3DMark runs fixed graphics, processor, storage, and feature workloads so results from equivalent runs can be compared. Each benchmark targets a defined class of hardware and a particular graphics path. The overall number is not a general speed rating for every task, and one 3DMark test does not share a score scale with another. A fair comparison uses the same test and its default settings.
Choose the workload
The Home screen recommends a test after examining the system, while the Benchmarks screen exposes the other installed choices. Heavy ray-tracing tests target different hardware than lightweight raster tests. CPU Profile changes the focus to processor thread scaling, and the Storage Benchmark measures game-related storage activity. Selecting a test that overwhelms an older device can produce an unusably low result rather than a clearer comparison.
Each graphics test renders at a fixed internal resolution. 3DMark then scales those frames to the current display, so a high-resolution benchmark does not require a monitor with the same resolution. Changing the desktop resolution does not change the internal workload. A custom run can change that workload, but it creates a different experiment.
Default scores
A default benchmark run produces the official overall score and the component results defined for that test. The result page also records detected hardware and performance-monitoring data. Comparisons remain meaningful only when the benchmark name, benchmark revision, and settings match.
Scores from separate tests cannot be placed on one shared ladder. A higher number in a lightweight test does not beat a lower number in a heavier test. Results from custom settings also should not be compared with default scores because the rendered work has changed.
Custom runs
Custom settings can change rendering resolution, aspect ratio, device selection, and other options supported by the chosen workload. 3DMark returns subtest measurements and monitoring information for these runs, but it does not calculate an official overall score. Reset returns the test to its default settings.
On a system with several graphics processors, the rendering-device menu selects the tested GPU. The display choice matters too. If the selected GPU does not connect directly to that display, frames may cross another device or bus, and the transfer can limit the measured rate. A laptop with hybrid graphics needs particular attention to this path.
Stress loops
A stress test repeats a graphics workload continuously without the ordinary loading breaks. It measures whether performance remains stable as temperature rises and boost behavior settles. The default loop set takes long enough to expose many cooling or stability faults that a single score run can miss.
A crash, hang, visual artifact, or large performance drop can indicate unstable hardware, an aggressive overclock, or insufficient cooling. The selected stress test must still match the device. If the system lacks the video memory required by that workload, the run can fail even when the hardware is otherwise stable. That failure should not be treated as proof of a defective GPU.
Explorer scenes
Explorer Mode opens supported benchmark scenes for free camera movement and visual inspection. It can pause a looping scene and expose display options that do not belong to the scored run. Moving around a scene does not produce a benchmark score because the camera path and workload no longer remain fixed.
3DMark can show repeatable performance differences, but it does not diagnose their cause by itself. Driver changes, power limits, cooling, background tasks, and hardware configuration can all move a result. Repeating the same default test after changing one variable gives a clearer record than changing several settings at once.
