Intel has launched two new products: The Xeon 6 CPU and the Gaudi 3 AI GPU. These chips aim to meet the growing demand for AI compute solutions. The Xeon 6 is priced as a high-end CPU, while the Gaudi 3 is positioned as a competitive AI accelerator.

The Xeon 6 ("Granite Rapids") features performance-cores (P-cores) and reportedly offers 2x the performance of its predecessor. It includes increased core count, double the memory bandwidth, and embedded AI acceleration capabilities. The Gaudi 3 ("Falcon Shores") AI Accelerator is designed for large-scale GenAI, featuring 64 Tensor processor cores, eight matrix multiplication engines, and 128GB of HBM2e memory. Additionally, the company claims that Gaudi 3 also offers compatibility with the PyTorch programming framework and Hugging Face transformer and diffuser models.

Intel claims the Gaudi 3 offers favorable cost/performance compared to NVIDIA's H100 GPU, delivering ~1.09x inference throughput and 1.8x performance per dollar on Meta's LLaMA 3 8B large language model. The company also emphasizes that these new products enable an open ecosystem, allowing customers to implement all their workloads with greater performance, efficiency, and security.

Analyst QuickTake: Intel is aggressively trying to position itself in the AI compute market with this recent launch. These products aim to compete directly with NVIDIA, offering allegedly enhanced performance and cost efficiency against its previous flagship model. While Intel has been on a rocky footing lately, receiving a potential acquisition offer from Qualcomm, a rumored USD 5 billion investment from Apollo Global Management could provide much-needed respite for the company.