Cisco UCS C880A M8: HGX B300 + Intel Xeon 6 — The New Vanguard of AI Infrastructure

Cisco continues to redefine AI infrastructure with the launch of the UCS C880A M8 Rack Server, now equipped with the ultra-powerful NVIDIA HGX B300 SXM GPUs and the latest Intel Xeon 6th-Generation (Granite Rapids) CPUs. This marks a convergence of state-of-the-art GPU acceleration, high-throughput CPU compute, and enterprise-grade infrastructure management, optimized for modern AI workloads.

NVIDIA: HGX B300 — Unprecedented AI Performance

Drawing on NVIDIA’s messaging around the HGX B300 platform:

• Inference performance: Up to 11× higher throughput for models like Llama 3.1 405B versus the previous Hopper generation, thanks to Blackwell Tensor Cores and Transformer Engine optimizations.
• Training speed: Up to 4× faster for large-scale LLMs like Llama 3.1 405B via FP8 precision enhancements, NVLink 5 interconnects (with 1.8 TB/s bandwidth), InfiniBand, and Magnum IO software.
• SXM form factor advantage: High-bandwidth socket architecture enables superior GPU-to-GPU NVLink connectivity and power delivery without cabling, minimizing bottlenecks and simplifying scaling.

Intel: Xeon 6th-Gen CPUs — CPU Power Meets AI Acceleration

Powered by the newly released Intel Xeon 6 (6700P/6500P) P-core processors, the UCS C880A M8 delivers:

• Up to 86 cores per socket, doubled memory bandwidth, and integrated AI accelerators—including Trust Domain Extensions (TDX), Advanced Matrix Extensions (AMX), Data Streaming Accelerator (DSA), QuickAssist (QAT), and In-Memory Analytics Accelerator (IAA).
• Support for DDR5-6400 and MRDIMMs, boosting memory throughput further.
• Optimized for compute-intensive and hybrid AI workloads—from inference pipelines to large-scale training nodes.

Cisco: Intersight Management + AI POD Integration

Cisco’s AI infrastructure goes beyond raw compute:

• The UCS C880A M8 integrates seamlessly with Cisco Intersight, Cisco’s SaaS-based management platform offering centralized control, visibility, and policy management across the distributed AI stack.
• It fits within Cisco AI PODs, modular AI infrastructure solutions designed for rapid deployment at scale. These validated data center units simplify AI factory rollouts while ensuring interoperability with compute, networking, and security.

Key Use Cases Enabled by HGX B300 (SXM)

The integration of HGX B300 SXM GPUs and Xeon 6th-Gen CPUs enables a suite of AI workloads:

1. Real-Time LLM Inference
   • Run massive models like Llama 3.1 405B with ultra-low latency and high throughput, ideal for chatbots, agents, and real-time reasoning.
2. Large-Scale Model Training & Fine-Tuning
   • Take advantage of 4× faster training and massive inter-GPU bandwidth to train or fine-tune models with hundreds of billions of parameters.
3. High-Performance AI Pipelines
   • Leverage CPU offload for data preparation and orchestration, alongside GPU acceleration, for ETL, multimodal processing, and inference workflows.
4. AI-Native Data Centers / AI Factories
   • Build composable, secure, and scalable AI infrastructure blocks with Cisco AI PODs, ready for integration in data centers or at the edge.
5. HPC & Scientific Simulation
   • Run contiguous memory models and multi-GPU workloads with enhanced NVLink connectivity for high fidelity simulations and analytics.

Summary Table

Final Thoughts

The Cisco UCS C880A M8: HGX B300 with Intel Xeon 6th-Gen sets a new benchmark in AI infrastructure. It offers hyperscale-level AI performance, rock-solid CPU support, enterprise-grade manageability via Intersight, and secure deployments through Cisco Secure AI factory with Nvidia and Cisco AI PODs scalable architectures. Whether you’re building an AI training cluster, LLM inference engine, or composable AI infrastructure, this platform is purpose-built for the next frontier of AI.

Discover the power of next-gen AI infrastructure—read the Cisco UCS C880A M8 Data Sheet