Intel’s Gaudi 3 AI accelerator is now available on IBM Cloud, offering enterprises a scalable and cost-effective solution for AI workloads. This integration positions IBM Cloud as the first cloud service provider to adopt Gaudi 3, providing clients with enhanced performance and flexibility for AI applications.
Key Features of Gaudi 3 on IBM Cloud:
- High-Performance AI Processing: Gaudi 3 accelerators are designed with AI-specific features, including matrix math engines, tensor processing cores, high-bandwidth memory, and built-in Ethernet ports. These components facilitate accelerated inferencing of deep neural networks, supporting the deployment of large language models (LLMs) and retrieval-augmented generation (RAG) applications.
- Scalability and Flexibility: IBM Cloud enables customers to scale AI workloads from single-node configurations with eight accelerators, offering a throughput of 9.6 TB/s, to expansive 1,024-node clusters with 8,192 accelerators, achieving a throughput of 9.83 PB/s. This scalability ensures that enterprises can tailor their AI infrastructure to meet specific performance and budgetary requirements.
- Cost-Effective Solutions: Deploying Gaudi 3 accelerators on IBM Cloud provides a more affordable alternative to existing AI accelerators. Preliminary tests have shown that Gaudi 3 delivers superior performance at a lower cost compared to NVIDIA’s H100 and H200 GPUs, making it an attractive option for enterprises seeking efficient AI solutions.
Performance Benchmarks:
Initial evaluations by Signal65 have demonstrated that Gaudi 3 outperforms NVIDIA’s H100 and H200 GPUs in several AI models. For instance, in tests involving models like IBM Granite (8B) and Meta Llama-3.1, Gaudi 3 achieved higher throughput rates, particularly as batch sizes increased. These results highlight Gaudi 3’s capability to handle complex AI workloads effectively.
Integration with IBM’s AI Platform:
IBM plans to integrate Gaudi 3 support into its watsonx AI and Data Platform, further enhancing AI infrastructure resources for clients. This integration aims to optimize model inference performance, providing a comprehensive environment for developing and deploying AI applications.
Strategic Implications:
The adoption of Gaudi 3 accelerators on IBM Cloud signifies a strategic move to diversify AI hardware options, reducing reliance on a single supplier and fostering a more competitive market. This collaboration between IBM and Intel reflects a commitment to offering clients innovative and secure compute solutions tailored to the evolving demands of AI workloads.
In summary, the availability of Intel’s Gaudi 3 AI accelerators on IBM Cloud provides enterprises with a high-performance, scalable, and cost-effective platform for AI applications. This development enhances IBM Cloud’s AI capabilities, offering clients advanced tools to meet the growing demands of AI workloads.
