ML Commons released a new AI benchmark to test the speed of responses to user queries

Artificial Intelligence Benchmarking Group ML Commons A fresh set of tests and results were released Wednesday that rank the speed at which top-of-the-line hardware can run AI applications and respond to users.

Two new benchmarks added by ML Commons measure the speed at which AI chips and systems can generate responses from powerful AI models loaded with data. The results show roughly how quickly the AI ​​application Chat GPT Can answer a user’s question.

One of the new standards added the ability to measure the speed of query-and-answer scenarios for large language models. Called Lama 2It contains 70 billion parameters and is developed by Meta Platforms.

ML Commons officials also added a second text-to-image generator to their suite of benchmarking tools, called ML PerfBased on Stability of AI Stable dispersion XL model.

Powered by servers Nvidia’s H100 Chips made by the likes of Alphabet GoogleSuper Micro and Nvidia On raw performance alone it easily won both new benchmarks. Several server makers have submitted designs based on the company’s less powerful L40S chip.

Server builder Crye presented a design for an image generation benchmark with a Qualcomm AI chip that draws significantly less power than Nvidia’s latest processors.

Intel also presented a design based on its Gaudi2 accelerator chips. The company called the results “solid”.

Raw performance isn’t the only metric that matters when deploying AI applications. Advanced AI chips suck a huge amount of energy and one of the most important challenges for AI companies is deploying a chip that provides the best performance for the least amount of energy.

ML Commons has a separate benchmark category for measuring power consumption.

© Thomson Reuters 2024

---

(This story has not been edited by NDTV staff and is auto-generated from a syndicated feed.)