Intel's plans for 2024 are progressing steadily. The company announced that its latest AI accelerator, Gaudi3, is expected to be officially launched next week. Gaudi3 is hailed as Intel's AI hero, designed to handle large-scale training and inference tasks with limitless scalability.
Intel CEO Pat Gelsinger first teased this accelerator at last year's "Intel AI Everywhere" conference. At the subsequent "Intel Vision 2024" event, Intel reiterated the launch plan for Gaudi3, aiming to address the growing demand for high-performance semiconductors capable of training and deploying large AI models.
Intel claims that Gaudi3's energy efficiency is more than double that of its predecessor and runs AI models 1.5 times faster than NVIDIA's H100 GPU. Compared to NVIDIA, Intel states that its chips have lower power consumption. Notably, NVIDIA currently holds about 80% of the AI chip market share, and its GPUs have been the preferred choice for AI developers over the past year.
Although NVIDIA has technically surpassed Intel and its Blackwell series has been delayed due to mask issues in the chip production process rather than design flaws, the Blackwell series promises significant AI performance improvements over competitors like AMD's MI300X, which AMD also plans to release soon.
Now, with the timing of the product launch, Intel has gained a more competitive position against NVIDIA's Blackwell and AMD's MI300X. The Gaudi3 AI accelerator is designed to significantly enhance the performance of AI applications, supporting the latest generative AI and large language models in data centers. Paired with Intel Xeon processors, Gaudi3 can provide enterprise-level performance and reliability in various applications such as 3D generation, language translation, sentiment analysis, and image generation.
Looking ahead to 2025, Gaudi3's successor, Falcon Shores, will combine Gaudi's AI capabilities with powerful Intel GPUs to form a new product, adding competitive advantages for Intel in the market.
Gaudi3 has performed exceptionally well in tests, supporting Meta's open-source Llama and TII's Falcon models, demonstrating its potential in training and deploying models like Stable Diffusion and OpenAI's Whisper speech recognition.
Key Points:
🌟 Gaudi3 accelerator will be launched next week, featuring high energy efficiency and fast computing capabilities.
⚡️ Intel claims its energy efficiency is double that of its predecessor, outperforming NVIDIA H100.
🚀 Falcon Shores, set to launch in 2025, will further enhance Intel's competitiveness in the AI field.