The processing of massive amount of data in a parallel fashion and use of GPU hardware acceleration is essential for any analytic or machine learning algorithm in an AI pipeline.

HPC Cluster

 

Historically, VITO has a couple of HPC clusters available to cater to the needs of the 7 different units. To harmonise things and streamline efficiency, VITO has almost completed building a central HPC cluster that is available to all units in the same way. This enables calculations to be completed in a much more performant way. The focus of the new HPC cluster lies on GPU accelerated computing, which is very efficient for machine learning and AI.

GPU Computing & Parallel Processing

 

GPUs have been increasingly used for scientific calculations because they can be up to 100 times as fast as traditional CPUs when it comes to certain types of calculations. At VITO, we make use of GPUs in several projects that demand a lot of computing power. Our data scientists develop specific frameworks for our programmers, shielding the complexity from the researchers.

In a project like Characterise-to-Sort (CtS), an innovative technology specifically developed for the inline characterisation of complex heterogeneous material streams, different cameras provide huge volumes of high-resolution data. GPU and parallel computing prove to be essential to do the sensor fusion, image processing and identification of the particles in real time.