Skip to main content

Xsight Labs Partners with Edgecore Networks to Bring World’s First 800G DPU to a PCIe® Add-In Card Form Factor

Xsight’s E1-AIC features 64 Arm® N2 Cores to deliver up to 800 Gbps throughput, redefining performance, efficiency, and adaptability. It offers Host Virtualization/Offloading for AI security, storage, and networking.

  • Utilizing Xsight Labs’ E1 SoC, the Edgecore Networks Add-In Card (AIC) delivers unprecedented bandwidth with the industry’s first 800G DPU, incorporating 64 Arm N2 Cores in a low power PCIe AIC form factor.
  • The E1-AIC delivers high performance, lower latency, and greater efficiency for today’s most demanding AI, security, storage, and networking workloads.
  • The E1-AIC utilizes innovative Software Defined Networking (SDN) and a fully programmable RoCEv2 software stack to enhance data center performance, efficiency, and security.

Xsight Labs and Edgecore Networks today announced availability of the first 800G PCI Express® (PCIe®) add-in card (AIC), delivering unprecedented performance to hyperscale AI workloads. The E1 AIC is purpose-built to meet the rapidly growing demands of AI, cloud and high-performance computing (HPC) workloads, extending Xsight Labs’ leadership in high-performance intelligent connectivity. The E1 AIC delivers unmatched bandwidth, low latency, and advanced data offload capabilities designed to allow greater utilization of accelerated compute and GPU clusters.

Why It Matters

The gigawatt-scale era requires exponential growth in network bandwidth, model complexity, and cluster sizes. As agentic AI, large language models (LLMs), and distributed training continue to push the limits of data center infrastructure, traditional CPUs and NICs are bottlenecks. As Neo-clouds offer greater specialization and multi-tenancy, advanced security becomes vital. Xsight Labs’ new DPU (Data Processing Unit) add-in card is engineered to overcome these challenges — enabling smarter, secure, faster, and more efficient data processing from the edge to the large-scale AI and data center infrastructure.

Building a New DPU Solution

Xsight Labs, in collaboration with Edgecore Networks has produced the industry’s first PCIe add-in card delivering 800 Gbps throughput (2x400G, 4x200G, 8x100G) for PCIe 5.0 Host Systems. It is designed to be ARM System Ready Level 6 and can work with virtually any standard server. The E1 AIC offers the greatest compute density per watt of alternative DPUs, enabling up to 320 Arm N2 cores per 19” Rack Unit (RU) in an air-cooled data center.

“As evidenced by merchant-built revenues more than doubling year-over-year, the DPU market is moving beyond a few large hyperscalers building their own in-house products," said Seamus Crehan, president of Crehan Research. "Expansion of existing use cases, in addition to newer applications including generative AI front-end, DPU-enabled Ethernet switches, and CPU replacement, are driving this strong growth. With its E1-AIC offering a compelling set of features at the industry’s highest DPU speed of 800G, Xsight Labs is delivering the right product at the right time to enable customer success and differentiation."

From a security perspective, the E1-AIC is a self-contained PCIe add-in card with its own 64 Arm® N2 cores and can run workloads independently of the host x86 CPU. This creates an air gap between the DPU and the x86 core so sensitive operations like packet inspection, encryption/decryption, and traffic analysis can be executed entirely on the card without exposing them directly to the host CPU’s attack surface, enabling more robust, stateful security at scale.

Powering Workloads with the E1

The E1-AIC utilizes innovative Software Defined Networking (SDN) and Software Defined Storage (SDS), resulting in enhanced data center performance, efficiency, and security. The E1-AIC also features a fully programmable RoCEv2 (RDMA over Converged Ethernet) software stack. Bypassing the CPU and handling data movement directly with RoCEv2, the E1-AIC delivers faster transfers, low latency, and greater efficiency for high-performance computing, cloud, AI/ML workloads, and data-intensive applications. The E1 800G DPU completely passed the SONiC-DASH full-scale Hero 800G test with zero dropped packets at 14.25 million connections per second, validating the industry-leading performance and readiness of the E1 for robust at scale deployments.

For a full list of the key features, view the E1-AIC Product Brief.

Perspectives:

“We are excited to introduce the world’s first 2x400G QSFP-112 Data Processing Unit, designed for host virtualization across security, networking, storage, and even ARM-based general compute— delivered in a full-height, three-quarter length form factor powered by Xsight Labs’ Edge Server Infrastructure SoC,” said Mingshou Liu, President of Edgecore. “As cloud and AI workloads continue to scale, we see opportunities to accelerate storage virtualization at the host and eliminate storage bottlenecks at the front end, moving seamlessly from 200G and 400G to 800G over Ethernet networks.”

"Xsight Labs' E1-AIC uniquely combines the openness of an Arm server with 800 Gbps of network throughput," said Bob Wheeler, Principal Analyst at Wheeler's Network. "This combination makes the E1-AIC a powerful building block for a wide variety of SDN applications using open-source stacks such as DPDK and SONiC."

About Xsight Labs

Xsight Labs is a leading fabless semiconductor company providing intelligent connectivity solutions for next-generation hyperscale, edge and AI data center networks. Xsight Labs’ technology delivers exponential bandwidth growth and unmatched versatility while lowering power and total cost of ownership. Founded in 2017, Xsight Labs is headquartered in Kiryat Gat, Israel, with additional offices in Tel-Aviv and Haifa, and international offices in Yerevan, Armenia, as well as in Boston, Raleigh, and San Jose in the United States. For additional information, visit www.xsightlabs.com.

Contacts

Stock Quote API & Stock News API supplied by www.cloudquote.io
Quotes delayed at least 20 minutes.
By accessing this page, you agree to the following
Privacy Policy and Terms Of Service.