See how KDB-X GPU Acceleration speeds up transaction cost analysis (TCA), qSQL queries, and real-time trading analytics without requiring developers to rewrite their applications in CUDA or C++.
KX CTO Paul Colgan explains how the KDB-X GPU module works before demonstrating CPU and GPU performance side by side across historical and streaming TCA workloads.
In this video, you’ll see:
- How KDB-X distributes workloads between CPUs and NVIDIA GPUs
- How q and qSQL operations are GPU-accelerated
- TCA over 20 million trades and 200 million quotes
- A query completing in 0.7 seconds on GPU versus 10.8 seconds on CPU
- Continuous TCA recalculation over streaming market data
- Customer results reaching approximately 10× end-to-end acceleration
- Workloads suited to GPU acceleration, including joins, sorting, backtesting, risk simulation, and end-of-day processing
- How the KX GPU Migration Skill identifies acceleration opportunities in existing q code
Explore KDB-X GPU acceleration
View the GPU-accelerated TCA blueprint