End-to-End Networking Solutions
The NVIDIA Quantum-X800 platform represents the latest advancement in NVIDIA Quantum InfiniBand technology. Designed specifically for AI models at the trillion-parameter scale, this platform includes the NVIDIA Quantum-X800 InfiniBand switch, NVIDIA ConnectX®-8 SuperNIC, and LinkX cables and transceivers.
AMAX Solutions Powered by NVIDIA InfiniBand
NVIDIA DGX GB200 NVL72
| Component | Specification |
|---|---|
| CPU + GPU | 72x NVIDIA Blackwell GPUs in Grace Blackwell Superchips |
| Total Fast Memory | 30.2TB |
| Interconnect | 72x OSFP single-port NVIDIA ConnectX®-7 VPI with 400Gb/s InfiniBand |
| NVIDIA NVLink Switch System | 9x L1 NVIDIA NVLink Switches |
AMAX Industrial HPC Cabinet
| Component | Specification |
|---|---|
| CPU | Intel 4th/5th Gen Xeon Scalable Processor |
| CPU Cores | 8192 cores (P-Core) |
| Memory | 128 TB |
| Networking | 32x OSFP single-port NVIDIA ConnectX-7 VPI with 400Gb/s InfiniBand |
NVIDIA Quantum-X800 InfiniBand
The new platform supports advanced hardware-based, In-Network Computing with Scalable Hierarchical Aggregate Reduction Protocol (SHARP)™ v4, adaptive routing, and telemetry-based congestion control, enabling a new frontier of AI innovation.
| Data Throughput | Higher Bandwidth (800Gb/s) | In-Network Computing |
|---|---|---|
| 5X | 2X | 9X |
Accelerated In-Network Computing
Advanced technologies including NVIDIA SHARP v4, Message Passing Interface (MPI) tag matching, MPI_Alltoall, and programmable cores enhance NVIDIA In-Network Computing. Adaptive routing utilizes the switch and ConnectX-8 SuperNIC collaboratively to optimize bandwidth and maintain network resilience for AI fabrics. Furthermore, telemetry-based congestion control techniques offer noise isolation for multi-tenant AI workloads.
Network resiliency
The platform actively addresses hardware issues to maintain continuous application performance. Acceleration engines reduce latency and increase data throughput, improving network efficiency. Self-healing interconnect boosts the resilience and reliability of the NVIDIA Quantum-X800 InfiniBand network, ensuring consistent network integrity.
Full Offload Capabilities
Remote direct-memory access (RDMA), NVIDIA GPUDirect® RDMA, and GPUDirect Storage optimize investment returns. Advanced power efficiency: Techniques such as power capping and transitions to low-power states reduce power consumption during idle periods.
AMAX as your Total Solution Provider
AMAX offers expert NVIDIA InfiniBand solutions, ideal for high-performance computing and AI environments. Our advanced IT infrastructure and liquid cooling technologies ensure efficient, scalable, and reliable InfiniBand deployments.