Performance Rating Methodology

Understanding how we calculate and adjust GPU performance scores across different resolutions.

Overview

For each benchmark, every graphics card receives a score from 0 to 100, calculated as a percentage relative to the top performer in that specific benchmark. Our methodology combines pure raster performance data with practical considerations like VRAM capacity and modern GPU features.

Raw Performance Calculations

The raw performance rating represents pure raster performance without upscaling or ray tracing, based on data from eight different 3DMark benchmarks and TechPowerUp performance rankings. Each resolution category uses a different weighting formula:

Overall Performance

(TPU × 2) + Fire Strike Ultra + Wild Life Extreme + Night Raid + Firestrike + Steel Nomad (DX12) + Steel Nomad Light (DX12) + Time Spy + Time Spy Extreme

÷ 10

1080p Performance

TPU + Night Raid + (Fire Strike × 2.5) + (Time Spy ÷ 2)

÷ 5

1440p Performance

TPU + (Steel Nomad Light × 1.5) + (Time Spy × 2) + (Port Royal ÷ 2)

÷ 5

4K Performance

(TPU × 2) + Fire Strike Ultra + Wild Life Extreme + Steel Nomad + Time Spy Extreme

÷ 6

VRAM Adjustments

The raw performance score is first adjusted to account for VRAM limitations, as cards with insufficient VRAM may experience performance degradation in modern games. This adjustment penalizes cards with less than 16GB of VRAM:

Overall

(Score × 5) + ((Score × min(VRAM, 13)) ÷ 13) ÷ 6

1080p

(Score × 6) + ((Score × min(VRAM, 12.5)) ÷ 12.5) ÷ 7

1440p

(Score × 4) + ((Score × min(VRAM, 13)) ÷ 13) ÷ 5

4K

(Score × 4) + ((Score × min(VRAM, 13.5)) ÷ 13.5) ÷ 5

Features & Functionality Multipliers

After VRAM adjustment, scores are further modified based on modern GPU features and capabilities. These multipliers vary by resolution to reflect real-world usage patterns:

Upscaling Technology

Weight: 10

Blackwell DLSS SR (best) to FSR 2 SR (baseline)

1080p: Not considered 1440p: Half weight

Frame Generation

Weight: 4

DLSS 4 MFG (best) to no frame generation (baseline)

I/O Bandwidth

Weight: 3

DP 2.1 UHBR20 (80 Gbps) to DP 1.2 HBR2 (21.6 Gbps)

1080p: Half weight

Hardware Ray Tracing

Weight: 8

Nvidia Blackwell (best) to no hardware RT (baseline)

1440p: Not considered

Driver Support

Weight: 4

Full support (best) to legacy/extended support (reduced)

Resolution-Specific Adjustments

1080p Considerations

Upscaling technologies are not factored in, as native 1080p performance is typically sufficient. I/O bandwidth receives half weight since high refresh rate limitations are less critical at this resolution.

1440p Considerations

Upscaling receives half weight as it becomes more relevant but not critical. Ray tracing is not separately weighted since Port Royal benchmark data already incorporates RT performance.

4K Considerations

All feature multipliers are applied at full weight, reflecting the demanding nature of 4K gaming where every advantage becomes crucial for playable performance.

Methodology Notes

The weighting assigned to each factor will always be somewhat subjective and dependent on individual use cases. Our approach aims to be as methodical and factually grounded as possible, balancing raw performance with practical considerations that affect real-world gaming experience.

Benchmark data is sourced from 3DMark test suites and TechPowerUp's comprehensive GPU database, ensuring consistent and reliable performance measurements across all tested graphics cards.