552 Search Results for "NVIDIA ConnectX-7"
-
Vendor:Cisco
Cisco UCSC-885A-M8-M356 | High-density GPU rack server | MI350X accelerators | 12x ConnectX-7 NICs (8+4) | 3TB DDR5 RAM
UCSC-885A-M8-M356
Stock Status: Enquire
Vendor:CiscoCisco UCSC-885A-M8-M356 | High-density GPU rack server | MI350X accelerators | 12x ConnectX-7 NICs (8+4) | 3TB DDR5 RAM
Compute-optimized rack server configured for large-scale AI and HPC. MI350X accelerators pair with high-memory capacity for model training, inference, and data analytics. Dual-tier ConnectX-7 networking (8 + 4 ports) enables flexible east–west and uplink fabrics, while redundant power and robust out-of-band management support enterprise operations. -
Vendor:NVIDIA
NVIDIA 900-9X7AH-0079-DTZ | 200GbE 1x QSFP56, PCIe 5.0 x16, HHHL
900-9X7AH-0079-DTZ
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-9X7AH-0079-DTZ | 200GbE 1x QSFP56, PCIe 5.0 x16, HHHL
Data‑center‑class 200GbE PCIe 5.0 adapter built for intensive east‑west traffic, storage acceleration, and GPU‑centric workloads. Delivers RDMA, inline crypto, and overlay offloads to maximize application performance and efficiency. -
Vendor:NVIDIA
NVIDIA MTM016919 | PCIe bracket | low-profile | 2x zQSFP cutouts | fits ConnectX-6/5/4 dual-port adapters
MTM016919
Stock Status: Enquire
Vendor:NVIDIANVIDIA MTM016919 | PCIe bracket | low-profile | 2x zQSFP cutouts | fits ConnectX-6/5/4 dual-port adapters
Low-profile (short) field-replaceable PCIe bracket for dual-port zQSFP adapters. Optimized for half-height chassis with precise dual-cage cutouts for supported ConnectX NICs. Use this when converting a dual-port card to low-profile installations. -
Vendor:NVIDIA
NVIDIA 900-53669-0000-000 | NVLink Bridge 3-Slot 2-Way, RTX A-Series Compatible
900-53669-0000-000
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-53669-0000-000 | NVLink Bridge 3-Slot 2-Way, RTX A-Series Compatible
The NVIDIA 900-53669-0000-000 is a high-speed NVLink bridge designed to connect two RTX A-series professional GPUs with 3-slot spacing. This interconnect enables GPU-to-GPU memory pooling and performance scaling, delivering up to 112 GB/s of total bandwidth for AI, deep learning, simulation, and professional visualization workloads. Compatible with RTX A6000, A5500, A5000, A4500, A40, and select datacenter GPUs. -
Vendor:NVIDIA
NVIDIA MCP7Y70-H001 | Speed: 400G→4x100G | Form: OSFP to 4xQSFP56 | Standard: InfiniBand HDR | Wavelength: n/a (passive copper) | Reach: 1m | Connector: OSFP<>QSFP56
MCP7Y70-H001
Stock Status: Enquire
Vendor:NVIDIANVIDIA MCP7Y70-H001 | Speed: 400G→4x100G | Form: OSFP to 4xQSFP56 | Standard: InfiniBand HDR | Wavelength: n/a (passive copper) | Reach: 1m | Connector: OSFP<>QSFP56
High-density passive copper breakout DAC that converts a 400G HDR OSFP host into four 100G QSFP56 endpoints. Engineered for ultra-low latency and minimal power within top-of-rack and short interconnects in HPC/AI environments. -
Vendor:Cisco
Cisco UCSC-885A-M8-M3X3 | Rack server | MI300X GPU | CX7: 8 + 2 ports | 2.3TB RAM
UCSC-885A-M8-M3X3
Stock Status: Enquire
Vendor:CiscoCisco UCSC-885A-M8-M3X3 | Rack server | MI300X GPU | CX7: 8 + 2 ports | 2.3TB RAM
High-performance rack-mount server optimized for accelerated AI/HPC workloads with MI300X GPU capability. This configuration provides ConnectX-7 networking with 8 + 2 ports to deliver high-throughput, low-latency fabric connectivity and 2.3TB of system memory for large models and datasets. Flexible I/O and storage options (as configured) and remote management enable reliable deployment in training, inference, and data analytics clusters. -
Vendor:NVIDIA
NVIDIA 900-9X7AH-0039-STZ | 400GbE 1x QSFP112, PCIe 5.0 x16, HHHL (Full-Height Bracket)
900-9X7AH-0039-STZ
