New Delhi: Intel has recently introduced a new AI chip, the Gaudi 3, at the Vision event with the aim of challenging Nvidia’s dominance in AI semiconductor technology.
The newly launched chip is capable of training large language models 50% faster than Nvidia’s H100 processor and can easily generate AI responses at a faster rate.
This announcement was made amid a rush by tech firms to find alternative chip sources that are important for AI.
Revealed at the Intel Vision 2024 event, the Gaudi 3 forms an essential part of Intel’s fresh, open, scalable systems, including next-generation offerings and strategic partnerships that are aimed at increasing the adoption of generative AI.
Pat Gelsinger, CEO of Intel, said, “Every company is quickly becoming an AI company, and Intel is bringing AI everywhere across the enterprise.”
Well, Gaudi 3 will be made available to leading tech giants, including Dell Technologies, HPE, Lenovo, and Supermicro, in the second quarter of 2024.
In addition, Intel has also decided to collaborate with Bharti Airtel, Infosys, and Ola/Krutim so that they can deploy Gaudi accelerator solutions.
Bharti Airtel: Airtel wants to use its phone data to improve its AI and customers’ experience of its services. This will also help Airtel make more money.
Infosys: Infosys is teaming up with Intel to use its latest technologies, like powerful processors and AI tools, in Infosys Topaz. This will help Infosys to provide better AI services to its businesses.
Ola/Krutrim: Krutrim is an AI project that will use Intel’s new tech to train its first big AI model. In return, this model will help them to generate text in 10 different languages.
Intel has also decided to collaborate with Google Cloud, Thales, and Cohesity, which will make use of Intel’s computing capabilities in their cloud services.
This will consist of Intel Trust Domain Extensions, Intel Software Guard Extensions, and Intel’s attestation service.
The Gaudi 3 chip is built using Taiwan Semiconductor Manufacturing Co’s 5nm process and will be available to different server builders, including Hewlett Packard Enterprise and Supermicro, in the 2nd quarter of 2024.