How The CPU Stress Test Works
The test spawns a configurable number of WebWorkers (defaulting to navigator.hardwareConcurrency, which matches your logical core count). Each worker runs a tight while loop that feeds the previous hash into a new crypto.subtle.digest('SHA-256', data) call, so every iteration is dependent on the last and the CPU cannot optimize it away. Every 500 ms, each worker posts its operation count and timestamp to the main thread, which converts that to operations per second and sums across threads. SHA-256 was chosen because it is implemented in hardware on every modern CPU, so the throughput ceiling reflects real compute capability rather than JavaScript engine overhead.
What The Test Measures (And What It Doesn't)
This tool measures sustained integer throughput across multiple cores. It is a good proxy for workloads like compilation, compression, video encode, and 3D rendering — all of which are multi-threaded and compute-bound. It is a poor proxy for single-thread gaming performance, memory bandwidth, or GPU-bound work. For a comprehensive picture, pair this test with our GPU stress test (which exercises the graphics pipeline) and memory test (which tracks allocation and heap behavior). Numbers from this browser test are not directly comparable to native benchmarks like Cinebench or Geekbench, but they are comparable across runs on the same machine and browser, which is what matters for throttling detection.
Spotting Thermal And Power Throttling
The single most informative pattern in this test is a full-speed burst followed by a sustained drop. A thin-and-light laptop will typically hold peak ops/sec for 20-40 seconds, then step down 30-50% as the CPU hits its thermal envelope. Desktop gaming rigs with good cooling should hold steady throughout. If you see throttling on a desktop, check: CPU cooler fans spinning, thermal paste age, case airflow, or an overclock that was stable on short benchmarks but falls over on sustained load. On laptops, plugging in vs battery makes a huge difference; "power saver" profiles cap clocks below thermal limits.
Hybrid CPUs And Per-Thread Variance
On Intel 12th-gen and later, Apple Silicon (M1/M2/M3/M4), and various mobile SoCs, not all cores are equal. "Performance" cores (Intel P-cores, Apple P-cores) run at 3-5 GHz and handle heavy threads; "efficiency" cores (E-cores) run at 2-3 GHz and handle background work. When you spawn one worker per logical core, you will see the P-cores hitting 2-3× the ops/sec of the E-cores — that is working as designed. For the most useful benchmark of a hybrid CPU, limit threads to the P-core count and run again.
