Nvidia unveils DGX GH200 AI computing platform for ChatGPT successors and ACE for Games service for ultrarealistic NPCs

With AI being the talk of the town and stock riding highest of all chip makers thanks to the insatiable demand for its specialized silicon, Nvidia is not resting on its nearly US$1 trillion valuation laurels. From the stage of the Computex expo in Taiwan, Nvidia CEO Huang announced a number of new products and services that will capitalize on the trend further.

First in line is a new platform for artificial intelligence calculations called DGX GH200 that Nvidia says will empower its customers to create the next ChatGPT success story. Google, Microsoft, and Meta are to be the first big customers of the DGX supercomputer that connects 256 Nvidia GH200 Grace Hopper Superchips with a NVLink Switch System for the first time. Individually, the Grace Hoppers offers up to 900 GB/s bandwidth, 7x more than the PCIe Gen5 lanes in current accelerated systems, for a total of 1-Exaflop, 144TB GPU DGX GH200 supercomputing power for demanding generative AI and HPC applications.

According to Mark Lohmeyer, vice president of Compute at Google Cloud, the "new NVLink scale and shared memory of Grace Hopper Superchips address key bottlenecks in large-scale AI and we look forward to exploring its capabilities for Google Cloud and our generative AI initiatives."

Besides AI hardware, Nvidia also unveiled a number of services for developers that want to dabble in the possibility to spice up their existing code via artificial intelligence on the cheap. The new Avatar Cloud Engine (ACE) for Games, for instance, is "a custom AI model foundry service that transforms games by bringing intelligence to non-playable characters (NPCs) through AI-powered natural language interactions."

Instead of painstakingly coding background characters, game developers will have the option to empower them with tailored speech, body language animation, and conversational behavior on the fast and easy via a bunch of new AI tools.

Those include NVIDIA NeMo for creating language models that can make NPCs more interesting by fleshing them out more with backstories, while keeping the words and conversations clean with the proprietary NeMo Guardrails censoring system. To make background characters in games more relatable, Nvidia ACE uses Riva for real-time conversations powered by its automatic speech recognition and text-to-speech engines.

Last but not least, Nvidia's new Avatar Cloud Engine (ACE) for Games employs the Omniverse Audio2Face animation service that can alter the facial expression of a game character in real time according to the speech track assigned to it by Riva. Audio2Face features Omniverse can be plugged into Unreal Engine 5, allowing easy facial expression manipulation of its MetaHuman characters by game devs.

According to the founder of Convai, whose Kairos game demo below Nvidia used to demonstrate how ACE can liven up NPCs, the Avatar Cloud Engine allows indie game studios to "achieve the latency and quality needed to make AI non-playable characters available to nearly every developer in a cost-efficient way." 

Get the ASUS ROG Strix GeForce RTX 4090 Gaming Graphics Card on Amazon