NVIDIA H100 Tensor Core GPU for AI, HPC & Enterprise Workloads
- Product: NVIDIA H100 Tensor Core Graphics Card
- Architecture: Hopper GPU Architecture
- Interconnect: NVIDIA NVLink & PCIe Gen5
- Power Consumption: Up to 700W
- Memory Bandwidth: Up to 3.35TB/s
- Tensor Performance: Up to 3,958 TFLOPS
- Form Factor: SXM Module / PCIe Accelerator
- Warranty: 1 Year Assured Warranty
- Quality Assurance: Professionally tested by certified technicians
- Support: 24/7 Technical Support
- Use Cases: Deep Learning, Generative AI, AI Training, Large Language Models (LLMs)
Are you facing issues of slow performance of the GPU on your server system when training AI models in your data center for real-time inferencing or HPC workloads? The NVIDIA H100 Tensor Core GPU is designed to help with demanding tasks such as AI processing, DL, analytics, and scientific computing using NVIDIA’s Hopper architecture for faster performance and increased scalability. Some of the features of the NVIDIA H100 GPU include 80GB of HBM3 memory, 4th Generation Tensor Cores, PCIe Gen 5, NVLink support for multiple GPUs, and MIG, which provide enhanced performance and increased processing speed for enterprise AI servers and HPC systems. ServerBasket provides a complete range of reliable NVIDIA H100 graphics card solutions with 24/7 technical support and a one-year warranty. Order yours today!
Technical Specifications
| FP64 | 34 teraFLOPS |
|---|---|
| FP64 Tensor Core | 67 teraFLOPS |
| FP32 | 67 teraFLOPS |
| TF32 Tensor Core | 989 teraFLOPS |
| BFLOAT16 Tensor Core | 1,979 teraFLOPS |
| FP16 Tensor Core | 1,979 teraFLOPS |
| FP8 Tensor Core | 3,958 teraFLOPS |
| INT8 Tensor Core | 3,958 TOPS |
| GPU Memory | 80GB |
| GPU Memory Bandwidth | 3.35 TB/s |
| Decoders | 7 NVDEC, 7 JPEG |
| Max Thermal Design Power (TDP) | Up to 700W (configurable) |
| Multi-Instance GPUs | Up to 7 MIGs @ 10GB each |
| Form Factor | SXM |
| Interconnect | NVIDIA NVLink: 900GB/s |
| PCIe | PCIe Gen5: 128GB/s |
| Server Options | NVIDIA HGX H100 (4 or 8 GPUs), NVIDIA DGX H100 (8 GPUs), Partner and NVIDIA-Certified Systems |
Technical Specifications
| FP64 | 34 teraFLOPS |
|---|---|
| FP64 Tensor Core | 67 teraFLOPS |
| FP32 | 67 teraFLOPS |
| TF32 Tensor Core | 989 teraFLOPS |
| BFLOAT16 Tensor Core | 1,979 teraFLOPS |
| FP16 Tensor Core | 1,979 teraFLOPS |
| FP8 Tensor Core | 3,958 teraFLOPS |
| INT8 Tensor Core | 3,958 TOPS |
| GPU Memory | 80GB |
| GPU Memory Bandwidth | 3.35 TB/s |
| Decoders | 7 NVDEC, 7 JPEG |
| Max Thermal Design Power (TDP) | Up to 700W (configurable) |
| Multi-Instance GPUs | Up to 7 MIGs @ 10GB each |
| Form Factor | SXM |
| Interconnect | NVIDIA NVLink: 900GB/s |
| PCIe | PCIe Gen5: 128GB/s |
| Server Options | NVIDIA HGX H100 (4 or 8 GPUs), NVIDIA DGX H100 (8 GPUs), Partner and NVIDIA-Certified Systems |
No FAQ Found
Reviews
There are no reviews yet.