In a bid to challenge Nvidia's dominance in the AI chip market, Intel has revealed its latest offering, the Gaudi 3 chip. Per their latest announcement, the new AI chip promises enhanced efficiency and speed advantages over its competitor's GPUs.
Now, this move comes as chipmakers race to meet the demand for semiconductors capable of training and deploying large AI models, including those powering platforms like OpenAI's ChatGPT.
About Intel's Upcoming Gaudi 3 Chip
Intel claims the Gaudi 3 chip is more than twice as power-efficient and one-and-a-half times faster than Nvidia's H100 GPU, offering various configurations to suit different system setups.
It offers up to 50% faster time-to-train on average across various models compared to Nvidia H100, along with 40% better power efficiency for inference tasks.
Additionally, it features a 1.5x increase in memory bandwidth over its predecessor and supports various configurations, from single nodes to mega-clusters. Industry-standard Ethernet networking ensures compatibility with existing systems.
The Intel Gaudi 3 accelerator is set to be available to original equipment manufacturers (OEMs) in the second quarter of 2024.
While Nvidia currently holds an estimated 80% share of the AI chip market, Intel aims to compete aggressively with its new release, expecting strong demand from customers such as Dell, HP, and Supermicro, according to CNBC.
As the data center AI market continues to expand, fueled by growing demand from cloud providers and businesses, Intel anticipates ample opportunities for competition despite Nvidia's stronghold. To bolster its position, Intel is not only focusing on hardware advancements but also collaborating with major players like Google, Qualcomm, and Arm to create open software solutions.
To date, however, Intel is not just going to compete with NVDIAI globally. VCPost reported that Huawei is already on the move as well.
Join the Conversation