Artificial Intelligence Solutions High-Performance Servers & GPU Platforms for AI Workloads

• Form Factor: 5U rack server

• Processor: 2x AMD EPYC™ 9005 (up to 500W TDP)

• Memory: 24x DDR5 6400MT/s DIMMs

• GPU: 8x NVIDIA Blackwell HGX B200, up to 1000W TDP each

• Storage: Up to 12x SATA/SAS/NVMe + 2x NVMe/SATA M.2

• PCIe: Up to 12x PCIe 5.0 x16 slots, supports NVIDIA® BlueField®-3 NIC

• Networking: Optional 1x OCP 3.0 with NC-SI support

• Cooling: Direct-to-chip liquid cooling for CPU, GPU & NVSwitch

• Power Supply: 5x 54V 3200W (GPU module, 4+1 redundancy), 2x 12V 3200W (CPU module, 1+1 redundancy)

Technical Specifications

• Form Factor: 10U rack server

• Processors: 2x Intel® Xeon® 6 or 2x AMD EPYC™ 9005

• Memory: Up to 32x DDR5 6400MT/s DIMMs (max 8TB DDR5 ECC)

• GPU: NVIDIA Blackwell Ultra 8-GPU (HGX™ B300)

• Storage: 8x NVMe U.2, 2x NVMe/SATA M.2, optional 4x SAS/SATA 2.5"

• PCIe: 8x CX8 ports + 2x PCIe 5.0 x16 FHFL

• Networking: 1x OCP 3.0 (optional)

• Ports: 3x USB 3.0, 1x VGA, 1x RJ45, 1x MicroUSB, 1x Type-C

• Cooling: Air cooling

• Management & Security: ASPEED AST2600, TPM 2.0

• Power Supply: 6x 3300W 54V Dual AC, 2x 3200W 12V (N+1/N+N Redundant)

• Dimensions: 447 x 440 x 860 mm

• Form Factor: 6U rackmount

• Processor: Dual Intel® Xeon® 6 (6500P/6700P series), up to 350W TDP

• Memory: 32x DDR5 6400MT/s DIMMs

• GPU Support: 8x PCIe GPUs (H200 NVL / RTX PRO™ 6000 Blackwell)

• Storage: Up to 16x E3.S/NVMe + 12x 3.5’’/2.5’’ SAS/SATA HDD/NVMe

• PCIe Slots: Up to 8x FHHL PCIe 5.0 x16 Dual Slot

• Power Supply: Redundant 3200W PSUs with N+N, N+1 or Non-Redundant configurations

Technical Highlights:

• Form Factor: 6U rackmount

• Processor: Dual Intel® Xeon® 6 (6500P/6700P series), up to 350W TDP

• Memory: 32x DDR5 6400MT/s DIMMs

• GPU Support: 8x PCIe GPUs (NVIDIA HGX H200)

• Storage: Up to 16x E3.S/NVMe + 12x 3.5’’/2.5’’ SAS/SATA HDD/NVMe

• PCIe Slots: Up to 8x FHHL PCIe 5.0 x16 Dual Slot

• Power Supply: Redundant 3200W PSUs with N+N, N+1, or non-redundant configurations

• Form Factor: 4U rackmount

• PCIe Architecture: PCIe Gen 5.0 with 11–13 rear expansion slots

• GPU Support: Up to 8 FHFL PCIe GPUs

• Memory & Storage: Flexible system memory and storage options

• Scalability: Modular design for diverse AI workloads

• Processors: Dual AMD EPYC™ 9005 Series CPUs (SP5 socket, up to 500W TDP, air-cooled)

• GPU Platform: 8x AMD Instinct™ MI350X GPUs

• GPU Memory: 2TB HBM3e total with 48 TB/s memory bandwidth

• Workload Focus: LLM inference, cloud AI, heterogeneous HPC computing

• Form Factor: 6U rackmount

• GPU Platform: 8x AMD Instinct™ MI300X accelerators

• GPU Memory: 192GB HBM3 per GPU, 5.3 TB/s peak memory bandwidth

• Compute TDP: 750W per GPU

• Workload Focus: LLM, generative AI, HPC, cloud AI

• Open Ecosystem: Compatible with AI frameworks, ROCm, and AI development tools

• Energy Efficiency: Optimized GPU-accelerated workloads with mature open ecosystem

