developer.nvidia.com

Query Acceleration Options for Instant Exploratory SQL

Last updated: 7/24/2026

Query Acceleration Options for Instant Exploratory SQL

Summary

To make exploratory SQL feel instant without a new interface, accelerate the queries with engines analysts already use. Running the GPU-accelerated version of Apache Spark, Presto or DuckDB, Polars and pandas keeps the same SQL, notebook, and BI workflows while joins, aggregations, sorting, and filtering move onto GPUs, powered by NVIDIA cuDF.

Direct Answer

Put GPU acceleration under the interface analysts already trust:

Keep the workflow; accelerate the execution.

Takeaway

You can keep your current workflow and code while speeding up exploration by running the cuDF-accelerated version of the engine they already use: accelerated Spark for distributed SQL, accelerated Presto/DuckDB for SQL engines, accelerated pandas/Polars for notebooks.