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-9X7AH-0039-STZ | 400GbE 1x QSFP112, PCIe 5.0 x16, HHHL (Full-Height Bracket)
Next‑generation 400GbE server adapter engineered for maximum throughput and ultra‑low latency in AI clusters, storage fabrics, and cloud-scale data centers. Delivers advanced RDMA, inline crypto, and rich overlay offloads over a single QSFP112 port. -
Vendor:NVIDIA
NVIDIA 900-5G153-2550-000 | RTX PRO 5000 Graphics Card, 48GB GDDR7, 300W, 4x DP 2.1b
900-5G153-2550-000
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-5G153-2550-000 | RTX PRO 5000 Graphics Card, 48GB GDDR7, 300W, 4x DP 2.1b
The NVIDIA RTX PRO 5000 Blackwell is a professional graphics card built on the Blackwell architecture, delivering breakthrough AI and visual computing performance for demanding creative, engineering, and scientific workflows. With 48GB of GDDR7 memory with ECC support and 14,080 CUDA cores, it accelerates AI development, LLM inference, generative AI workflows, high-fidelity simulations, video production, and complex 3D modeling. Featuring 440 fifth-generation Tensor Cores and 110 fourth-generation RT Cores, the RTX PRO 5000 enables photorealistic rendering, neural rendering capabilities, and supports up to four 8K displays via DisplayPort 2.1b outputs. -
Vendor:NVIDIA
NVIDIA 920-9B210-00FN-0M0 | Quantum-2 QM9700 64-Port NDR InfiniBand Switch, 51.2Tbps
920-9B210-00FN-0M0
Stock Status: Enquire
Vendor:NVIDIANVIDIA 920-9B210-00FN-0M0 | Quantum-2 QM9700 64-Port NDR InfiniBand Switch, 51.2Tbps
The NVIDIA Quantum-2 QM9700 is a high-performance InfiniBand switch engineered for AI training clusters, HPC environments, and hyperscale data centers. Delivering 64 ports of NDR 400Gb/s connectivity via 32 OSFP connectors in a compact 1U chassis, it achieves 51.2 Tbps aggregate bidirectional throughput with sub-100ns port-to-port latency. Ideal for NVIDIA DGX SuperPOD deployments and large-scale GPU compute fabrics requiring ultra-low latency RDMA, adaptive routing, and in-network computing acceleration. -
Vendor:Cisco
Cisco UCSC-885A-M8-M3X0 | High-memory GPU rack server | MI300X accelerators | 2x ConnectX-7 NICs | 15TB DDR5 RAM
UCSC-885A-M8-M3X0
Stock Status: Enquire
Vendor:CiscoCisco UCSC-885A-M8-M3X0 | High-memory GPU rack server | MI300X accelerators | 2x ConnectX-7 NICs | 15TB DDR5 RAM
AI- and data-intensive rack server built around MI300X accelerators and ultra-high memory capacity for large models, vector databases, and analytics. Dual ConnectX-7 NICs provide low-latency fabric connectivity, while enterprise-grade power and management ensure reliable 24/7 operation. Ideal for scaling model training or multi-tenant inference clusters. -
Vendor:NVIDIA
NVIDIA 900-53625-2500-000 | 2-Way NVLink Bridge | 2-Slot Spacing | For GV100-class GPUs
900-53625-2500-000
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-53625-2500-000 | 2-Way NVLink Bridge | 2-Slot Spacing | For GV100-class GPUs
High-bandwidth 2-way NVLink bridge connecting two compatible GV100-class GPUs to enable fast peer-to-peer communication and enlarged unified memory spaces for demanding compute and visualization workloads. -
Vendor:NVIDIA
NVIDIA 900-9X766-003N-ST0 | 200GbE 1x QSFP56, PCIe 5.0 x16, HHHL
900-9X766-003N-ST0
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-9X766-003N-ST0 | 200GbE 1x QSFP56, PCIe 5.0 x16, HHHL
High-performance 200GbE server adapter delivering ultra‑low latency, advanced RDMA (RoCE v2), and rich offloads for AI/HPC, storage, and cloud networking. Ideal for leaf/spine and GPU-accelerated platforms with PCIe 5.0 bandwidth headroom and backward compatibility. -
Vendor:NVIDIA
NVIDIA 900-54934-2500-000 | NVLink 2-Slot Bridge for Quadro RTX 5000, 50GB/s
900-54934-2500-000
