ApexMark › CPU stress test
CPU stress test online
See your CPU build a ray-traced scene and follow its frame rate live. Use a short session for a quick performance preview, or keep rendering to observe sustained load. Adjust resolution and complexity while the graph records the effect.
Open CPU tortureA CPU workload you can actually see
The scene contains 143 spheres. Browser workers calculate separate strips of the image; only an entirely finished image counts as a frame.
The worker pool follows the logical concurrency reported by your browser, with a maximum of 16 workers. This is a browser rendering workload: it does not measure a physical CPU utilization percentage or guarantee that every processor core is fully occupied.
A quick demo or a sustained rendering session
Open Live test in the menu and choose CPU torture. Rendering starts automatically after preparation while the page is visible. If your browser requests reduced motion or the page is hidden when opened, choose Start once it is visible.
For a longer observation, leave the page visible and active. There is no fixed duration limit. Keep the browser, view, power conditions and background activity comparable when repeating a session.
Adjust the image and the amount of work
Resolution changes the number of pixels calculated for each complete image. The five complexity levels select 1, 2, 4, 8 or 16 samples per pixel.
Change one control at a time if you want to understand its effect. Current statistics restart when the replacement workload is ready, while the graph keeps its recent history.
- Choose a resolution between 160 × 90 and 1280 × 720.
- Adjust complexity, keeping the camera view steady for comparisons.
- Use Stop to end the workload; restart explicitly for a fresh run.
Read FPS and 1% low together
Live FPS reflects completed images, including worker delivery, scheduling and canvas delivery. The large reading averages frame times over the latest second.
The rolling 1% low describes the slowest completed frame durations. It becomes available after 100 frames and uses up to the latest 1,000 frames. It can take longer to appear at low FPS. A 1% low well below the current FPS indicates uneven frame delivery during that window; it does not identify the cause.
Keep a useful record of setting changes
Changing resolution or complexity preserves the recent graph and session duration. Applied changes appear as orange markers, with the time and before-and-after settings.
Current FPS, average, minimum and the 1% low window restart for the new settings. Stop freezes the traces; an explicit restart begins a fresh run.
Choose live observation or a scored benchmark
Use this continuous test to demonstrate browser ray tracing and observe its response to adjustable load. Use the separate ApexMark CPU benchmark when you want a fixed protocol and a verified score.
A session runs in your browser without an account; no live score is saved or submitted. These observations do not certify hardware health, thermal stability or game performance. A slowdown alone cannot identify its cause or justify a hardware upgrade.
Frequently asked questions
What is the difference between a CPU stress test and a CPU benchmark?
This live test keeps rendering at adjustable settings until you stop it. The CPU benchmark runs a fixed, verified protocol to produce a score. Choose live rendering for observation and the benchmark for comparable scored runs.
Does it use all my CPU cores?
It creates workers using the logical concurrency exposed by the browser, capped at 16. Browser scheduling and other activity still matter. The worker count does not establish a physical CPU utilization percentage.
Are these FPS the same as gaming FPS?
No. Each frame is a complete image from this browser ray tracer. The measurement includes delivery and scheduling. It neither predicts a commercial game’s frame rate nor measures monitor refresh rate.
Why is the 1% low initially unavailable?
It needs at least 100 complete frames. Once enough frames exist, it uses the slowest 1% of durations from a rolling window of up to 1,000 completed frames. Changing settings restarts that window.
Why does the test stop when I switch tabs?
Hiding the page stops the workers and releases the renderer. Returning to the tab requires an explicit restart, so a hidden session cannot quietly continue consuming resources.
Can this diagnose a failing processor or prove stability?
No. It observes a specific browser workload and its variation. It does not certify hardware health, identify thermal throttling or replace an appropriate hardware diagnostic. No account or public score submission is involved.