About this test

A free PC stress test that runs in your browser

SystemCheck loads your processor, memory and graphics hardware for a fixed length of time and records what they deliver the whole way through, not just the best moment. Most benchmarks report a peak, which is the number a machine can hold for a few seconds. The number that decides whether a laptop is pleasant to work on is the one it still holds in the third minute, and that is the number this test is built to find.

There is nothing to install and no account to make. The run happens in this tab, and the report opens in plain words before any of the technical sections.

What the stress test measures

Sustained CPU load

Every logical core is loaded for a full minute and throughput is recorded in 250 ms windows, so the result is what the processor delivered rather than what it reached once. The shape of that curve is what tells you whether the machine held its speed, stepped down when a power budget expired, or slid steadily as cooling fell behind.

Memory bandwidth and latency

Read and write bandwidth from one thread and from all threads at once, and the round-trip time to cache and to main memory. A machine can have a fast processor and still feel slow if it is waiting on memory, and the two are measured separately here so you can tell which.

GPU raster at a fixed resolution

A lit 3D model renders at a fixed 1600 by 900 on the same camera path on every machine, and every frame is timed until the graphics chip has finished it, so frame rate, frame pacing and decay are comparable between runs. Pacing matters more than the average: a steady 45 frames per second looks better than an average of 60 delivered unevenly.

Path tracing

A path tracer runs samples back to back at a fixed 960 by 540, so the graphics chip stays loaded for the whole stage, and reports samples per second and how consistently they were delivered. It is the heaviest thing a browser can ask of a GPU and the stage where a thermally limited laptop usually separates from a desktop.

A deep run adds four more stages: your display’s refresh rate and frame pacing, a combined load with the processor and graphics working at once, soft shadows against hard shadows on one fixed scene, and a repeat of the CPU stage on the already-warmed machine, which is the clearest single test of whether cooling is the limit. The methodology page sets out each stage in full.

What your results mean

Run quality comes first

Before any figure, the report says what would make this run less trustworthy: a tab that was hidden part of the time, a run on battery with its charge level, a fallback to a single core, or contention where it could be measured. If one of these is present, fix it and run again before reading the rest.

Peak, sustained and decay

Peak is the best one-second mean after a short warm-up. Sustained is what the stage delivered over its later part. Decay compares the first seconds after warm-up with the end of the stage, so a machine that boosts hard and then settles lower shows a large decay even when its peak looks strong.

The shape of the curve

A flat curve means the machine held its speed. A single step down part way through is consistent with a power budget ending, and a steady slide is consistent with cooling falling behind, though a run on battery or another program on the cores can look similar. The report names the shape with a confidence that is capped by the run's own measured noise floor.

Frame rate and frame pacing

For the graphics stages, the average frame rate tells you how much work was done, and pacing tells you how evenly it arrived. Uneven pacing is what you notice as stutter, so a lower but steadier result can feel better than a higher but uneven one.

For a longer walk through a report, read why sustained performance matters more than a peak and how to tell a CPU bottleneck from a GPU one. If you want to know how far a browser measurement can be trusted, the guide to browser benchmark accuracy covers it.

Related tests

Each of these looks at one part of the machine on its own, with a page that explains that stage in more depth.

• CPU stress test
• GPU stress test
• RAM speed test
• Laptop stress test
• GPU benchmark
• Bottleneck test
• PC performance test
• Every test

What it cannot measure

A browser is a restricted place to measure from, and pretending otherwise is how benchmarks end up quoting numbers they never read. These are the limits:

• Temperatures in degrees. A browser has no sensor access, so decay is measured as lost throughput and the cause is inferred from its shape, never reported as a reading.
• Fan speed, voltages and power draw in watts. None are exposed to a web page on any browser.
• Your exact processor and graphics model, unless you name the processor yourself. The browser reports a rendering string, not a part number.
• Anything about other machines. Every figure is measured on the device in front of you, in this tab, during this run.

Questions people ask before running it

Where to go next