Aiming to reclaim its dominance in the datacenter market, Intel unveiled a couple of new AI chips to give the unchecked advances of NVIDIA and AMD some serious pause. The two new AI server chips, the Xeon 6 CPU and Gaudi 3 AI accelerator, promise enhanced performance and energy efficiency, aimed to reinforce Intel’s position as a formidable contender in the rapidly evolving AI landscape.
This dual chip launch announcement comes hot on the heels of a Wall Street Journal report that Qualcomm is looking into a potential takeover of Intel to bolster its own chip business. Rumours of potential investment into Intel have also surfaced, suggesting that even its rivals and private equity firms are recognising the company’s strategic value in the semiconductor industry.
Also read: How Intel’s Lunar Lake is Taking the Fight to its Competition
Let’s take a quick dive into the all new Xeon 6 and Gaudi 3 AI accelerator chips unveiled by Intel as successors to Xeon 5 and Gaudi 2, respectively.
Intel Xeon 6 is a significant upgrade over its predecessor, offering double the performance for AI and HPC workloads. This powerful processor features increased core count, double the memory bandwidth, and AI acceleration capabilities embedded in every core. Designed to handle compute-intensive workloads with exceptional efficiency, Intel Xeon 6 is engineered to meet the performance demands of AI from edge to datacenter and cloud environments.
The launch of Intel Xeon 6 with Performance-cores (P-cores) marks a major milestone for the company’s commitment to delivering powerful AI systems. With this new processor, customers can expect improved performance per watt and lower total cost of ownership (TCO), making it an attractive option for businesses looking to deploy high-performance AI workloads.
Intel Gaudi 3 is a cutting-edge AI accelerator specifically optimised for large-scale generative AI applications, featuring 64 Tensor processor cores (TPCs) and eight matrix multiplication engines (MMEs) to accelerate deep neural network computations. With 128 gigabytes (GB) of HBM2e memory for training and inference, Intel Gaudi 3 offers seamless compatibility with the PyTorch framework and advanced Hugging Face transformer and diffuser models.
Also read: Intel Vision 2024: Intel Gaudi 3 boasts 50% better performance over NVIDIA H100 AI processor
In short, the Gaudi 3 is specifically designed by Intel for powering generative AI applications, such as natural language processing (NLP) and image generation. What’s more, Intel Gaudi 3 is designed to be energy-efficient, making it a cost-effective choice for large-scale generative AI deployments.
The launch of Intel Gaudi 3 marks a significant step forward in the company’s efforts to deliver powerful AI systems that meet the demands of modern businesses. With its improved performance and efficiency gains, Intel Gaudi 3 is poised to become an attractive option for customers looking to deploy large-scale generative AI workloads.
It’s all about doubling AI performance, enhancing throughput, and optimising cost. The Intel Xeon 6 and Gaudi 3 AI Accelerator offer several key benefits that make them ideal for businesses looking to deploy high-performance AI workloads.
For instance, Intel claims Xeon 6 delivers twice the performance of its predecessor, making it an attractive option for businesses looking to improve their AI and HPC workloads. Similarly, Gaudi 3 boasts up to 20% more throughput compared to H100 for inference of LLaMa 2 70B, making it a powerful tool for large-scale generative AI.
Intel Gaudi 3 offers 2x price/performance compared to H100, making it an attractive option for businesses looking to optimise their costs while maintaining high-performance capabilities. Both Intel Xeon 6 and Gaudi 3 are designed to deliver improved performance per watt, reducing energy consumption and lowering total cost of ownership (TCO). The new processors offer seamless compatibility with popular AI frameworks, such as PyTorch, making it easy for businesses to integrate them into their existing infrastructure.
Also read: From silicon to cloud: CEO Pat Gelsinger shares Intel’s vision of AI everywhere
Intel also announced collaboration with OEMs like Dell Technologies and Supermicro to create custom-built systems for their customers’ AI workload requirements. These co-engineered solutions are designed to optimise AI performance and efficiency by leveraging the Open Platform Enterprise AI (OPEA) platform – an industry ecosystem orchestration framework supported by 40+ tech companies committed to efficiently integrating GenAI technologies and workflows quicker to unlock business value. Optimised for Xeon 6 and Gaudi 3 AI systems, these solutions enable customers to seamlessly integrate applications from Kubernetes, Red Hat OpenShift AI, and Red Hat Enterprise Linux AI seamlessly, according to Intel.
Just to give you a sense of Intel’s footprint in the datacenter market, as much as 73% of GPU-accelerated servers in the cloud use Intel Xeon as the host CPU – which means datacenters that are running massive AI workloads on NVIDIA chips in the cloud still largely need an Intel Xeon CPU to run at the heart of each system. And with Intel’s latest Xeon 6 chip boasting of powerful P-cores, delivering twice the performance of its predecessor and specifically designed for AI and high-performance computing, the company believes Xeon 6 is well-suited for both edge as well as cloud environments.
The Gaudi 3, Intel’s dedicated generative AI processor, is a direct competitor to NVIDIA’s H100 and AMD’s MI300X. This chip is specifically engineered to excel in generative AI applications. Ultimately, the Xeon 6 and Gaudi 3 are designed to complement each other, allowing Intel to offer a comprehensive solution for customers aiming to deploy all sizes and shapes of AI workloads.
These AI chips couldn’t have launched at a more pivotal time in Intel’s history, where it’s facing unprecedented competition from the likes of NVIDIA and AMD just as its own stock performance continues to slide. Intel has also announced it would cut 15% of its workforce. Amid these challenges, CEO Pat Gelsinger is attempting to return Intel to its former glory by pushing teams to build more advanced chips for datacenters and consumer PCs, while simultaneously expanding its manufacturing capabilities.
The reported acquisition interest is another knock on Intel, as companies like Qualcomm look to significantly expand their chip business into the datacenter and PC markets. Qualcomm relies heavily on its smartphone segment, but smartphone sales have slowed over the years as customers hold onto their handsets longer, prompting the company to seek new growth opportunities. One such opportunity includes building laptop chips meant to rival Intel’s own line of processors, though it will take considerable time for Qualcomm to chip away at Intel’s PC market share if it manages to do so at all.
Also read: Intel’s Santhosh Viswanathan opines on what role will AI play in India’s future
