NVIDIA Mellanox MUA9652H-2SF UFM 4.0 Cyber-AI Appliance

• AI and GPU cluster monitoring – Provides an appliance-based platform for analyzing network behavior across demanding accelerated-computing environments.
• HPC InfiniBand operations – Helps administrators monitor and maintain high-performance InfiniBand fabrics where network degradation can directly affect compute utilization.
• Cyber-AI network analytics – UFM Cyber-AI uses AI-powered analytics to identify abnormal usage and network behavior, giving infrastructure teams earlier visibility into potential issues. 
• Preventive maintenance – Continuous system-data collection and predictive analysis can help teams identify conditions that warrant preventive maintenance before they become larger operational problems. 
• Performance degradation detection – Helps identify changes in network performance so administrators can investigate problems affecting cluster efficiency.
• Abnormal cluster behavior analysis – Correlates operational information to help expose unusual behavior across the infrastructure.
• Enterprise InfiniBand management – Supports deployments using UFM Enterprise and UFM Telemetry alongside the Cyber-AI platform.
• Large-scale research computing – Suitable for research environments where network availability and predictable performance are important to expensive compute resources.
• Data-center infrastructure security monitoring – AI-powered analysis can help identify unusual usage patterns that may warrant investigation. 

• NVIDIA UFM 4.0 Cyber-AI appliance
• 2U rackmount form factor
• NVIDIA A30 GPU with 24GB memory
• 2 × NVIDIA ConnectX-6 VPI dual-port adapters
• Up to 200Gb/s Ethernet support
• HDR InfiniBand support
• 128GB DDR4 ECC system memory
• 6 × 2TB enterprise SATA HDDs
• 2 × 3.84TB SATA SSDs
• RAID 1 and RAID 10 support
• Dual 1+1 redundant hot-swappable power supplies
• TPM 2.0 security
• Secure Boot with Intel Boot Guard
• Front-to-rear airflow
• BMC remote management
• Designed for AI, HPC and InfiniBand data centers

• Brand: NVIDIA Mellanox
• Product Family: NVIDIA UFM
• Model: MUA9652H-2SF
• NVIDIA SKU: 920-9B020-00FH-0D0
• Product Type: UFM 4.0 Cyber-AI Appliance
• Supported UFM Platforms: UFM Cyber-AI, UFM Enterprise, UFM Telemetry
• Form Factor: 2U rackmount, 19-inch
• System Depth: Up to 24 inches
• GPU: NVIDIA A30 24GB
• CPU: 2 × Intel Xeon Silver 4214R, 2.40GHz, 16.5MB cache
• Network Adapters: 2 × NVIDIA ConnectX-6 VPI dual-port adapters
• InfiniBand Speeds: SDR / QDR / HDR100 / HDR
• Ethernet Speeds: 25 / 50 / 100 / 200Gb/s
• Documented Bandwidth: Up to 100Gb/s bidirectional per port
• System Memory: 128GB total; 16 × 8GB DDR4 ECC 2666MHz modules
• HDD Storage: 6 × 2TB 2.5-inch SATA 7200RPM enterprise HDD
• SSD Storage: 2 × 3.84TB 2.5-inch SATA 6Gb/s 3D2 TLC SSD
• RAID Support: RAID 1 for 2 HDD; RAID 10 for 4 HDD; RAID 1 for 2 SSD
• Management Networking: 2 × 1GbE management ports with IPv4/IPv6; 2 × 10GbE
• Serial Interface: 1 × DB9 RS-232 male
• PCIe: 2 × PCI Express 3.0 x16
• TPM: TPM 2.0

Why Choose This Product?

• Purpose-built UFM appliance – Instead of assembling a separate server, GPU and network adapters, buyers receive a purpose-configured platform for UFM deployments.
• AI-accelerated analytics – The integrated A30 24GB GPU provides hardware acceleration for supported deep-learning frameworks used by UFM Cyber-AI. 
• Dual ConnectX-6 VPI connectivity – Two dual-port adapters provide multiple high-speed interfaces for connecting the appliance to supported InfiniBand infrastructure.
• Predictive operational visibility – UFM Cyber-AI continuously collects system data and uses AI-powered analytics to identify degradation, abnormal behavior and potential maintenance requirements. 
• Reduced infrastructure downtime risk – Earlier detection of performance and component issues can help operations teams investigate problems before they cause wider disruption.
• Enterprise storage architecture – Dedicated enterprise HDD and SSD storage with documented RAID configurations provides a structured platform for appliance data and software requirements.
• Resilient hardware design – Redundant hot-swappable PSUs and multiple cooling fans improve serviceability and operational resilience. 
• Security-conscious platform – TPM 2.0 and secure boot based on Intel Boot Guard provide hardware-level security mechanisms for the appliance.