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-54934-2500-000 | NVLink 2-Slot Bridge for Quadro RTX 5000, 50GB/s
The NVIDIA 900-54934-2500-000 is a 2-slot NVLink bridge designed exclusively for connecting two identical Quadro RTX 5000 GPUs. This high-speed interconnect delivers 50 GB/s bidirectional bandwidth, enabling peer-to-peer GPU communication that significantly outperforms traditional PCIe for multi-GPU professional workloads. Ideal for rendering, simulation, and AI workflows that support NVLink technology. -
Vendor:NVIDIA
NVIDIA 920-9B36F-00RX-8S0 | Quantum‑3 XDR InfiniBand switch | 144x 800G XDR over 72x OSFP | 4U chassis | Non‑PoE
920-9B36F-00RX-8S0
Stock Status: Enquire
Vendor:NVIDIANVIDIA 920-9B36F-00RX-8S0 | Quantum‑3 XDR InfiniBand switch | 144x 800G XDR over 72x OSFP | 4U chassis | Non‑PoE
Quantum‑3 based XDR InfiniBand data center switch delivering extreme radix and bandwidth for next‑gen AI/HPC clusters. Provides 144 XDR ports via 72 OSFP cages in a compact 4U chassis, enabling dense spine/leaf deployments with advanced congestion management and telemetry at line rate. -
Vendor:Cisco
Cisco CAI-GPU-RTXP6000 | RTX Pro 6000 Graphics Card, 96GB GDDR7, 600W, PCIe 5.0
CAI-GPU-RTXP6000
Stock Status: Enquire
Vendor:CiscoCisco CAI-GPU-RTXP6000 | RTX Pro 6000 Graphics Card, 96GB GDDR7, 600W, PCIe 5.0
The NVIDIA RTX Pro 6000 Blackwell Workstation Edition delivers professional-grade AI and visual computing for workstations and servers. Built on the Blackwell architecture with 96GB GDDR7 ECC memory, 24,064 CUDA cores, and 600W TDP, this dual-slot GPU accelerates AI inference, large language models, 3D rendering, simulation, and data science workloads. Ideal for engineering, content creation, scientific computing, and enterprise AI applications requiring massive memory capacity and proven professional reliability. -
Vendor:Cisco
Cisco UCSC-885A-M8-M352 | Rack server | 8x MI350X GPUs | 2x ConnectX-7 | 23TB RAM
UCSC-885A-M8-M352
Stock Status: Enquire
Vendor:CiscoCisco UCSC-885A-M8-M352 | Rack server | 8x MI350X GPUs | 2x ConnectX-7 | 23TB RAM
High‑density 8‑GPU rack platform built for large‑scale AI/ML training, LLM inference, and HPC workloads. Factory‑configured with 8x MI350X accelerators, ConnectX‑7 networking, and a massive 23TB memory footprint for memory‑bound models and in‑memory analytics. -
Vendor:NVIDIA
NVIDIA 900-53669-2500-000 | NVLink Bridge 3-Slot, 112GB/s, Ampere Generation
900-53669-2500-000
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-53669-2500-000 | NVLink Bridge 3-Slot, 112GB/s, Ampere Generation
The NVIDIA NVLink Bridge 3-slot enables high-speed GPU-to-GPU interconnect for dual Ampere-based RTX A-series professional graphics cards. Designed for systems with three PCI-Express slot spacing, this bridge provides up to 112 GB/s bidirectional bandwidth, enabling memory pooling and performance scaling for demanding visual computing, AI, data science, rendering, and simulation workloads. -
Vendor:NVIDIA
NVIDIA 980-9I17Q-00CM00 | 100GbE | QSFP28 | 100GBASE-CWDM4 | 1310 nm | 2 km SMF | LC duplex
980-9I17Q-00CM00
Stock Status: Enquire
Vendor:NVIDIANVIDIA 980-9I17Q-00CM00 | 100GbE | QSFP28 | 100GBASE-CWDM4 | 1310 nm | 2 km SMF | LC duplex
Cost‑effective 100G CWDM4 optics for campus and data center interconnect up to 2 km. QSFP28 with LC duplex uses four CWDM lanes near 1310 nm over OS2 single‑mode fiber for simple, standards‑based deployment. -
Vendor:NVIDIA
NVIDIA 900-9X721-003N-DT0 | 1x 400G Eth/IB VPI, PCIe Gen5 x16, RDMA RoCEv2, SR-IOV, GPUDirect, PTP
900-9X721-003N-DT0
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-9X721-003N-DT0 | 1x 400G Eth/IB VPI, PCIe Gen5 x16, RDMA RoCEv2, SR-IOV, GPUDirect, PTP
