NVIDIA H200 NVL Accelerator - GPU computing processor - H200 Tensor Core - 141 GB HBM3e - PCI Express 5.0
-
Power efficiency
Designed to function with a power consumption of 600 watts, ensuring that high performance does not come at the cost of excessive energy use.
-
Video memory capacity
The 141 GB installed size of HBM3e memory facilitates efficient handling of complex computations and large data sets.
-
High bandwidth
With a bandwidth of 4800 GBps, this accelerator supports fast data processing, crucial for demanding applications.
-
Computational power
Capable of achieving up to 1979 Tflops in peak half precision and 67 Tflops in peak single precision, it empowers users to run intensive workloads seamlessly.
-
Connectivity options
The PCI Express 5.0 interface and NVLink support allow for high-speed data transfers and better scalability in multi-GPU setups.
The NVIDIA H200 NVL Accelerator from Hewlett Packard Enterprise is designed to elevate the performance and efficiency of compute-intensive tasks. With an operational power consumption of 600 watts and advanced video memory technology utilizing HBM3e, this accelerator ensures performance in demanding applications. Its capabilities are enhanced by a 141 GB installed size and a bandwidth of 4800 GBps, catering to the needs of high-performance computing environments.
Equipped with features such as the NVIDIA Tensor Core graphics processor, this model delivers computational power with 34 Tflops for peak double precision (FP64), 67 Tflops for peak single precision (FP32), and 1979 Tflops for peak half precision (FP16). The PCI Express 5.0 interface allows for seamless integration into various systems, while support for NVLink enhances data throughput across connected devices. This accelerator is built to handle taxing tasks, making it a choice for professionals seeking to optimize their workloads and achieve results.
- 600 watts operational power consumption for robust performance
- 141 GB installed video memory with HBM3e technology
- 4800 GBps bandwidth for rapid data processing
- 34 Tflops peak double precision performance
- Supports NVLink for enhanced connectivity
