← Back to Articles Explainer

3DMark Scores Explained: What TimeSpy, Port Royal, and Speed Way Actually Measure

3DMark is the most widely used GPU benchmark, and its scores appear constantly in hardware discussions. But TimeSpy, Port Royal, Speed Way, and FireStrike all measure different things at different conditions. Knowing what each test is actually doing is necessary to interpret the number correctly.

Published June 2026 — VoltGround

3DMark benchmarks are produced by UL (formerly Futuremark) and are the closest thing GPU testing has to a standardized score. The value of a standardized test is that results from different testers on different systems can be directly compared—provided they ran the same test, which is the part most discussions skip over. TimeSpy and FireStrike produce entirely different numbers and measure different capabilities.

FireStrike and FireStrike Extreme

FireStrike is a DirectX 11 benchmark that renders at 1080p (standard) or 1440p (Extreme). It is an older test but still used for comparisons with legacy hardware. FireStrike scores are significantly higher numerically than TimeSpy scores for the same GPU because of how the score formula is calibrated. A high-end desktop GPU might score 40,000 to 55,000 in FireStrike and 20,000 to 26,000 in TimeSpy Extreme on the same run. These numbers are not directly comparable: FireStrike is DX11 at 1080p, TimeSpy Extreme is DX12 at 4K.

TimeSpy and TimeSpy Extreme

TimeSpy is the standard modern raster performance benchmark. It uses DirectX 12 and renders its two graphics tests at 2560x1440. TimeSpy Extreme doubles the resolution to 3840x2160. The Extreme variant is more GPU-limited and less CPU-limited because the increased pixel count drives the GPU harder relative to the CPU overhead of the DX12 workload. For GPU performance comparison without CPU influence, TimeSpy Extreme is more useful than standard TimeSpy.

Both tests produce a single overall score derived from the average FPS across Graphics Test 1 and Graphics Test 2. The CPU score (from a shorter CPU test) is provided separately and does not factor into the main score.

Port Royal

Port Royal is a dedicated ray tracing benchmark. It uses DX12 with DXR (DirectX Raytracing) to render a single complex scene with real-time reflections, shadows, and global illumination computed via hardware ray tracing. A GPU without dedicated ray tracing hardware (RT cores on NVIDIA, Compute Units on AMD) will score much lower relative to its raster performance than one with dedicated hardware.

Port Royal scores are much lower numerically than TimeSpy or FireStrike. An RTX 4090 at stock settings scores around 23,000 to 25,000 in Port Royal versus 24,000+ in TimeSpy Extreme. The similar numbers are coincidental; the tests are measuring completely different rendering workloads.

Speed Way

Speed Way is a newer test designed to represent a realistic DirectX 12 Ultimate workload combining rasterization, ray tracing, and mesh shaders in a single pass. It targets hardware capable of running modern engines at high quality settings and is not comparable in scale to either TimeSpy or Port Royal scores. A high-end GPU scoring 10,000 to 12,000 in Speed Way is normal.

Comparison rule: Only compare scores from the same benchmark. A FireStrike score and a TimeSpy score tell you nothing about relative performance between two GPUs. Always specify which test produced the number.

Using 3DMark to measure overclock gains

3DMark is well-suited for measuring overclock gains because the tests are deterministic and repeatable. Run three passes of TimeSpy Extreme at stock settings, average the scores. Apply your overclock. Run three more passes, average. The percentage difference is a reliable measure of raster performance improvement. Port Royal gives you the same measurement for ray tracing workloads, which often respond differently to memory overclocking than raster benchmarks.