Intel announces list prices of its Gaudi 3 and Gaudi 2 AI accelerators and we’re in for a shock: rivals to Nvidia’s iconic H100 GPU have a much better performance per dollar ratio, but will that matter?
Intel introduced the Intel
Intel is positioning its Gaudi architecture as a viable and more affordable alternative to Nvidia’s H100. The prices announced at Computex showed that a standard AI kit consisting of eight Intel Gaudi 2 accelerators with a universal base board (UBB) costs $65,000.
Not long ago we reported how Supermicro was selling an Intel server with eight Gaudi 2 AI accelerators, 76 cores, 1TB RAM and 100GbE for $90,000. Intel also revealed that a kit of eight Intel Gaudi 3 accelerators with a UBB will cost $125,000.
Better performances
“AI is driving one of the most consequential eras of innovation the industry has ever seen,” said Pat Gelsinger, CEO of Intel.
“Intel is one of the few companies in the world innovating across the full spectrum of capabilities in the AI market – from semiconductor manufacturing to PC, networking, edge and data center systems. Our latest Xeon, Gaudi and Core Ultra platforms, combined with the power of our hardware and software ecosystem, deliver the flexible, secure, sustainable and cost-effective solutions our customers need to realize the immense opportunities ahead maximize.”
As the market leader, Nvidia can command a premium for its H100 GPU-based hardware. While the Gaudi offering may be more attractive in terms of affordability, Intel has a battle to win over buyers who view Nvidia’s products as superior and are willing to pay more to get the best.
For that reason, Intel not only undercuts its main rival on price, but also wants to sell its products on performance. The veteran tech giant says its Gaudi 3 accelerators deliver performance improvements for both training and inference tasks on leading GenAI models. A cluster with 8,192 accelerators is claimed to offer up to 40% faster time-to-train compared to a similarly sized Nvidia H100 GPU cluster. Intel says the Gaudi 3 also delivers up to 15% faster training throughput for a 64 accelerator cluster compared to Nvidia H100 on select models, along with up to 2x faster inference speed on popular LLMs.
Dell, Hewlett Packard Enterprise, Lenovo and Supermicro are already on board to bring Intel Gaudi 3 to market, but Intel plans to add six more partners to make its AI systems widely available, and these are Asus, Foxconn , Gigabyte, Inventec, Quanta and Wistron.