- Home
- PC Benchmark Software
- OpenCL-Benchmark
OpenCL-Benchmark
What it is
OpenCL-Benchmark is a free command-line program for measuring peak GPU and CPU performance. It runs on Windows, Linux, macOS, and Android, and is aimed at comparing compute, memory, and PCIe behavior. Compute tests cover FP64, FP32, FP16, INT64, INT32, INT16, and INT8 workloads. Memory tests measure both coalesced and misaligned reads and writes. PCIe tests report send, receive, and bidirectional bandwidth, while estimating PCIe generation on the assumption of an x16 link. You can pass device IDs to restrict a run to selected devices, making it useful on systems with more than one accelerator or processor. The project README explains building from source with Visual Studio on Windows or make.sh on Linux, macOS, and Android. Its license permits educational use, non-military research, and non-military commercial use, with attribution and marking requirements for modified versions. Product documentation must include attribution when the software is used in a product. The license provides the program as is and disclaims warranties.
For and against
For
- Measures several GPU and CPU numeric workloads
- Tests coalesced and misaligned memory transfers
- Reports PCIe send, receive, and bidirectional bandwidth
- Runs on Windows, Linux, macOS, and Android
- Supports device IDs and command-line automation
Against
- Requires building from source
- License requires attribution and modification marking
- No warranty is provided
- No free trial; free plan only
The verdict
OpenCL-Benchmark focuses on detailed compute, memory, and PCIe measurements rather than a general-purpose interface. It is a practical fit when command-line operation and cross-platform source builds match your benchmarking workflow.
Who it suits
It suits users comparing GPU or CPU compute and memory performance across supported operating systems. Command-line and device-selection options also fit automated or targeted benchmark runs.
Plans and prices
OpenCL-Benchmark
Free to use for educational use, non-military research, or non-military commercial use · attribution required in product documentation when used in a product · modified versions must be marked
Read 11 Oct 2026 · github.com
On the record
- Audience
- The stated purpose and performance measurements make it relevant to users comparing GPU or CPU compute and memory performance.github.com · read 11 Oct 2026
- Build and use
- The README gives instructions to build from source using Visual Studio on Windows or make.sh on Linux, macOS, and Android.github.com · read 11 Oct 2026
- CLI or automation
- Yesgithub.com · read 22 Sep 2026
- Compute benchmarks
- It measures FP64, FP32, FP16, INT64, INT32, INT16, and INT8 compute performance.github.com · read 11 Oct 2026
- Device selection
- Users can pass device IDs to run the benchmark only on selected devices.github.com · read 11 Oct 2026
- License terms
- The license permits educational use, non-military research, and non-military commercial use, subject to attribution and modification-marking conditions.github.com · read 11 Oct 2026
- Maker
- The GitHub profile identifies ProjectPhysX as Dr. Moritz Lehmann and lists Germany as the location.github.com · read 11 Oct 2026
- Memory bandwidth
- It measures coalesced and misaligned memory reads and writes.github.com · read 11 Oct 2026
- Operating systems
- Windows; Linux; macOS; Androidgithub.com · read 22 Sep 2026
- PCIe bandwidth
- It measures PCIe send, receive, and bidirectional bandwidth, and estimates PCIe generation assuming an x16 link width.github.com · read 11 Oct 2026
- Purpose
- OpenCL-Benchmark is a small program for measuring peak GPU and CPU performance.github.com · read 11 Oct 2026
- Warranty
- The license provides the software as is and disclaims warranties.github.com · read 11 Oct 2026
- Workload areas
- compute performance; memory bandwidth; PCIe bandwidthgithub.com · read 22 Sep 2026
