Products2024-09-20T13:56:19+00:00

Solutions That Power the Next-Generation of Cloud Infrastructure

Solutions That Power the Next-Generation of Cloud Infrastructure

Solutions That Power the Next-Generation of Cloud Infrastructure

Xsight Labs’ innovative X-Switch is designed to revolutionize how you optimize data center workloads, ensuring your AI clusters operate at peak efficiency. With key features tailored specifically for AI workloads, the X-Switch family of data center switches are architected to deliver unparalleled benefits.

Xsight Labs’ innovative X-Switch is designed to revolutionize how you optimize data center workloads, ensuring your AI clusters operate at peak efficiency. With key features tailored specifically for AI workloads, the X-Switch family of data center switches are architected to deliver unparalleled benefits.

Ultra-low Power Efficiency to Maximize Bottom-Line Savings

Ultra-low Power Efficiency to Maximize Bottom-Line Savings

At the forefront of energy efficiency, the X-Switch family of data center switches sets a new benchmark in ultra-low power efficiency, designed to maximize energy savings and minimize environmental impact without compromising performance or reliability. As the industry’s most energy-efficient 12.8 terabit switch, X-Switch data center switches redefine industry standards by delivering unparalleled performance while consuming just 200 watts, compared to the typical 350 watts of other switches. This remarkable 40% reduction in power usage per switch translates to an estimated savings of $300 to $400 million over five years. Cloud service providers can expect to achieve substantial operational savings and OpEx benefits while minimizing their AI infrastructure’s environmental footprint.

At the forefront of energy efficiency, the X-Switch family of data center switches sets a new benchmark in ultra-low power efficiency, designed to maximize energy savings and minimize environmental impact without compromising performance or reliability. As the industry’s most energy-efficient 12.8 terabit switch, X-Switch data center switches redefine industry standards by delivering unparalleled performance while consuming just 200 watts, compared to the typical 350 watts of other switches. This remarkable 40% reduction in power usage per switch translates to an estimated savings of $300 to $400 million over five years. Cloud service providers can expect to achieve substantial operational savings and OpEx benefits while minimizing their AI infrastructure’s environmental footprint.

Flexible Performance with Xsight Labs’ Programmable Switch Architecture

Flexible Performance with Xsight Labs’ Programmable Switch Architecture

The X-Switch’s architecture enables lightning-fast packet processing through the switch fabric, ensuring low latency and seamless data flow within your AI clusters. The X-Switch provides the industry’s most flexible platform, allowing administrators to define packet processing rules, routing protocols, and network policies, enabling tailored optimizations for specific applications or workloads. Accelerate data processing and improve your AI applications’ responsiveness to maximize network utilization, saving you time and money.

X-IQ™ for Reduced Tail Latency and Improved Flow Completion

X-IQ™ for Reduced Tail Latency and Improved Flow Completion

Say goodbye to tail latency and hello to improved flow completion times. X-IQ’s intelligent queuing dynamically optimizes packet processing, minimizing delays and maximizing throughput for your AI tasks. In AI clusters, where rapid data processing is crucial, high tail latency can disrupt real-time applications, introduce inconsistencies in processing times, and lead to bottlenecks in distributed systems. The X-Switch family’s intelligent queuing not only optimizes AI application performance with lower flow completion time (FCT) but also leads to significant CapEx and OpEx savings by reducing the need for over-provisioning of resources and minimizing operational costs associated with inefficient networks.

X-View™ Real-time Telemetry for Congestion Avoidance in AI Fabrics

X-View™ Real-time Telemetry for Congestion Avoidance in AI Fabrics

Real-time telemetry allows for congestion avoidance, ensuring smooth and efficient operation of your AI fabrics. X-View provides hyperscalers with real-time insights into network metrics like bandwidth, packet loss, and latency issues across AI fabrics. X-View’s real-time telemetry empowers cloud operators to proactively manage resources and optimize network efficiency by swiftly identifying and resolving issues, resulting in faster processing, reduced latency, and enhanced scalability. With X-View, the X-Switch family maximizes AI cluster performance and cost-effectiveness, leading to significant savings in both CapEx and OpEx by avoiding costly upgrades and downtime.

