Intel is leading the way for a new generation of high-performance AI systems for businesses.
Key factors to consider
-
Intel launches Xeon 6 with P-cores, doubling AI and high-performance computing (HPC) performance.
-
Gaudi 3 AI accelerators provide 20% more processing power and offer double the cost-performance compared to H100 for handling LLaMa 2 70B1.
SANTA CLARA, Calif., September 24, 2024
As artificial intelligence (AI) continues to grow across industries, businesses need cost-effective and fast infrastructure. To address this need, Intel has launched the Xeon 6 with Performance-cores (P-cores) and Gaudi 3 AI accelerators. This move reinforces Intel’s focus on delivering strong AI systems that offer excellent performance per watt and a lower total cost of ownership (TCO).
AI is driving a big change in data centers, and companies need choices in hardware, software, and development tools," said Justin Hotard, Intel’s executive vice president. "With Xeon 6 and Gaudi 3, Intel is creating an open system that lets customers run their workloads with better performance, efficiency, and security.
Introducing Xeon 6 and Gaudi 3 AI Accelerators
Intel’s new AI infrastructure brings two key updates for data centers:
-
Intel® Xeon® 6 with P-cores
Built to handle demanding workloads, Xeon 6 offers double the performance of the previous model. It has more cores, twice the memory bandwidth, and AI features in every core. It is designed for AI tasks from edge devices to data centers and cloud platforms.
-
Intel® Gaudi® 3 AI Accelerator
Designed for large-scale AI tasks, Gaudi 3 has 64 Tensor processor cores and eight matrix multiplication engines to speed up AI computations. It includes 128 GB of memory for training and inference, plus 24 ports for fast networking. Gaudi 3 works well with the PyTorch framework and advanced AI models like Hugging Face transformers. Intel also teamed up with IBM to offer Gaudi 3 as a service on IBM Cloud. This partnership aims to reduce costs while boosting AI performance.
AI Systems with Lower Costs
Deploying AI at a large scale requires flexible options, good cost-performance, and easy access to technology. Intel’s x86 system and open ecosystem help businesses build AI systems with great TCO and efficiency. Notably, 73% of GPU-accelerated servers use Intel Xeon as their main processor.
Intel is working with partners like Dell Technologies and Supermicro to build systems that meet customer needs for AI deployment. Dell is co-developing solutions using Gaudi 3 and Xeon 6.
Moving from Prototypes to Production with AI
Building AI systems from prototypes to production is not easy. It requires monitoring, error management, security, and scalability. Intel addresses these challenges through joint efforts with OEMs and partners. They focus on delivering ready-to-use AI solutions for businesses.
These solutions, based on the Open Platform Enterprise AI (OPEA) system, use microservices and are optimized for Xeon and Gaudi AI systems. They are also compatible with applications like Kubernetes, Red Hat OpenShift AI, and Red Hat Enterprise Linux AI.
Increasing Access to Enterprise AI
Intel’s Tiber portfolio helps businesses overcome obstacles like access, cost, complexity, security, efficiency, and scalability in AI, cloud, and edge systems. Intel’s Tiber Developer Cloud now offers preview systems of Xeon 6 for testing. Selected customers will also get early access to Gaudi 3 for validating AI models, with Gaudi 3 clusters expected to roll out soon for large-scale use.
New services, such as SeekrFlow, an AI platform for building reliable AI applications, will also be available. These updates include Intel Gaudi software, Jupyter notebooks with PyTorch 2.4, and AI tools from Intel, supporting the latest Xeon 6 processors.
Explore Our Global Dedicated Server Locations
Your Voice Matters: Share Your Thoughts Below!
Recent Topics for you 
