AI and HPC Workload Convergence

Discover how AI and HPC workload convergence running on an Intel®-based HPC infrastructure can accelerate your innovation. It starts with Intel’s common HPC reference architecture—a combination of high-performance hardware with optimized software that can simultaneously support analytics, simulation and modeling, machine learning, deep learning and other compute and data-intensive workloads. An HPC infrastructure featuring Intel® Xeon® Scalable processors and Cornelis Networks fabric delivers the highly-scalable compute performance AI and analytics applications need.

Creating a converged platform to run simulation and modeling, artificial intelligence (AI), and analytics workloads in a single cluster infrastructure supports breakthrough innovation while increasing the value and utilization of resources. Servers based on Intel® architecture are the ideal foundation for that convergence. This solution brief introduces the challenges and opportunities around running AI and analytics workloads on existing high performance computing (HPC) clusters.

Overcoming Challenges to Realize the Value of HPC: 

• Coexistent frameworks and software stacks for simulation and modeling, analytics, and AI must be made to work together harmoniously.
• Non-HPC-oriented hardware such as servers equipped with accelerators and GPUs draws the interest of many organizations as a potential means to drive greater performance for AI workloads in particular.
• Cultural and operational separation between HPC teams’ focus on bare metal, on-premises clusters, as opposed to the cloud orientations of many AI and analytics teams, which may extend to functional approaches such as DevOps.

Building on a Broad Intel® Technology Foundation

Rather than prescribing a rigid solution, Intel’s architectural approach for simulation and modeling, AI, and analytics includes openness and flexibility as a primary design requirement. In addition to supporting hardware building blocks from the entire range of popular manufacturers, the stack draws on a vast ecosystem of software. Along with tools provided directly by Intel, many open-source and commercial third-party software packages are optimized for performance, scalability, stability, and security on Intel architecture.

Pre-validated combinations of hardware and software building blocks, tailored to specific organizational needs, are available as Intel® Select Solutions. This infrastructure helps accelerate performance while simplifying implementation and reducing risk for end customers associated with data center modernization, using systems that are available from a wide choice of popular server manufacturers.