GPU support for UDFs

This article is a short introduction to how you can use GPU support for UDFs in Exasol.

Introduction

In on-premises installations of Exasol 2025.2 and later you can utilize GPUs to accelerate parallel processing for user defined functions (UDFs) in Exasol clusters. This article is a general overview of how GPU support is implemented in Exasol, and how you can use it with your custom UDFs.

Utilizing GPUs for parallel processing in an Exasol cluster offers a significant performance increase for certain types of workloads, specifically those that require repetetive computations. For example, GPUs can tangibly accelerate AI and machine learning workloads, including model training and model inference. GPUs can also accelerate vector and semantic search workloads.

Although GPUs are more expensive than CPUs, the performance improvements can often justify or even offset the cost, for example, compared to running a cloud based CPU workload for a long period of time.

Prerequisites

The support for GPUs in Exasol allows you to offload some of the computation from the CPU to the GPU. However, the CPU must be able to handle preprocessing and feeding the data to the GPU. When planning the hardware for a new Exasol system, you need to consider additional requirements on CPU and RAM to enable GPU support.

To learn more about hardware and software system requirements and how to enable GPU support for UDFs in a new deployment, see Install and configure GPU support in the Administration section.