NVIDIA H200
The NVIDIA H200 is a Data Center graphics card based on the Hopper architecture. Released on 2024-11-01 with a launch price of $35,000.00, it features 16,896 CUDA cores and 141 GB of HBM3e memory.
General info about H200
| Architecture | Hopper (GH200) |
| Market Segment | Data Center |
| Release Date | 2024-11-01 |
| Launch Price | $35,000.00 |
| Manufacturing Process | 4nm |
Technical specs of H200
| CUDA Cores | 16,896 |
| Base Clock Speed | 1450 MHz |
| Boost Clock Speed | 1900 MHz |
| Transistor Count | 90B |
| VRAM Capacity | 141 GB HBM3e |
| Memory Bus Width | 5120 bits |
| Memory Bandwidth | 4800.0 GB/s |
| TDP | 700 W |
Key Technical Parameters
Key technical parameters of the NVIDIA H200 include its 16,896 CUDA cores, 1450 MHz base clock, and 1900 MHz boost clock, delivering high performance for data center applications.
Benchmark performance of H200
FP32 Performance
75
TFLOPS
FP16 Performance
2400
TFLOPS
The NVIDIA H200 achieves 75 TFLOPS in FP32 performance, making it ideal for data center workloads.
Gaming performance of H200
Gaming Performance
The NVIDIA H200 is not designed for gaming. It is optimized for AI training, machine learning, and high-performance computing workloads.
Features and connectivity of H200
Tensor Cores
Gen 4+
Transformer Engine
Enabled
Flash Attention
Supported
Connectivity
NVLink 4, PCIe Gen5
Physical dimensions of H200
| Length | 10.5in |
| Width | 4.4in |
| Height | 2-slot |
Rankings and efficiency of H200
| Ranking Position | #1 |
| Popularity Ranking | #4 |
| Cost Effectiveness | 0.9/5 |
| Power Efficiency | 1.3/5 |
Cloud Availability
Available Instances
0
Starting Price
$0.00/hr
Top GPUs for Training and Inference
| 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.