CUDA Cores
VRAM
GB/s
16,896
CUDA Cores
1450
Base MHz
1900
Boost MHz
141GB HBM3e
5120-bit bus
75
FP32 TFLOPS
2400
FP16 TFLOPS
700W
TDP
2
Available Instances
$1.50/hr
Starting Price
| Architecture | Hopper (Unknown) |
| Release Date | 2023-11-13 |
| Launch Price | $35,000.00 |
| Process | 4nm |
| Transistors | 90B |
Gen 4+
Tensor Cores
Enabled
Transformer Engine
Supported
Flash Attention
10.5in
Length
4.4in
Width
2-slot
Height
The NVIDIA H200 is a powerful GPU designed for AI/ML workloads, offering exceptional performance for both training and inference tasks. With 141GB of VRAM and 16,896 CUDA cores, it provides the memory capacity and computational power needed for modern deep learning models.
Released in 2023, the H200 features Hopper architecture with advanced AI accelerators including Tensor Cores and Transformer Engine support. This makes it ideal for large language models, computer vision tasks, and generative AI applications.
When considering cloud rental options for the H200, pricing starts at $1.50/hour from various providers. This GPU offers excellent price-to-performance for AI training workloads, with its high memory bandwidth of 4800 GB/s enabling fast data transfer for large datasets.
The H200 features CUDA compute capability 9.0 and is compatible with all major deep learning frameworks including PyTorch, TensorFlow, and JAX. Its 4nm manufacturing process ensures efficient power consumption relative to performance output.
Get started quickly with these trusted GPU cloud providers. We may earn a commission when you sign up.
Learn more about GPUs from these authoritative sources:
Official CUDA programming guide
NVIDIA GPU Specifications →Official NVIDIA GPU specs
TechPowerUp GPU Database →Comprehensive GPU specifications
CUDA Compute Capability Guide →GPU compute capability reference
| Category | Rank 1 | Rank 2 | Rank 3 |
|---|---|---|---|
| Best for Training | NVIDIA H200 | NVIDIA H100 | NVIDIA B200 |
| Best for Inference | NVIDIA A40 | NVIDIA A100 | NVIDIA A10 |
Compare GPU specifications and cloud instances to find the best GPU for your workload.