Rent GPU
NVIDIA H200
The NVIDIA H200 GPU supercharges generative AI and high-performance computing (HPC) workloads with game-changing performance and memory capabilities. delivering massive throughput improvements for generative AI and large-scale inference workloads.
Memory
141GB
GPU RAM
Bandwidth
4.8TB/s
Memory Bandwidth
Form factor
SXM & NVL
Architecture
Interconnect
900GB/s
NVLink Switch
Accelerated
Workload & Computing
H200 is purpose-built for next-generation AI models, enabling faster inference, reduced latency, and lower total cost for large-scale deployments across data centers and cloud platforms. 1.6x faster for GPT-3 175B Inference, 1.9X Faster for Llama2 70B Inference, 110X Faster for High-Performance Computing
- Ultra Large Model Inference
- High Throughput AI Serving
- Generative AI at Scale
Architecture Comparison
Select the form factor optimized for your specific workload
| Feature | H200 SXM | H200 NVL |
|---|---|---|
| architecture | Hopper | Hopper |
| memory | 141GB HBM3e | 141GB HBM3e |
| memory Bandwidth | 4.8 TB/s | 4.8 TB/s |
| recommended For | Better Tensor Cores for Advanced Generative AI & LLM Inference | High Performance Computing |