Intel has unveiled its highly anticipated Intel Gaudi 3 AI accelerator at the Intel Vision event, directly targeting Nvidia in the AI training and inference markets. The new processor boasts significant improvements over its predecessor, Gaudi 2, and other competing products, particularly in processing BF16 data types essential for AI workloads. Manufactured using a 5nm process technology, Gaudi 3 incorporates architectural advancements such as more TPCs and MMEs, providing the computing power needed for parallel AI operations, reducing training and inference times for complex AI models.
The Gaudi 3 AI accelerator expands its hardware capabilities with more Matrix Math Engines and Tensor Cores than its predecessor, significantly increasing processing power for AI tasks. It showcases improved FP8 precision throughput of 1835 TFLOPS, doubling the performance of Gaudi 2, and enhances BF16 performance. With 128GB of HBMe2 memory and 3.7TB/s memory bandwidth, Gaudi 3 efficiently processes large datasets, crucial for training and running large AI models. Additionally, it features high-speed networking with twenty-four 200Gb Ethernet ports, enhancing system connectivity for efficient scaling of AI compute clusters.
In terms of performance, Intel projects that Gaudi 3 will outperform competing products like Nvidia’s H100 and H200 in training speed, inference throughput, and power efficiency for various AI models. The accelerator is expected to deliver a 50% faster training time, superior inference throughput, and power efficiency for parameterized models. Analysts view Intel’s Gaudi 3 AI accelerator as a strategic move to challenge Nvidia and capitalize on the growing demand for advanced AI compute solutions. The improvements in AI compute, memory bandwidth, and networking capabilities position the Gaudi 3 as a powerful solution for next-generation AI applications.
Intel’s emphasis on open community-based software and standard Ethernet networking differentiates its approach from Nvidia, aligning with the industry trend towards open standards and interoperability. By partnering with key OEMs like Dell Technologies, HPE, Lenovo, and Supermicro for the Gaudi 3 rollout, Intel is poised for success in the accelerator market. The launch of Gaudi 3 not only addresses the current AI accelerator market but also paves the way for Intel’s next-generation GPU, Falcon Shores, which will integrate Gaudi and Xe IP for enhanced AI and HPC capabilities.
Overall, the introduction of the Gaudi 3 AI accelerator signifies a significant milestone for Intel, showcasing technological advancements, strategic market positioning, and commitment to meeting the evolving needs of the AI industry. Through substantial performance improvements, open standards adoption, and strategic partnerships, Intel is challenging the status quo in the AI accelerator market and positioning itself as a leader in the next phase of AI infrastructure development.
