NVIDIA CEO personally brought the first DGX H200 AI accelerators to OpenAI's offices

By: Bohdan Kaminskyi | 26.04.2024, 17:52

gdb/X

NVIDIA CEO Jensen Huang personally brought the first DGX H200 systems to OpenAI's San Francisco office. This gesture was noted in a tweet by the startup's co-founder Greg Brockman.

Here's What We Know

The DGX H200 is NVIDIA's latest and most powerful processor for artificial intelligence. Its delivery to OpenAI, a leader in AI research, marks an important milestone as the company has received cutting-edge AI hardware.

Key improvements to the DGX H200 include a 40 per cent and 80 per cent increase in bandwidth and memory capacity, respectively, over the previous model. This is achieved by integrating HBM3e memory, which is critical for training large generative AI models.

According to NVIDIA's vice president, the expanded H200 memory is designed to significantly improve performance when working with complex AI models.

For OpenAI, the accelerator purchase is an important step to expand its research capabilities, especially in terms of the long-awaited GPT-5. The increased computational power will boost the performance of AI models.

The arrival of the H200 will catalyse progress across the AI industry, paving the way for ambitious new projects and breakthroughs in various fields.

At the same time, there is a high risk of shortages, as was the case with the previous H100. NVIDIA is committed to allocating resources fairly among system manufacturers and cloud providers.

Source: @gdb, VentureBeat