Single‑port 400G VPI adapter for high‑throughput Ethernet or InfiniBand connectivity in AI/HPC clusters and NVMe‑oF storage. PCIe Gen5 x16 host interface with advanced offloads: RDMA (RoCE v2), GPUDirect RDMA, NVMe‑oF, inline IPsec/TLS, and PTP hardware timestamping for microsecond‑level determinism. -
Vendor:NVIDIA
NVIDIA 900-9X721-003N-DT1 | 1x 200G Eth/IB VPI, PCIe Gen4/5 x16, RDMA RoCEv2, SR-IOV, GPUDirect, PTP
900-9X721-003N-DT1
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-9X721-003N-DT1 | 1x 200G Eth/IB VPI, PCIe Gen4/5 x16, RDMA RoCEv2, SR-IOV, GPUDirect, PTP
Single‑port 200G VPI adapter for high‑performance Ethernet or InfiniBand connectivity in dense compute and storage nodes. PCIe Gen4/5 x16 interface with advanced offloads: RDMA (RoCE v2), GPUDirect RDMA, NVMe‑oF, inline IPsec/TLS, and IEEE 1588 PTP hardware timestamping. -
Vendor:NVIDIA
NVIDIA 900-5G172-2580-000 | RTX A1000 8GB GDDR6 Graphics Card, 50W, 4x Mini-DP
900-5G172-2580-000
In stock
Vendor:NVIDIANVIDIA 900-5G172-2580-000 | RTX A1000 8GB GDDR6 Graphics Card, 50W, 4x Mini-DP
The NVIDIA RTX A1000 8GB is a professional low-profile graphics card built on Ampere architecture, designed for compact workstations requiring powerful visualization and AI acceleration. Featuring 2,304 CUDA Cores, 72 Tensor Cores, and 18 RT Cores with 8GB GDDR6 memory, it delivers real-time ray tracing and AI-enhanced workflows for CAD, 3D rendering, and data visualization. The single-slot, 50W design with four Mini DisplayPort 1.4a outputs supports up to four 5K displays, making it ideal for professional applications in space-constrained environments. -
Vendor:NVIDIA
NVIDIA 900-9X6AF-0018-MT0 | 200Gb/s InfiniBand HDR, QSFP56, PCIe Gen4 x16, FHHL
900-9X6AF-0018-MT0
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-9X6AF-0018-MT0 | 200Gb/s InfiniBand HDR, QSFP56, PCIe Gen4 x16, FHHL
High‑throughput InfiniBand adapter delivering HDR 200 Gb/s for HPC and AI clusters. Provides ultra‑low‑latency RDMA, GPUDirect RDMA, and in‑hardware offloads to accelerate MPI messaging, storage fabrics, and distributed training workloads. Ideal for scaling InfiniBand fabrics and enabling NVMe‑oF/IB in performance‑critical environments. -
Vendor:NVIDIA
NVIDIA 920-9B210-00FN-0D0 | Quantum-2 QM9790 64-Port 400G NDR InfiniBand Switch, 1U
920-9B210-00FN-0D0
Stock Status: Enquire
Vendor:NVIDIANVIDIA 920-9B210-00FN-0D0 | Quantum-2 QM9790 64-Port 400G NDR InfiniBand Switch, 1U
The NVIDIA QM9790 (920-9B210-00FN-0D0) is an unmanaged 1U NDR InfiniBand switch delivering 64 ports of 400Gb/s connectivity via 32 OSFP cages with 51.2Tb/s aggregate bidirectional throughput and 66.5 billion packets per second capacity. Built on the Quantum-2 ASIC, this switch is engineered for extreme-scale AI training clusters, HPC environments, and data-intensive workloads requiring ultra-low latency interconnects. The P2C airflow design ensures optimal thermal performance in high-density rack deployments. -
Vendor:NVIDIA
NVIDIA 900-2G179-0020-100 | A2 Tensor Core GPU, 16GB GDDR6, 40-60W, PCIe 4.0 x8
900-2G179-0020-100
Stock Status: Enquire
Vendor:NVIDIANVIDIA 900-2G179-0020-100 | A2 Tensor Core GPU, 16GB GDDR6, 40-60W, PCIe 4.0 x8
The NVIDIA A2 Tensor Core GPU delivers entry-level AI inference acceleration in a compact, low-power form factor ideal for edge computing deployments. Built on the NVIDIA Ampere architecture with 1280 CUDA cores and 40 third-generation Tensor Cores, it provides 4.5 TFLOPS FP32 performance and up to 72 TOPS INT4 precision. The single-slot, low-profile design with passive cooling and configurable 40-60W TDP makes it perfect for space-constrained servers handling intelligent video analytics, machine learning inference, and virtual desktop infrastructure workloads.
