 NVMe SSDs

 

Supported Platforms

  • Altera Agilex evaluation boards via provided reference designs
  • Contact us to confirm device coverage, resource fit, and SSD recommendations

 

 

 

Get Started

 

Tell us your number of streams, target IOPS, queue depth, and SSD count. We will recommend a sizing path, a demo setup, and the right adapter set for your Agilex platform.

 

 

 

 

Similar Products