[BC Award 2023 Winners Interview] Golden Award – NVIDIA DGX H100

 Publish Date :2023/06/27

NVIDIA A100 Tensor core GPUs has delivered unprecedented performance in AI, data analytics, and high-performance computing (HPC) for various jobs. In 2023, NVIDIA released the NVIDIA H100 Tensor core GPU which features higher computing and energy efficiency. NVIDIA’s dedication to providing the best solutions for AI applications has awarded them the Best Choice Award Golden Award with the NVIDIA DGX H100.

This year’s NVIDIA GTC 2023 gathered not only global ICT giants such as Microsoft, Google, and AWS, it was also joined by several well-known companies such as Netflix, Ford Motor Company, and Nike. Why do people value AI so high? ChatGPT has shown how industries can leverage AI to improve efficiency and put business strategies into practice, primarily in the medical industry, cloud, and communications.

NVIDIA has repeatedly emphasized that GPU is the foundation of all developments. As the best choice for enterprise AI, the DGX H100 is equipped with a dedicated Transformer Engine and utilizes the NVIDIA® NVLink® Switch system. It is also optimized for AI R&D, generative AI training & application, large language model (LLM), and recommendation system. The system provides unprecedented performance for large-scale AI and HPC such as chatbots, recommendation engines, and visual recognition.

While LLM has existed for quite some time, it has recently grown in complexity and the rise of ChatGPT has shown the level of complexity and sophistication that can be expected in the future. To enhance these models in performance, speed, and accuracy; the H100 not only accelerates inference by up to 30x, it also delivers the lowest latency. Fourth-generation Tensor Cores speed up all precisions, including FP64, TF32, FP32, FP16, INT8, and now FP8, to reduce memory usage and increase performance while still maintaining accuracy for LLMs.

NVIDIA has committed to the development of GPU and parallel computing and expanded into various applications. From the technical perspective, the NVIDIA H100 Tensor Core GPU provides unprecedented performance. The NVIDIA H100 Tensor Core GPU has up to 6x more AI performance and up to 3x more HPC performance compared to the prior-generation A100 GPU. It’s designed to perform with incredible efficiency. Its second-generation Multi-Instance GPU technology can partition the GPU into smaller compute units, dramatically boosting the number of GPU clients available to data center users.

Various emerging technologies have improved industrial capabilities. For example, the combined application of sensors, AI, prediction models, etc. has significantly improved weather prediction accuracy. Today’s scientific computing workflow is very complex with large and static datasets with limited access and processing methods.

Based on the A100, the H100 is custom built according to the users’ needs. Its new architecture, high performance, and low power consumption can support multiple applications and solve various problems faced during the development phase.

As energy demand has increased and it is not an unlimited resource, all manufacturers are trying to find ways to do more while consuming less energy. For energy intensive applications such as data analysis, deep learning, computer vision, graphic processing, etc. it is important to find and utilize energy efficient solutions while pursuing greater performance. NVIDIA’s goal is to strike balance between performance and energy efficiency and currently, the NVIDIA DGX H100 might be the best answer to that demand.

For more winners of the Best Choice Award 2023, visit the Official Best Choice Award Winner page here.