Intel has revealed details about its upcoming chip for artificial intelligence (AI) computing during a recent conference in Germany.
Intel, which is struggling to grow its business, aims to launch the chip, codenamed “Falcon Shores,” in 2025. The launch of AI chip is part of Intel’s strategic shift to compete against Nvidia and Advanced Micro Devices (AMD), Reuters news report said.
The Falcon Shores chip boasts impressive technical specifications, including 288 gigabytes of memory and support for 8-bit floating point computation. These specifications are crucial as AI models, such as those powering services like ChatGPT, continue to increase in size, driving the demand for more powerful chips to handle their computational requirements.
These details mark an important step for Intel as it pursues its strategy to catch up with Nvidia, the current market leader in AI chips, and AMD, which is set to challenge Nvidia’s position with its upcoming MI300 chip. In contrast, Intel currently holds minimal market share due to delays in the release of its Ponte Vecchio chip, which was intended to compete with Nvidia’s offerings.
Despite the setbacks with Ponte Vecchio, Intel has made progress in delivering the Argonne National Lab’s Aurora supercomputer, which utilizes the Ponte Vecchio chip. Intel claims that Ponte Vecchio outperforms Nvidia’s latest AI chip, the H100.
However, Intel’s next-generation chip, Falcon Shores, will not be available until 2025, by which time Nvidia will likely have released another chip of its own.
Jeff McVeigh, corporate vice president of Intel’s super compute group, explained that the company is taking the opportunity to redesign the chip after abandoning its previous strategy of integrating graphics processing units (GPUs) with central processing units (CPUs). This shift allows Intel to offer discrete CPU and GPU options, enabling customers to select the best combination of components and vendors at the platform level.
