How Our Bottleneck Calculator Works

Component Performance Scoring

We start with normalized CPU and GPU performance fields compiled from component specifications and public benchmark context. These scores are comparison inputs, not measurements made on your PC.

For CPUs, we measure:

• Single-thread performance (critical for game physics and AI)
• Multi-thread performance (important for open-world games and multitasking)
• Cache size and speed (affects how fast data reaches cores)
• Memory controller efficiency (impacts RAM speed utilization)
• Instruction per clock (IPC) improvements across generations

For GPUs, we measure:

• CUDA/Stream processor count and clock speeds
• VRAM capacity and bandwidth (crucial for high-res textures)
• Ray tracing performance (RT cores/Ray Accelerators)
• Tensor core performance for DLSS/FSR upscaling
• Architecture efficiency (RDNA 3, Ampere, Ada Lovelace)

Workload-Specific Analysis

Different games stress hardware differently. A strategy game like Civilization VI hammers the CPU with AI calculations, while Cyberpunk 2077 pushes the GPU with complex lighting and textures.

The selected game category applies a documented CPU/GPU workload weighting. Individual games and patches can behave differently:

Resolution Impact Calculation

Resolution changes the modeled graphics workload. The same pairing can appear more CPU-limited at 1080p and more GPU-limited at 4K.

Here's why: At 1080p, your GPU renders 2.1 million pixels. At 4K, that jumps to 8.3 million pixels—nearly 4x more work for the GPU. Meanwhile, the CPU still handles the same game logic regardless of resolution.

This is why a Ryzen 5 5600 might bottleneck an RTX 4090 at 1080p but pair perfectly at 4K—the higher resolution shifts the workload to the GPU.

Advanced Feature Adjustments

Modern rendering features change GPU pressure, so the model applies explicit ray tracing and upscaling modifiers when selected.

Ray Tracing Impact:

• Reduces FPS by 20-40% depending on GPU generation
• Nvidia RTX cards handle RT better than AMD/Intel equivalents
• We adjust GPU scores based on RT core count and efficiency

DLSS/FSR Upscaling:

• Reduces GPU load by rendering at lower resolution then upscaling
• The modeled adjustment depends on the GPU's supported upscaling generation
• Actual image quality and gain depend on the game implementation
• The selected modifier is shown in each result's calculation-factor table
• Users should verify the feature in the target game

Conditions the model does not measure:

• Thermal throttling and case airflow
• Background tasks, driver versions, and game patches
• Per-system overclocks, power limits, and silicon variation

Bottleneck Percentage Calculation

The model compares the adjusted CPU and GPU scores to identify the practical limiter for the selected scenario.

The adjusted bottleneck percentage is a relative mismatch indicator inside this model. It is not a literal promise that replacing one part will increase FPS by the same percentage.

FPS Prediction Model

The FPS estimate converts normalized component scores through workload, resolution, memory, ray tracing, and upscaling modifiers.

Our prediction process:

• The model starts from normalized CPU and GPU performance scores
• Workload weighting changes the relative CPU and GPU pressure
• Resolution changes the graphics-side score
• RAM, ray tracing, and upscaling apply visible modifiers
• Each result exposes its assumptions and scenario tables

For example, the same CPU and GPU can receive different FPS and limiter estimates at 1080p and 4K because the modeled GPU workload changes.

Utilization Percentage Display

The utilization gauges are modeled indicators of relative component pressure in the selected scenario; they are not readings from your computer.

What the numbers mean:

• 98-100% utilization: Component is maxed out (likely the bottleneck)
• 85-97% utilization: Component is working hard but has slight headroom
• 70-84% utilization: Component is moderately utilized
• Below 70%: Component has significant unused capacity

In a balanced system, both CPU and GPU should sit in the 85-100% range. If one component is at 100% while the other is below 80%, you have a clear bottleneck.

Recommendation Engine

The result suggests practical next checks and upgrade priorities based on the modeled limiting component.

Our recommendation algorithm considers:

• Current bottleneck severity (mild, moderate, severe)
• Your target resolution and refresh rate
• Whether settings or upscaling can reduce the limit
• Whether CPU or GPU replacement is the more direct path
• Power, cooling, memory, and physical-clearance checks

Recommendations are decision support. Confirm motherboard, memory, power-supply, cooling, and case compatibility before buying hardware.