To see the performance difference of running on GPU over CPU, manually switch where your queries are run:
- Enable timing reporting in
- Ensure that you are in GPU mode (the default):
- Run your queries a few times. Because data is lazily moved to the GPUs, the first time you query new data/columns takes a bit longer than subsequent times.
- Switch to CPU mode:
\cpuOr add the g_cpu_mode hint to the query and run your queries a few times.
- Compare the different runtimes.
If you are using a large data set, you should see a significant difference between the two. However, if the sample set is relatively small (for example, the sample 7-million flights dataset that comes preloaded in OmniSci), some of the fixed overhead of running on the GPUs causes those queries to appear to run slower than on the CPU.
Article is closed for comments.