Packet Spraying and Reordering to Eliminate Need for Extensive Buffering

Packet Spraying and Reordering to Eliminate Need for Extensive Buffering

Bid farewell to extensive buffering requirements. Our packet spraying and reordering technology eliminates the need for bulky buffers, streamlining data transmission and enhancing overall cluster performance essential for the rapid data processing needs of AI clusters. The X-Switch family strategically distributes data packets across network paths, minimizing buffering needs and reducing latency, resulting in enhanced cluster performance and responsiveness, optimized network resources, and reduced computational overhead, while minimizing expenses in AI cluster deployments.

Experience the power of the X-Switch architecture and unlock the full potential of your AI clusters. Maximize network utilization, minimize latency, and elevate your AI projects to new levels of performance and efficiency with our cutting-edge switch architecture.

Experience the power of the X-Switch architecture and unlock the full potential of your AI clusters. Maximize network utilization, minimize latency, and elevate your AI projects to new levels of performance and efficiency with our cutting-edge switch architecture.

X2 Programmable Ethernet Switch

X2 Programmable Ethernet Switch

X2 is the next-generation chipset in the X-Switch family of fully programmable Ethernet switches. Supporting up to 12.8Tbps full-duplex, non-blocking throughput, X2 is optimized for interconnecting AI, storage, and compute clusters in data centers.

The combination of a programmable packet processor and the X-PNDTM engine allows for flexible sharing of control tables. X2 also supports application-specific optimization, emerging protocols, and vendor-specific features.

  • At less than 200W of power – 40% lower power consumption compared to other 12.8Tbps switches in the market today – X2 enables cost and power-optimized systems that drive lower CapEx and OpEx.
  • Industry-leading 100G PAM4 LR SerDes, enabling designs with in-rack DAC connectivity without the need for re-timers. Future-proofed for 800G optics ecosystem.
  • Supports 50, 100, 200, and 400GbE port speeds.
  • Network port configuration of 128 LR SerDes @ 50Gbps or 100Gbps+ enables support for 4-meter DAC without retimers. + 800Gbps optical transceivers.

Key Features

Key Features

The application-optimized switching architecture and parallel hardware acceleration engines, powered by a programmable packet header processor, deliver 256B parsing and packet editing capabilities. This results in high performance, reduced processing latency, and increased power efficiency.

Elastic resource allocation of control tables, maximizing adaptation to application needs.

Elastic resource allocation of control tables, maximizing adaptation to application needs.

Fully shared packet buffer for incast burst handling.

Fully shared packet buffer for incast burst handling.

700ns first-bit-in-first-bit-out for typical data center applications.

700ns first-bit-in-first-bit-out for typical data center applications.

X1 Programmable Ethernet Switch

X1 Programmable Ethernet Switch

X1 data center switch delivers 25.6Tbps and 12.8Tbps of fully programmable application-optimized switching for top-of-rack, data center fabric, and HPC and AI clusters. It is the industry’s first, low power 25.6Tbps (32 x 800G) data center switch with 100G SerDes and is designed from the ground up to address the bandwidth, power, form factor, and radix requirements for current and next-generation cloud deployments and hyperscale networks.

X1’s architecture incorporates e application-optimized switching, X-PND™, and X-IQ™ enabling customers’ switch deployments to achieve optimized latency and power efficiency.

