Free calculatorLive resultsSteps and sourcesEmbed on your site, free

Bottleneck Calculator

Bottleneck Calculator works out Limiting component and CPU headroom from Measured CPU utilisation, Measured GPU utilisation and Utilisation counted as saturated. The default figures on this page give limiting component of GPU; change any entry and the result, the working and the sources update together.

Limiting component
GPU
Bottleneck Calculator
CPU headroom
40.0 %
GPU headroom
2.0 %
Gap between the two
38.0 percentage points
Best-case throughput multiplier if the limiter were removed
1.633
Best-case uplift
63.3 %
Both components saturated
no
What this means
The GPU is the limiter in this workload: it sits at 98% while the other sits at 60%.
Worked out from Theory of constraints: system throughput is set by the most utilised resource and 1 other source.
These numbers describe the workload you measured. A different game, resolution, setting or scene will move both utilisations and can change which component limits.
The uplift figure is a ceiling from the utilisation law, not a prediction. Frame time is not linear in utilisation, and a faster part often shifts the limit somewhere else entirely - memory, storage or a single game thread.
This is a comparison of utilisation figures YOU measured, not a prediction from component names. There is no published model that converts a CPU and GPU model pair into a bottleneck percentage, and this tool does not pretend to have one. The uplift figure assumes throughput scales linearly with utilisation, which is a ceiling rather than a forecast; the only way to know what a part swap gains is to test it.

Your figures

Updates as you type
Average over the run, from an overlay such as the one in your GPU driver or a system monitor. Use total CPU utilisation, or the busiest thread if the workload is single-threaded.
Average over the same run and the same window as the CPU figure.
Optional. Used only to express the best-case uplift in the same unit you measured.

How this is worked out

4 steps
  1. Measured: CPU 60%, GPU 98%.
  2. Headroom: CPU 40 points, GPU 2 points; the two differ by 38 points.
  3. The busier component, the GPU, is the limiter.
  4. Utilisation law ceiling: 98 / 60 = 1.633x, i.e. at most 63.3% more throughput if the limiter stopped limiting and everything else scaled perfectly.

Its 4 worked examples are re-run on every release, last on 15 September 2026. What changed

Sources and references

2 sources
Theory of constraints: system throughput is set by the most utilised resource
Goldratt, E.M. (1984), The Goal. The limiting resource is the one closest to saturation; relieving any other cannot increase throughput.
Utilisation law (Denning & Buzen, 1978, "The Operational Analysis of Queueing Network Models")
U = X S: utilisation is proportional to throughput for a fixed service demand. This is the assumption the best-case uplift figure rests on, and it is stated rather than hidden.

What the Bottleneck Calculator calculates

Compares the CPU and GPU utilisation you measured while running your own workload, names which component is the limiter, reports the headroom left on each, and gives the best-case uplift removing that limiter could produce.

How to use this Bottleneck Calculator

Enter Measured CPU utilisation, Measured GPU utilisation, Utilisation counted as saturated and Call it balanced when the two are within. Limiting component updates as you type, and “Use a worked example” loads checked figures you can adjust. Change one figure at a time to see what drives the result.

How the result is worked out

Open “How this is worked out” beside the result for the step-by-step calculation behind Limiting component, CPU headroom and GPU headroom.

The method follows Theory of constraints: system throughput is set by the most utilised resource and Utilisation law (Denning & Buzen, 1978, "The Operational Analysis of Queueing Network Models").

GPU at 98 percent, CPU at 60 percent - a GPU-limited system example

With Measured CPU utilisation: 60, Measured GPU utilisation: 98, the calculator returns Limiting component: GPU, GPU headroom: 2, CPU headroom: 40. This example is checked against an independent source in the test suite.

What you can compare

Change one input at a time to see how it affects Limiting component, CPU headroom and GPU headroom. The inputs you can vary are Measured CPU utilisation, Measured GPU utilisation, Utilisation counted as saturated, Call it balanced when the two are within, Measured frame rate or throughput and Workload.

Understanding Limiting component, CPU headroom and GPU headroom

This is a comparison of utilisation figures YOU measured, not a prediction from component names. There is no published model that converts a CPU and GPU model pair into a bottleneck percentage, and this tool does not pretend to have one. The uplift figure assumes throughput scales linearly with utilisation, which is a ceiling rather than a forecast; the only way to know what a part swap gains is to test it.

Results are general information, not financial, tax, legal, medical or engineering advice. Check any decision that matters with a qualified professional.

Embed this calculator on your website

Copy this code into any page. The calculator appears with the figures currently entered, sizes itself to fit, and a small credit line links back here. Free, no account, no ads inside it, no tracking of your readers. Set your own figures and fixed fields; Personal removes the credit line, sets your colour and shows you which sites use it.

Preview

See something wrong? Report a problem with this calculator

Frequently asked questions

Is there a bottleneck calculator?

Compares the CPU and GPU utilisation you measured while running your own workload, names which component is the limiter, reports the headroom left on each. On the figures this page opens with — measured CPU utilisation 60% and measured GPU utilisation 98% — the limiting component is GPU. Put your own figures in their place and the working updates beside the result.

How do you calculate a bottleneck?

Compares the CPU and GPU utilisation you measured while running your own workload, names which component is the limiter, reports the headroom left on each. Open “How this is worked out” beside the result to see each step for your own figures.

Can you trust bottleneck calculators?

It needs Measured CPU utilisation, Measured GPU utilisation, Utilisation counted as saturated, Call it balanced when the two are within, Measured frame rate or throughput and Workload. Nothing else is required, and the figures stay in your browser.

What information do I need for the Bottleneck Calculator?

You need Measured CPU utilisation, Measured GPU utilisation, Utilisation counted as saturated, Call it balanced when the two are within, Measured frame rate or throughput and Workload. The page opens with default figures that give limiting component of GPU, so you can see the scale before entering your own.

What does this Bottleneck Calculator show?

It shows Limiting component, CPU headroom, GPU headroom, Gap between the two, Best-case throughput multiplier if the limiter were removed, Best-case uplift, Best-case frame rate at that ceiling and Both components saturated, with the working and any warnings for your entries.

Can I check how the Bottleneck Calculator result was calculated?

Yes. The “How this is worked out” section shows the calculation steps, and the sources section identifies the references used for the method.

Related calculators

View all related calculators12 more