Enquire: Asia & Africa - +65-98008081 USA - +1-919-995-4114

Home Networking Broadcom N1800GO 800GbE OCP 3.0 Ethernet Adapter

Broadcom N1800GO 800GbE OCP 3.0 Ethernet Adapter

Brand:

Broadcom N1800GO is a single-port 800GbE OCP 3.0 Ethernet adapter built for AI/ML data centers, HPC, cloud infrastructure, and high-performance storage. Based on the BCM57708 Ethernet controller, it combines PCIe 6.0 x16 host connectivity with an OSFP112 network port and line-rate 800GbE throughput. Advanced RoCEv2, packet-level multipathing, congestion control, telemetry, security, and hardware acceleration make it suitable for next-generation XPU servers and high-bandwidth AI networks.

OVERVIEW

  • Broadcom N1800GO is a single-port 800GbE OCP 3.0 AI Ethernet adapter with PCIe 6.0 x16 host connectivity and OSFP112 networking. Compatibility: requires a qualified OCP 3.0 server with PCIe 6.0 x16 support. Model comparison: N1800GO is the OCP 3.0 version, while P1800GO is the PCIe NIC variant. Need: select N1800GO for high-bandwidth AI, HPC, cloud, and storage servers requiring 800GbE.

USE CASES

  • AI training clusters – Provides 800GbE line-rate connectivity for high-volume communication between GPU/XPU servers and the network fabric.
  • AI inference infrastructure – High network throughput helps move model inputs, outputs, and distributed inference traffic efficiently across accelerated servers.
  • HPC clusters – Supports high-bandwidth Ethernet networking for distributed scientific and technical computing workloads.
  • GPU/XPU server networking – Peer Memory Direct and advanced RoCE capabilities help optimize communication between accelerators and network resources. 
  • Cloud-scale data centers – OCP 3.0 integration provides a standardized server-adapter approach for high-density cloud infrastructure.
  • Disaggregated storage – RoCEv2 and high-speed Ethernet provide a connectivity option for demanding storage environments and data-intensive server architectures.
  • AI back-end fabric – Packet-level multipathing, adaptive routing, dynamic load balancing, and congestion-control capabilities address the communication requirements of large distributed AI systems. 
  • Private cloud infrastructure – Advanced virtualization, tunneling, flow processing, and network management features support modern software-defined data-center environments.

KEY FEATURES

  • 800GbE single-port connectivity – Delivers up to 800Gb/s Ethernet bandwidth for high-throughput AI, HPC, and cloud servers.
  • PCIe 6.0 x16 host interface – Provides a high-bandwidth host connection designed to feed an 800GbE network interface without relying on an older PCIe generation. 
  • OSFP112 network connector – Provides the physical interface for supported 800GbE/400GbE and other link-speed configurations. 
  • Multiple Ethernet speeds – Supports 800G, 400G, 200G, 100G, 50G, and 25G link speeds, giving infrastructure teams greater deployment flexibility. 
  • Advanced RoCEv2 – Provides RDMA-based networking features optimized for AI/ML communication and other latency-sensitive workloads. 
  • Packet-level multipathing – Provides additional path-management capabilities for distributed AI traffic, helping improve network utilization and resilience. 
  • Out-of-order packet delivery – Helps the adapter process packets more flexibly in demanding RoCE workloads.
  • Hardware selective retransmission – Moves supported retransmission processing into the adapter, helping reduce host CPU involvement.
  • Programmable congestion control – Gives network operators more control over congestion behavior in large-scale AI fabrics.
  • Advanced telemetry – Provides network-performance visibility to help operators monitor and troubleshoot high-speed AI infrastructure. 
  • Peer Memory Direct – Enables optimized access paths for peer memory, supporting low-latency communication in accelerator-based systems. 
  • TruFlow hardware acceleration – Offloads supported virtual-switch and flow-processing functions, helping reduce CPU overhead in virtualized networking environments.

TECHNICAL SPECIFICATIONS

  • Brand: Broadcom
  • Product: N1800GO
  • Product Type: 800GbE Ethernet Network Adapter
  • Adapter Type: OCP 3.0
  • ASIC: BCM57708
  • Port Configuration: 1 × 800GbE
  • Maximum Network Speed: 800GbE
  • Supported Link Speeds: 800/400/200/100/50/25GbE
  • Network Connector: OSFP112
  • Host Interface: PCIe 6.0 x16
  • PCIe Generation: Gen 6
  • PCIe Lane Configuration: x16, x8, x4, x2, x1 supported by ASIC
  • SerDes: 100/50G PAM4 and 25G NRZ
  • RoCE: RoCEv2
  • Advanced RoCE: Yes
  • TruFlow: Yes
  • TruManage: Yes
  • Multi-Host: Yes
  • NIC Partitioning: 8 PFs
  • SmartNIC: No
  • Security: Secure Boot, Silicon Root of Trust, SPDM attestation
  • Encryption: Hardware-based encryption/decryption with PSP
  • Network Boot: UEFI PXE, UEFI L2 iSCSI boot

Why Choose This Product?

  • 800GbE bandwidth – Provides an extremely high-speed Ethernet connection for AI clusters and next-generation data-center fabrics.
  • PCIe 6.0 host architecture – Matches the adapter's high network bandwidth with a modern host interface designed for next-generation server platforms.
  • AI-focused networking – Advanced RoCEv2 features address the high communication demands of distributed AI and ML workloads.
  • Flexible link speeds – Supports 25G through 800G, allowing the adapter to operate across multiple network-speed environments. 
  • Strong security architecture – Silicon Root of Trust, secure boot, attestation, and hardware security processing provide a stronger foundation for enterprise AI infrastructure.
  • CPU offload capabilities – TruFlow and other hardware acceleration engines can reduce host processing requirements for supported network functions.
  • OCP 3.0 deployment model – Designed for modern server architectures where OCP NICs enable flexible, high-density networking.
  • Future-ready AI infrastructure – The 800GbE interface and PCIe 6.0 architecture provide substantial bandwidth headroom for evolving AI and HPC server designs.

CUSTOMER REVIEW

No Reviews Found

FAQ

No FAQs Found