H100 Advanced Accelerated Computing Platform

The H100 represents the latest in GPU technology, purpose-built for Artificial Intelligence (AI) and High-Performance Computing (HPC) workloads. This platform combines advanced processing power, scalability, and energy efficiency, setting new benchmarks for performance and innovation.

Why Choose H100?

The H100 stands out for its unparalleled performance, memory bandwidth, and advanced AI acceleration capabilities. It’s a versatile choice for enterprises seeking to deploy large-scale AI models, conduct complex simulations, or optimize cloud-based HPC operations.

Cutting-Edge Architecture

  • Transistor Density: The H100 integrates 80 billion transistors, leveraging the industry’s most advanced 4nm fabrication process.
  • Tensor Core Technology: Features fourth-generation Tensor Cores, optimized for high-precision AI computations. These cores provide unmatched flexibility for handling diverse data types, including FP8, FP16, TF32, and FP64.

Exceptional Memory Bandwidth

  • HBM3 Memory: Equipped with High Bandwidth Memory (HBM3), the H100 delivers up to 3 TB/s of memory bandwidth, 50% higher than its predecessor.
  • High-Capacity Memory: Ensures seamless data throughput for large-scale AI models, complex simulations, and real-time processing.

Performance Highlights

  • FP64 Precision for HPC: Delivers a 3x improvement in FP64 operations, ideal for scientific computing and simulations.
  • Transformer Engine: Designed to accelerate large-scale natural language processing (NLP) and transformer-based AI models with FP8 precision.
  • Scalability: Supports multi-instance GPU (MIG) technology, allowing multiple independent workloads to run simultaneously on a single GPU.

Interconnectivity

  • Scalable Multi-GPU Systems: The H100 supports 256-GPU clusters, connected via advanced interconnect technologies to enable distributed AI and HPC workloads.
  • Data Center Optimization: Ensures high throughput and minimal latency, making it ideal for cloud environments and enterprise-grade infrastructure.

AI Training and Inference

  • Training Models: Enables faster training cycles for deep learning models, reducing time to insights.
  • Inference Efficiency: Provides ultra-fast inference capabilities for large datasets, enabling real-time decision-making.

Energy Efficiency

  • Eco-Friendly Design: Designed to deliver maximum performance with optimal power consumption, reducing operational costs in data centers.
  • Advanced Thermal Management: Maintains stable operation in high-density deployments, ideal for modern cloud and enterprise workloads.

Specifications

Processor / System Bus

2 x Socket

4th Intel® Xeon® Processor Scalable Family (Up to 350W)
5th Intel® Xeon® Processor Scalable Family (Up to 350W)

Memory

Total Slots:
32 (8 channel per CPU, 16 DIMM per CPU)
Capacity:
Maximum up to 4TB per CPU socket
Memory Type:
4th: DDR5 4400 RDIMM/ 3DS RDIMM (2DPC)
5th: DDR5 4400 RDIMM/ 3DS RDIMM (2DPC)
Memory Size:
64GB, 32GB, 16GB RDIMM
256GB, 128GB RDIMM 3DS

* Refer to support page for more information

Storage

Optional Kits:
-Broadcom MegaRAID 9560-16i
-Broadcom MegaRAID 9540-8i

Drive Bays

10 x 2.5” hot-swap drive bays (8 NVMe, 2 NVMe/SATA/SAS*)

[PCIe Switch directly]
Front: 8 NVMe
[CPU directly]
Rear: 2 NVMe(CPU2)/SATA(PCH)

H100 SKU
2 x M.2 Gen5 x4 link (CPU1) / 2 x M.2 Gen3 x2 link (PCH)

H200 SKU
2 x M.2 Gen3 x2 link (PCH)

Networking

2 x 10 Gigabit LAN ports (Intel X710-AT2 Controller)
1 x Management Por

Dimensions

885mm x 447mm x 306.65mm

Form Factor

7U

Weight

Net Weight: 113 Kg.
Gross Weight: 174 Kg.

Power Supply

4+2 Redundant 3000W 80 PLUS Titanium Power Supply