All Benchmarks

Benchmarks provide performance measurements and comparative analysis of NVIDIA Holoscan capabilities. These examples demonstrate real-world performance characteristics including threading models, visualization techniques, and system optimization strategies. Use these benchmarks to understand performance implications and optimize your applications for specific hardware configurations and use cases.

New Green Context Benchmarking

Green Context Benchmarking

cpp

Benchmark CUDA Green Context scheduling latency and performance with CUPTI profiling

cuda green context scheduling cupti performance
Real-time Thread Scheduling Benchmark

Real-time Thread Scheduling Benchmark

cpp

This benchmark application demonstrates and evaluates the effectiveness of real-time thread scheduling in Holoscan ap...

benchmarking real-time performance
Holoscan Release Benchmarking

Holoscan Release Benchmarking

Reproducible benchmarking workflow to evaluate HoloHub performance with a Holoscan SDK release.

performance
Exclusive Display Benchmark

Exclusive Display Benchmark

This benchmark shows performance of exclusive display and default mode in presence of multiple headless applications.

benchmarking
Holoscan Flow Benchmarking

Holoscan Flow Benchmarking

This is a benchmarking tool to evaluate the performance of HoloHub and other Holoscan applications.

Model Benchmarking

Model Benchmarking

Python, C++

This application demonstrates an easy, quick and straightforward way to benchmark the scalaibility of inferencing wit...