Intel announced the new Jaguar Shores AI accelerator during the recent SC2024 supercomputing conference. This product fits into the company’s strategy to compete in the artificial intelligence market dominated by NVIDIA and AMD.
Jaguar Shores represents an important piece in Intel’s plans to regain market share in the AI sector. While not many technical details have been revealed, it is speculated that it could be the successor to Falcon Shores, scheduled for 2025, and that it is mainly intended for AI inference.
Intel hopes the 18A node will give it a manufacturing edge
Intel is focusing heavily on this segment, where it hopes to be able to exploit its Gaudi e chips upcoming technologies like Falcon Shores and Jaguar Shores to catch up with competitors. It’s not yet clear whether Jaguar Shores will be a GPU or an ASIC, but the product roadmap suggests it could be a next-generation GPU for enterprise AI applications.
In recent years Intel has struggled to keep pace in the AI hardware industryparticularly in the GPU segment. The previous attempt to develop a GPU, Rialto Bridge, was canceled due to lack of customer interest. Plans for Falcon Shores have also undergone several revisions.
The company has essentially ceded the AI training market to NVIDIA and AMD, with the former in a dominant position thanks to its GPUs and CUDA software stack. The now ex-CEO of Intel, Pat Gelsinger, admitted in the past that the company is currently behind NVIDIA, AWS, Google Cloud and AMD in the race for AI.
To catch up, Intel is betting big on its upcoming 18A manufacturing nodewhich incorporates new technologies such as RibbonFET transistors and backside power delivery. This process promises to improve the efficiency and performance of chips, potentially giving Intel an edge over rivals like TSMC.
Despite the difficulties, Intel continues to invest in the artificial intelligence sector, trying to leverage its manufacturing skills of semiconductors and expanding its AI product portfolio. The challenge for the company will be to translate these efforts into competitive products that can erode NVIDIA’s current dominance in the AI hardware market.
Source: www.tomshw.it
