Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore how NVIDIA's Grace-Blackwell GPU system architecture revolutionizes data analytics performance in this 36-minute conference talk from Databricks. Trace the journey of a single row through a GPU-enabled query engine to understand how GPUs accelerate data processing workflows. Learn about the Grace-Blackwell architecture's key innovation in addressing fast data access challenges, including hardware-accelerated decompression of compact formats and ultra-high-speed data transfer rates of up to 450 GB/s across the CPU to GPU interconnect. Discover how query engines leverage these capabilities when reading large datasets from CPU memory or disk storage. Examine real-world performance demonstrations using a prototype query engine with industry standard benchmark queries, comparing GPU-accelerated results against traditional CPU solutions. See practical applications through Apache Spark and the RAPIDS Accelerator for Apache Spark, showcasing GPU acceleration impact on SQL query performance at 100TB scale using NDS benchmark suite that simulates actual business scenarios. Gain insights from NVIDIA's Senior Developer Technology Engineer Clemens Lutz and Principal Systems Software Engineer Thomas Graves on the future of GPU-accelerated data analytics and query processing optimization.