  • Built with industry-leading 256 x 100G LR PAM4 SerDes, X1 enables retimer-less design, supporting in-rack DAC connectivity for the 100G ecosystem. X1 supports 800G optics ecosystem, delivering a 2x improvement in port density vs. 400G optics
  • Fully shared buffer architecture enables wire-speed in-cast performance for worst-case microburst absorption
  • Extensive suite of programmable Data Plane Telemetry for best-in-class network visibility for monitoring, troubleshooting, real-time analysis and decision-making
  • Broad range of switch configurations:
    25.6Tbps: 256 x 100G, 128 x 200G, 64 x 400G, 32 x 800G
    12.8Tbps: 256 x 50G, 128 x 100G, 64 x 200G, 32 x 400G, 16 x 800G

Key Features

Key Features

At less than 300W for 25.6Tbps and under 200W for 12.8Tbps for typical data center use cases, X1 is an ideal low-power and low-latency solution for customer-specific data center deployment use cases.

2x improvement in port density in a 1RU form factor by championing the transition to the 100G SerDes ecosystem.

2x improvement in port density in a 1RU form factor by championing the transition to the 100G SerDes ecosystem.

Extensive suite of programmable Data Plane Telemetry for best-in-class network visibility for monitoring, troubleshooting, real-time analysis and decision-making

Extensive suite of programmable Data Plane Telemetry for best-in-class network visibility for monitoring, troubleshooting, real-time analysis and decision-making

FAQs

FAQs

What is the X2 chipset, and how does it differ from the first generation X- Switch family?2024-05-27T10:55:53+00:00

The X2 chipset is the second generation chipset in the X-Switch family of programmable Ethernet switches. It offers enhanced performance and capabilities compared to its predecessor, supporting up to 12.8 Tbps full-duplex throughput.

What are the primary applications of the X2 switches?2024-05-27T10:56:02+00:00

The X2 switches are optimized for interconnecting AI, storage, and compute clusters in data centers, providing efficient and high-speed connectivity for these critical workloads.

What features distinguish the X2 switches in terms of connectivity and compatibility?2024-05-27T10:56:32+00:00

The X2 switches incorporate industry-leading 100G PAM4 LR SerDes, enabling seamless connectivity within the 100G ecosystem without the need for re-timers. They are also future proofed for compatibility with emerging 800G optics ecosystems. Additionally, the devices support port speeds of 50, 100, 200, and 400 GbE.

How does the X2 chipset ensure flexibility and optimization for diverse applications?2024-05-27T10:55:38+00:00

The combination of a programmable packet processor and the X-PNDTM engine allows for flexible sharing of control tables, enabling customization and optimization for various application requirements and protocols.

What are the key features that enhance performance and reliability in X2 switches?2024-05-27T10:57:39+00:00

X2 switches feature comprehensive device features aimed at optimizing AI, storage, and compute workloads. They minimize packet discard in incast scenarios with a fully shared packet buffer and utilize X-FCTM, a queuing system combined with a fine-grained flow control protocol. Hardware-based link failure detection and recovery mechanisms ensure high availability, while PFC and cut-through mechanisms optimize RDMA workload performance.

What is the X2 Data Center Switch optimized for?2024-06-03T05:27:01+00:00

The X2 Data Center Switch is optimized for interconnecting AI/ML, storage, and compute clusters within a data center’s ecosystem.

What are the key features of the X2 Data Center Switch?2024-05-27T10:59:22+00:00

The X2 Data Center Switch enables the construction of compact 1RU switches with large port densities of up to sixteen ports of 800G OSFP.

What benefits does the X2 Data Center Switch offer?2024-05-27T11:00:10+00:00

The X2 Data Center Switch provides enhanced connectivity and scalability for AI/ML, storage, and compute clusters, thereby facilitating efficient data center operations and workflows.

How does the X2 Data Center Switch support AI and ML applications?2024-05-27T11:00:57+00:00

The X2 Data Center Switch is designed to efficiently handle the networking requirements of AI and ML workloads, ensuring seamless communication and data transfer between different components of the infrastructure.

Can the X2 Data Center Switch accommodate different types of connectivity protocols?2024-05-28T10:11:40+00:00

Yes, the X2 Data Center Switch is versatile and can support various connectivity protocols commonly used in data center environments, making it adaptable to diverse networking requirements.

Stay tuned for New Product Announcements!

Stay tuned for the New Product Announcements!