A new way to benchmark games. Introducing the GFP Score.

How It Works

The GFP Score is a grading system designed to apply a score based on the level of performance you attain, versus how visually complex the game is. This is designed to be used as a tool to point out games that may look worse than others, while performing worse at the same time. Instead of benchmarking the hardware like we’ve always done, we’re benchmarking the game itself.

The tool can also be used to find games that have matching scores, so gamers can pre-emptively know how well they can expect a game to run on their system. Or, it can be used to roughly benchmark a game’s overall optimization. The “perfect” score would land at 10,000 points or above, while an ideal score is above 4,000 points, with 5,000 points being main target to reach. Anything below 2,500 points is considered a fail. The scores are based on unbiased observations of the game’s graphical complexity. This provides a new way to benchmark how well a game runs, completely unrelated to the hardware that runs it.

With each generation of PC hardware that gets released, our test bench will be updated to target the ideal “mid-range” PC to base our scores on hardware that the majority of consumers are actually in the market for. Since mid-range targets 1080p resolution, that is the resolution used for testing. The target settings will be set based on what is most efficient for the game. Not all setting presets are universally applicable. This allows you to be able to find games that look on par with others, despite using a higher/lower settings preset, while also providing a settings guide you can follow to get the best experience in the game.

(Example: Low/medium settings on Alan Wake 2 look noticeably better than High settings on a lot of other games.. so Low settings must be viewed differently for Alan Wake 2). Games will be tested to disable settings that don’t visually affect the game, or take a big performance impact without improving visuals enough to warrant, in order to preserve performance as much as possible. This settings preset will be called the GFP Preset, and we hope to see it implemented in games as a selectable preset in the future.
Separate testing may be done for scenarios that have Ray-Tracing On or Off.

Due to the nature of this test’s structure, the results will scale proportionately to the hardware used, so these scores are still relevant and applicable if you own better or worse hardware.

Example: (i5 13400F + RTX 4060),(Ryzen 5 7600x + RX 7600)

(Or any performance equivalent with minimum 16gb ram, running from an NVME)

DLSS, FSR, XeSS, and Frame Generation are all turned off. This score is purely based on rasterization.

Formula:
Initial quality evaluation: (texture quality, + image clarity, + animation quality, + character model quality, + physics quality, + geometrical complexity, + shader quality, + reflection quality). Each of these are scored 0-100, and added together into one value. Detailed explanations on how this algorithm and these scores are decided can be found here.

Next, we take the Average Framerate (AFR) and 1% Lows (LFR): We multiply the sum of our quality evaluation by the AFR. Once we have that value, we divide by 8 to get a normalized score. Then, a penalty is calculated using an exponential slope. on the difference between the AFR and LFR, and that percentage is subtracted from our normalized value, giving the final score.
Example Benchmark:
”Example Game, GFP Settings: (88, 75, 97, 82, 93, 88, 93, 84. AFR: 86, LFR: 70)”
Final Score: 6,321.35. The performance is above standard for it’s visual quality.

So, why?

The answer is simple. While this method doesn’t account for everything, it gives a base point for what level of performance should be achievable given the game’s visuals. Should this sort of score be followed, a game like Starfield, which is less visually complex compared Red Dead Redemption 2, (while performing much worse), would have a score to answer to.