Intel released the new Gaudi 3 series AI accelerator at the "Let AI Everywhere" event, aiming to seize the growing AI market. This accelerator, designed for deep learning and large-scale generative AI models, is scheduled to be launched next year and is expected to gain a larger market share in 2024 due to its performance advantages and extremely competitive pricing. Gaudi 3 will use the advanced 5nm process technology, which will significantly improve bandwidth, BF16 power and network computing power, and be equipped with up to 128GB of HBM3e memory, directly benchmarking Nvidia's H200 accelerator card.
At today's "Let AI Everywhere" event, Intel CEO Pat Gelsinger also revealed to the public for the first time the Intel Gaudi 3 series AI accelerator, which is designed for deep learning and large-scale generation of artificial intelligence models. The tool is scheduled to be launched next year. According to Intel, the new generation Gaudi 3 AI accelerator is expected to occupy a larger market share in 2024 due to its excellent performance advantages and highly competitive total cost of ownership and pricing. In terms of technology, the new generation of Gaudi 3 will use the advanced 5nm process technology. Compared with the previous generation Gaudi 2 (7nm process technology), its bandwidth will be increased by 1.5 times, BF16 power will be increased by 4 times, and network computing power will be increased by 2 times. At the same time, it will also be equipped with up to 128GB of HBM3e memory, which will greatly improve AI learning and training performance and directly compete with NVIDIA's H200 accelerator card.The release of Gaudi 3 marks Intel's further efforts in the field of AI accelerators. Its performance improvements and memory upgrades will have a significant impact on the development and deployment of AI applications. Future development is worth looking forward to. This will further intensify competition in the AI chip market and provide users with more choices.