Gaudi3 is expected to adopt a 5-nanometer chip. The accelerator will provide significant performance improvements, with up to 4 times the BFloat16 processing capability, double the computing power, 1.5 times the network bandwidth, and up to 1.5 times the HBM capacity (144 GB compared to 96 GB). Looking ahead to 2025, Falcon Shores, the successor to Gaudi3, will merge Gaudi's AI capabilities with Intel's powerful GPU in a single package.

All about AI?

Intel is making bold attempts in the field of AI to compete with NVIDIA and AMD. At the Intel Innovation event in September 2023, the company unveiled an ambitious development roadmap for the coming years, clearly indicating their full commitment to AI. This includes their monster-level 288-core Xeon CPU based on the Emerald Rapids architecture, which will be launched next year.

The company also announced that Stability.AI will purchase an AI supercomputer based on Gaudi2, which will be managed by Xeon processors and 4000 Gaudi2 accelerators.

Of particular interest are the processors planned for release in 2024 and 2025, such as Arrow Lake, Lunar Lake, and Panther Lake. These processors represent a major leap forward for Intel in pursuit of technological excellence.

Intel plans to release another version of the AI accelerator superchip, Falcon Shores 2, by 2026. "We have a simplified roadmap because we are merging the GPU and accelerator into a single product," said CEO Pat Gelsinger. While this is a distant vision, the imminent announcements should not be underestimated.

In addition, Intel's Falcon Shores chip was initially conceived as a fusion of CPU and GPU cores, representing the company's first foray into high-performance computing with the 'XPU' architecture. However, a few months ago, Intel shocked the industry by choosing to focus solely on GPUs and delaying the chip's release until 2025. The company's journey in AI and GPU fields has encountered a series of twists and turns.

Developer Focus

The announcement on Meteor Lake, scheduled for release on December 14th, is undoubtedly a headline news, as CEO Pat Gelsinger stated that the processor will "deliver efficient AI acceleration and on-device inference on PCs".

Gelsinger emphasized the company's commitment to engineering excellence and showcased its efforts to democratize AI through the "AI PC" concept, similar to AMD's Ryzen AI PC. This innovation is achieved through the upcoming "Meteor Lake" laptop chip, which combines new AI data processing capabilities.

The AI PC concept aims to bring AI capabilities directly to personal computers, allowing users to run generative AI chatbots like ChatGPT locally without relying on cloud data centers, and even for inference models like Llama 2.

Furthermore, to compete with AMD's ROCm and NVIDIA's CUDA, developers can use the oneAPI programming model to build and optimize AI and high-performance computing workloads. Intel also revealed the Project Strata commercial software platform, set to launch in 2024, which aims to support distributed edge infrastructure and applications, providing modular building blocks and advanced services.

With a focus on developers, Intel has announced the general availability of its Intel Developer Cloud platform, offering developers the opportunity to test and deploy AI and high-performance computing applications on the latest CPUs, GPUs, and AI accelerators.

The platform includes access to the 5th generation Xeon Scalable processors, Intel Data Center GPU Max series, Intel Gaudi2 deep learning processors, as well as Intel software and tools.

All of these clearly indicate that Intel may leave a significant mark on December 14th, as its current roadmap, including the updates to Gaudi3, the Xeon upgrade for AI PC, and the powerful capabilities of Gaudi2, all point in this direction.

While Intel undergoes strategic adjustments, it is worth noting that NVIDIA has also made a significant leap into the CPU market with the GH200 supercomputer, which is already being rolled out. This expansion complements NVIDIA's existing strengths in GPUs and AI technology and ventures into the CPU market.

AMD is also fully committed to AI computing, collaborating with Microsoft, Meta, and multiple OEM companies. In its push for AI activities, AMD has made several AI announcements, including the Instinct MI300X AI accelerator to compete with NVIDIA's H100, updates to ROCm, and the Ryzen AI PC for edge computing.