HPC and AI Convergence in a Heterogeneous Exascale Cluster
SessionExhibitor Forum
Presenter
Event Type
Exhibitor Forum
Data Management
Extreme Scale Comptuing
TP
XO / EX
TimeThursday, 18 November 20212:30pm - 3pm CST
Location263
DescriptionHigh-performance computing and artificial intelligence have many similarities. Both classes of workloads are very compute-intensive, highly parallel, and address problems that benefit from large clusters. One key difference between HPC and AI is in the level of precision that is often required. The level of data analysis required for HPC applications typically needs double-precision or possibly single-precision. AI, however, frequently requires lower precision, with the reduced precision enabling much higher performance.
An emerging trend gaining popularity for both HPC and AI is the use of various types of hardware accelerators to augment the processor performance for the highly parallel workloads characteristic of both HPC and AI. The heterogeneous compute environments utilizing these accelerators provide increased performance, but at a high cost. The accelerators are expensive, power-hungry, consume server space and require special software support, so they are more difficult to program and maintain.
In 2022 an exascale-class supercomputer is planned for deployment in Slovakia that will deliver mixed-precision floating-point operations that support a range of computations from double-precision to AI floating-point data types and will be deployed in a fully homogeneous environment. The wide range of mixed precision coupled with the homogeneous environment will deliver the highest performance supercomputer in the world for the most demanding HPC and AI workloads, providing 64 AI exaflops over 64 compute racks. This presentation compares HPC and AI workloads, and presents an overview of the Slovakian Supercomputer, which will deliver an optimal solution for both HPC and AI.
An emerging trend gaining popularity for both HPC and AI is the use of various types of hardware accelerators to augment the processor performance for the highly parallel workloads characteristic of both HPC and AI. The heterogeneous compute environments utilizing these accelerators provide increased performance, but at a high cost. The accelerators are expensive, power-hungry, consume server space and require special software support, so they are more difficult to program and maintain.
In 2022 an exascale-class supercomputer is planned for deployment in Slovakia that will deliver mixed-precision floating-point operations that support a range of computations from double-precision to AI floating-point data types and will be deployed in a fully homogeneous environment. The wide range of mixed precision coupled with the homogeneous environment will deliver the highest performance supercomputer in the world for the most demanding HPC and AI workloads, providing 64 AI exaflops over 64 compute racks. This presentation compares HPC and AI workloads, and presents an overview of the Slovakian Supercomputer, which will deliver an optimal solution for both HPC and AI.
Presenter
