3DMark

3DMark, from UL Solutions (formerly Futuremark), is the most widely recognized consumer and enthusiast graphics benchmark. It measures gaming-oriented GPU and system performance using purpose-built 3D scenes that stress modern rendering techniques. It is used by reviewers, overclockers, and manufacturers to compare graphics cards, gaming PCs, and laptops.

The suite contains multiple named tests targeting different hardware tiers and APIs: Time Spy (DirectX 12), Port Royal and Speed Way (real-time ray tracing), Fire Strike (DirectX 11), Steel Nomad (cross-platform), Wild Life (mobile and cross-platform), and feature tests for specific capabilities like mesh shaders, DLSS/upscaling, and sampler feedback. Each renders a demanding animated scene at a defined resolution.

What It Measures

3DMark produces an overall benchmark score weighted from a graphics score (GPU-bound rendering frame rate) and a CPU score (a physics or simulation test). Higher scores indicate stronger combined performance. Many tests report per-scene frame rates, and stress-test modes measure frame-time stability over repeated loops to detect thermal throttling and reliability problems.

Methodology

Each test renders fixed scenes deterministically so runs are repeatable. The graphics tests are designed to be GPU-bound, isolating rendering capability, while the CPU test loads processor cores. Scores are normalized against reference hardware so absolute numbers stay comparable across systems. UL maintains a large online results database, and benchmark integrity rules detect manipulated runs. A stress test runs the scene many times to report a stability percentage.

How to Interpret Results

Use the graphics score to compare GPUs and the overall score for whole-system gaming capability, but pick the test that matches your hardware tier and the APIs you use — ray-tracing tests favor cards with dedicated RT hardware. The stress test's stability percentage reveals cooling adequacy; a steep drop indicates throttling. Compare only identical tests and presets, and treat 3DMark as a synthetic proxy rather than a guarantee of real-game frame rates.

Limitations

3DMark scenes are synthetic and may not predict performance in a specific game, which depends on its engine and optimization. Scores are influenced by drivers, background load, and thermal conditions. The benchmark targets gaming and consumer graphics, not professional compute or AI workloads. Cross-version comparisons are invalid because scoring formulas change, and high scores can reflect aggressive overclocking unrepresentative of stock performance.