clpeak

clpeak is a cross-platform benchmark for people who need comparable measurements of CPUs, GPUs and NPUs rather than a single headline score. It suits hardware engineers, developers and technically minded buyers working across operating systems and accelerator APIs. Its breadth and automation are compelling for a free tool, while unsigned Windows and macOS releases make installation less seamless.

Overview

clpeak runs the same benchmark families through the device APIs exposed by each machine, allowing side-by-side comparison between systems. It measures compute throughput, memory bandwidth, matrix throughput, kernel-launch latency and AI inference, with additional CPU-focused operations. That makes it useful for checking whether a particular processor or accelerator delivers the expected performance in a real software stack, not just comparing nominal specifications.

Results remain dependent on the APIs and hardware available on each system. A test that cannot use a given backend or device will not provide an equivalent datapoint, so clpeak is best for controlled comparisons where you understand the underlying platforms.

Key features

Coverage spans general compute, matrix engines, vendor libraries, bandwidth, latency, AI workloads and CPU-specific tests. GPU backends include CUDA, ROCm/HIP, Metal, oneAPI/SYCL, Vulkan and OpenCL. AI runtime support covers ONNX Runtime, Core ML and LiteRT. This is an unusually broad combination for an open-source benchmark and lets teams evaluate the APIs they actually deploy, although cross-device results should be interpreted in the context of each backend.

The app lists detected devices and streams measurements as they are recorded, so you can see progress instead of waiting for a single final score. Run history can be renamed and exported as JSON, which helps preserve meaningful labels for different drivers, machines or configurations.

The command-line interface is the stronger choice for repeatable work. It can select devices, backends and test categories, explain what measurements mean, save runs as JSON and compare them with saved runs. Those controls support scripts, regression checks and lab workflows; casual users who only need an occasional number may find the configuration unnecessary.

Pricing

The Free plan costs nothing. It is GPL-3.0-licensed open-source software with no free trial because the product is already free. There are no paid tiers, seat charges or published usage quotas, so the full application and CLI are available without a subscription commitment. The trade-off is that support is project-based: bug reports are directed to a GitHub issue with verbose run diagnostics attached.

Platforms

clpeak supports Windows, macOS, Linux, iOS and Android. Releases include platform zip files with the app and CLI, backend variants, and a macOS DMG; Linux users can also install through Snap, while Android is distributed through Google Play. Windows and macOS users should expect a first-launch warning because release binaries are not signed with a developer certificate. That prompt is an installation hurdle, not a benchmark limitation, but it may matter in managed environments.

Who it's for

Choose clpeak if you compare heterogeneous hardware, need measurements from a specific GPU or AI runtime, or want JSON records that can be automated and compared later. Developers validating accelerator choices and engineers tracking performance across operating systems get the most from its backend selection and saved-run comparisons. It is less suitable for readers seeking a universally recognized public ranking, since it has no comparison database, or for nontechnical users who prefer a guided score with minimal setup.

Pros and cons

Pros

• Broad accelerator coverage: CUDA, ROCm/HIP, Metal, oneAPI/SYCL, Vulkan and OpenCL span major GPU ecosystems.
• AI-specific measurements: ONNX Runtime, Core ML and LiteRT make the tool relevant to inference hardware decisions.
• Automation-ready output: CLI controls, JSON export and saved-run comparisons support repeatable benchmarking.
• Cross-platform reach: Windows, macOS, Linux, iOS and Android allow one project to cover varied test machines.

Cons

• No comparison database: You must maintain and interpret your own saved runs rather than browse a central ranking.
• Backend-dependent results: Available APIs and devices determine which tests can run and how comparable results are.
• Unsigned desktop binaries: Windows and macOS show security prompts before first launch.
• Issue-based support: Troubleshooting goes through GitHub reports with verbose diagnostics, not a dedicated support channel.

Alternatives

PC Benchmark Software is a broader starting point when you want to survey benchmark tools by category.

Blender is a free, open-source Linux, macOS and Windows suite with offline 3D creation, better suited to production work than low-level throughput measurement.

CrossMark fits users who want a freemium benchmark with a dashboard and automatic online result submission across desktop and mobile platforms.

Geekbench is preferable when you want established CPU and GPU compute benchmarks with online results management and an optional 99.00 USD one-time Pro plan.

Nexus is another free, cross-platform option for readers whose priority is its product-specific workflow.

Basemark GPU suits home or private users who want a free GPU score and online submission, accepting internet dependence and limited breakdown analysis.

V-Ray Benchmark is the better fit for CPU and GPU rendering tests with online result comparison.

Blackmagic RAW Speed Test targets codec and SDK performance rather than general compute, with no ongoing licensing fees.

DaVinci Resolve is for editing and color workflows, not a dedicated hardware benchmark.

Verdict

clpeak is a strong choice for developers and hardware evaluators who need broad API coverage, machine-readable records and repeatable CLI comparisons at no cost. Pick it when your question involves a particular device backend or AI runtime. Look elsewhere if you need a public ranking database, a polished one-click experience or signed desktop installers.