• Form Factor: 6U rackmount

• Processor: Dual 4th/5th Gen Intel® Xeon® Scalable

• Memory: 32x DDR5 DIMMs, up to 8TB system memory

• GPU Support: 8x Intel® Habana® Gaudi®2 OAM accelerators

• GPU Memory & Bandwidth: 96GB HBM3 per GPU, 4.2TB/s P2P bandwidth

• Cooling: Advanced thermal design with intelligent thermal management

• Power Supply: Dual 12V/54V input, optimized for energy efficiency

• Form Factor: 2U rackmount

• Processor: Dual latest-generation CPUs

• Memory: Flexible DIMM configurations, up to max system memory

• Storage: Multiple front/rear access options for SAS, SATA, NVMe drives

• Cooling: Supports cold-plated liquid cooling with leak detection alerts

• Expansion: Scalable PCIe and I/O options for diverse workloads

• Power Supply: Configurable for redundancy and efficiency

• Processors: Dual SP5 AMD EPYC™ 9005 Series CPUs (up to 500W TDP, air-cooled)

• GPU Platform: 8x AMD Instinct™ MI350X GPUs

• GPU Memory: 2TB HBM3e total with 48 TB/s memory bandwidth

• Workload Focus: LLM inference, cloud AI, heterogeneous HPC computing

Scalable High Performance:

• 96GB HBM3 memory with 4.2TB/s converged P2P bandwidth

• Supports AMX/DSA AI engines

• 4000Gbps cross-node communication bandwidth enables pipeline and data parallelism

Optimized Energy Efficiency:

• Refined thermal design with intelligent thermal management

• 12V/54V dual power input

• Dynamic power matching for high-power workloads, zero conversion loss, resulting in 8% system power reduction

Open Ecosystem:

• Complies with OAM v1.1 and UBB v1.5 standards

• Supports mainstream AI frameworks and development tools

• Mature open ecosystem enables cost-efficient AI compute solution development

Technical Highlights:

• GPU Platform: 8x AMD Instinct™ MI300X

• Processor: Dual AMD EPYC™ 9004 CPUs

• GPU Memory: 192GB HBM3 with 5.3 TB/s peak bandwidth

• TDP: 750W

• Workload Focus: LLM training, generative AI, HPC computing

Technical Highlights:

• GPU Platform: 8x NVIDIA Blackwell GPUs with NVLink

• CPU Support: 2x CPUs up to 500W TDP each

• GPU TDP: 1000W total

• Cooling: Direct-to-chip liquid cooling for CPU, GPU, NVSwitch

• Form Factor: 5U

• Workload Focus: LLM training, HPC, generative AI, AI research

Technical Highlights:

• GPU Platform: 8x NVIDIA Blackwell GPUs, HGX™ B300 compatible

• OAM Support: Intel® Gaudi® accelerators

• Form Factor: 10U

• Workload Focus: LLM training, multimodal AI inference, HPC, generative AI

• GPU Platform: 8x NVIDIA H200 with Transformer Engine

• AI Performance: 32 PFlops

• Form Factor: 6U

• Workload Focus: GPT LLM training, hyperscale AI, HPC, metaverse workloads

• Power & Efficiency: Optimized modular design for high efficiency and flexible configuration

• GPU / OAM Support: 8 NVIDIA Blackwell Ultra GPUs (HGX™ B300) or Intel® Gaudi® 3 OAMs

• Processors: 2x Intel® Xeon® 6 (6700/6900P Series)

• Memory: 32/24x DIMM slots up to 8TB

• Interconnect: CX8 IB, 800Gb per SXM

• Storage: 8x Hot-swap U.2 NVMe drives, 2x M.2 NVMe

• Cooling: Efficient air-cooling design