Intel’s Gaudi 3 AI Accelerator, the latest product from Habana Labs, a division of Intel, is reportedly being built on the TSMC 5nm process node. The new chip is expected to bring significant improvements in memory, computing, and networking capabilities compared to its predecessor, Gaudi 2.
During the recent unveiling of Intel’s 5th-Gen Xeon Scalable and Core Ultra “Meteor Lake” CPUs, CEO Pat Gelsinger offered a preview of the company’s next-gen AI accelerator. While specific details were not disclosed, the glimpse was enough to stir up excitement in the tech community.
Reports from Korean media have since shed light on what Intel’s Gaudi 3 accelerator might entail. Gaudi 3, a product of Team Blue, is expected to be a formidable contender in the AI performance arena, going head-to-head with NVIDIA’s H100 and AMD’s Instinct MI300X AI GPUs.
Intel’s commitment to AI is evident in its product roadmap. The company is not only developing advanced AI chips like Gaudi 3 but also integrating AI capabilities into its other products. For instance, Intel recently launched the 5th Gen Intel Xeon enterprise and data center CPU family, code-named “Emerald Rapids,” built with AI acceleration in every core.
In conclusion, the development of the Gaudi 3 AI Accelerator marks an important step in Intel’s AI journey. As the AI market continues to grow, Intel’s innovative products like Gaudi 3 are set to play a crucial role in shaping the future of AI. Intel has shown strong belief in its strategy for the future of AI markets, with a hinted emphasis on inferencing. The evolution of the Gaudi platform will be intriguing to observe, especially as Team Blue is gearing up to launch its next-generation accelerator in 2024.
The Gaudi 3 AI Accelerator is a process shrink of Gaudi 2, which was built using the 7nm process. The new chip will have much more memory, computing, and networking than its predecessor. This development is part of Intel’s broader strategy to strengthen its position in the AI market, which is expected to exceed $300 billion by 2026.
Our Gaudi roadmap remains on track with Gaudi3 out of the fab, now in packaging and expected to launch next year. In 2025, Falcon Shores brings our GPU and Gaudi capabilities into a single product.
Pat Gelsinger (Intel CEO) via SeekingAlpha
While the Gaudi 3 AI chip has shown the potential to handle large language models powering applications like ChatGPT, it has been reported that the AI chip posted better training times for BERT – which is for large language models – than Nvidia’s A100 chips, but lagged the latest H100, which is based on Hopper architecture. However, it is expected that the Gaudi 3 accelerator will be very competitive in price to performance when compared to NVIDIA H100.
The Gaudi 3 accelerator is not just another chip in the market. It is expected to compete fiercely with NVIDIA’s H100 and AMD’s Instinct MI300X AI GPUs in terms of AI performance. While the specifics are yet to be revealed, the anticipation surrounding the Gaudi 3 suggests that it could potentially disrupt the current AI GPU market dynamics.
In conclusion, the unveiling of the Gaudi 3 AI Accelerator marks a significant milestone in Intel’s AI journey. As the AI market continues to grow, products like Gaudi 3 will play a crucial role in shaping the future of AI.
The AMD Radeon RX 7900 GRE (Golden Rabbit Edition) is an incredibly powerful graphics card, capable of handling almost anything…
In the ever-evolving realm of graphics processing units (GPUs), NVIDIA, a titan in the field, has consistently demonstrated its strategic…
AMD’s FidelityFX Super Resolution (FSR) is a cutting-edge open upscaling technology that has been making waves in the gaming world.…
There’s been a recent update to Windows 11, known as KB5035853, and it’s causing some problems. Specifically, it’s affecting devices…
Dragon Ball: Sparking! ZERO, the latest addition to the Dragon Ball gaming franchise, has been making waves in the gaming…
Nvidia’s Blackwell GPUs are the successors to the Ada Lovelace series. They are expected to hit the market in late…
This website uses cookies.