Unleash GPU-to-GPU Clustering Across Racks
- Extends PCIe 5.0, PCIe 4.0, and CXL signal reach across dense AI racks using Active Electrical Cables
- Built on the proven track record of our widely deployed and field-tested Aries PCIe/CXL Smart DSP Retimers
- Supports multiple form factors and configurations to accommodate diverse AI system topologies
- suite for performance optimization and seamless fleet management across dense AI platforms
Aries SCM Highlights
Enabling Reliable PCIe/CXL for Rack-to-Rack Connectivity
Why Use Aries Smart Cable Modules?
Aries SCMs enable robust, easy-to-design PCIe and CXL cabling while also offering COSMOS' suite of built-in advanced fleet management, security, and deep diagnostic capabilities critical to ensuring high reliability and up-time.
Robustness
Supported Data Rates: 32 GT/s, 16 GT/s, 8 GT/s, 5 GT/s, and 2.5 GT/s Data Rates with Automatic Link Equalization
Power Efficient: Low power CMOS process and L1.0 Low-Power Mode minimizes rack power/thermal density
Low Latency: Less than 10ns
Security: Advanced security features help prevent malicious attacks of unauthorized firmware loading, module diagnostics access, and configurations
Easy System Design
Thin Cable Gauge: Offering thin cables in various lengths for flexible bend radius
Extended Reach: Up to 7m for long reach inter-rack connectivity
Flexible Supply Chain: Aries SCMs are compatible with multiple cable vendors for ease of second sourcing of active cable assemblies
Interop Testing: Rigorous system testing with 50+ Endpoints and all major Root Complex
Fleet Management
Quick Debug: Built-in protocol analyzer with Link state history and timestamps, full non-destructive eye scan for RX Lane margining, self-test features to minimize link downtime and accelerate fault isolation
Deep Diagnostics: Firmware-driven link health monitoring to alert BMC of any possible link performance issues
Application Notes
Videos
First Demo of End-to-End PCIe over Optics for GPU Clusters Across the Data Center
Unprecedented low-latency connectivity scale for Generative AI and cloud infrastructure.
FAQ
Have more questions about general topics, such as PCIe/CXL, ordering, quality, etc.? See the full list of FAQs.