Intel IPU E2200 Adapter (Mount Morgan)

The Intel IPU E2200 Adapter, codenamed Mount Morgan, is a purpose-built Infrastructure Processing Unit (IPU) designed to accelerate and offload critical data-center functions—including networking, storage, and security—away from the host CPU. Built on Intel's IPU architecture, the E2200 delivers up to 400Gbps of throughput while exposing a programmable pipeline that enables operators to deploy stateful firewall, NVMe-oF storage offload, RDMA, and telemetry functions as software-defined services running directly on the adapter. This fundamentally changes CPU utilization economics in large-scale cloud and AI infrastructure by reclaiming host cores that would otherwise be consumed by infrastructure overhead.

Ideal for

• AI/ML GPU cluster fabric acceleration: offloading RDMA over Converged Ethernet (RoCEv2) transport from host CPUs in GPU nodes to sustain full 400Gbps all-reduce collective communication bandwidth for large-model training
• NVMe-oF disaggregated storage: enabling CPU-less NVMe-oF target and initiator termination so storage servers present shared NVMe namespaces over Ethernet with sub-10µs added latency without consuming host CPU cycles
• Inline network security for multi-tenant cloud: running hardware-accelerated IPsec, TLS 1.3, and MACsec at line rate on the IPU to enforce tenant-traffic encryption and micro-segmentation without a software firewall tax on the host
• 5G core and carrier edge NFV: hosting virtual network functions (UPF, SMF, AMF packet forwarding) directly on the IPU ARM cores to achieve carrier-grade throughput with deterministic latency in a PCIe form-factor appliance
• Telemetry and observability at scale: leveraging the P4 programmable pipeline to implement per-flow INT (In-band Network Telemetry) and streaming metrics to an external collector without adding CPU overhead on production servers
• Storage QoS and multi-path management: executing NVMe-oF multipath failover, I/O queuing, and quality-of-service enforcement on the IPU to guarantee latency SLAs for latency-sensitive database and AI inference workloads

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.