Bottleneck Calculator
Bottleneck calculator that works in frames per second instead of invented scores. See whether your CPU or GPU sets the ceiling, and how much sits idle.
Measuring it exactly, rather than estimating
Metric definitions from Intel PresentMon, a free frame-timing tool that works with any graphics card. When GPU Busy is close to the frame time, the graphics card is the pacing part; when the frame time is much longer than GPU Busy, the gap is processor-side.
| Metric | CSV column | What it measures |
|---|---|---|
| Frame Time | FrameTime | Total time between frames, measured on the CPU |
| CPU Busy | CPUBusy | Time the CPU spent generating the frame |
| GPU Busy | GPUBusy | Time the GPU spent working on the frame |
| GPU Wait | GPUWait | Time the GPU spent waiting during the frame |
How much work each resolution is
Pixel counts are the standard mode dimensions multiplied out. Graphics work scales roughly with pixel count, which is why the same machine can be graphics-limited at 4K and processor-limited at 1080p.
| Resolution | Dimensions | Pixels | vs 1080p |
|---|---|---|---|
| 720p | 1280 x 720 | 921,600 | 0.44× |
| 1080p | 1920 x 1080 | 2,073,600 | 1× |
| 1440p | 2560 x 1440 | 3,686,400 | 1.78× |
| Ultrawide | 3440 x 1440 | 4,953,600 | 2.39× |
| 4K | 3840 x 2160 | 8,294,400 | 4× |
Where the two numbers come from
- Planning a build: take the processor figure from a CPU review and the graphics figure from a GPU review, for the same game, the same preset and the resolution you will play at.
- Diagnosing a machine you own: run the same built-in benchmark twice, changing only the resolution, and use the diagnose mode above.
- Measuring frame timing directly: Intel PresentMon and NVIDIA FrameView are both free, both log per-frame data, and both work regardless of which brand of card you have.
The playable figure is a ceiling, not a prediction: a real system also loses frames to driver overhead and to moments when neither part is the constraint, so expect to land at or slightly below it. Results depend on the specific game, preset and scene the two figures came from.
Related Calculators
You might also find these calculators useful
How to Check if Your PC Has a Bottleneck
A PC bottleneck is one component setting a ceiling the other cannot get past — a fast graphics card waiting on a slow processor, or the other way round. This calculator answers that in frames per second, the one unit a CPU and a GPU are both measured in. Give it the frames each part can manage in the same game, or give it your own frame rate at two resolutions, and it tells you which part sets your ceiling and how much of the other one goes unused.
What Is a PC Bottleneck?
A bottleneck is a limit imposed by whichever component finishes last. Rendering a frame is a relay: the processor prepares the work and the graphics card draws it. Whichever of the two takes longer sets the frame time, and the other one waits. If the processor is slower, the graphics card sits partly idle and you have a CPU bottleneck. If the graphics card is slower, the processor waits and you have a GPU bottleneck. Neither is damage and neither is a fault — it is simply what happens whenever two parts of a pipeline are not equally fast. What matters is how much capability is going unused, and that is a number you can measure.
Bottleneck Formula
Playable fps = min(CPU fps, GPU fps) · Idle share = 1 − min ÷ maxHow to Use the Bottleneck Calculator
Decide which question you are asking: whether a pairing you are considering will bottleneck, or what is limiting a machine you already have
For a pairing, find the frames the processor can feed from a CPU review, which is measured at a low resolution so the graphics card is never the constraint
Find the frames the graphics card can draw from a GPU review, at the resolution you will actually play at, and make sure both figures are the same game and preset
For a machine you own, switch to the diagnose mode, run your game's built-in benchmark at your normal resolution and note the frame rate
Drop the resolution, leave every other setting alone, run the same benchmark again and enter both figures with the resolutions they came from
Read which part sets the ceiling, the frame rate that implies, and the share of the faster part that goes unused
For an exact answer instead of an estimate, use a free frame-timing tool and compare how long the graphics card was busy against the total frame time
Common Bottleneck Calculator Use Cases
New Build Planning
Before buying, check the frames each candidate part can manage in the game you care about. If one figure is far below the other, the money spent on the faster part buys nothing.
Deciding on a Graphics Card Upgrade
If your processor already caps you below what the new card could draw, the upgrade delivers a fraction of its benchmark numbers. Compare the two figures before you buy.
Low FPS Troubleshooting
Frame rate lower than reviews suggested? Run the same benchmark at two resolutions to find out which part is actually limiting you, instead of guessing.
Deciding Whether to Move Up a Resolution
Moving from 1080p to 1440p is 1.78 times the pixels and moving to 4K is four times. If you are already graphics-limited, that is where the frames go.
Judging a Second-Hand PC
Ask the seller to run a built-in benchmark at two resolutions. The two figures reveal a mismatched pairing that a specification list does not.
Streaming and Recording
Software encoding takes processor time away from the game, so a build that is graphics-limited while playing can become processor-limited while streaming. Measure with the stream running.
Checking Whether an Upgrade Worked
Take the two figures before and after. If the part you replaced was never the limit, both runs will land in the same place, which is the answer.
Why Use a Bottleneck Calculator?
One Unit, Not Two Scales
Frames per second is the only unit a processor and a graphics card are both measured in. Comparing a CPU benchmark score against a GPU benchmark score compares two different tests of two different kinds of work, and no published data connects them.
See Which Part Sets the Ceiling
The slower of the two figures is what you actually get. The calculator names it, states the frame rate it implies, and shows how much of the faster part is going unused.
Spend on the Right Component
Upgrading the part that was never the limit changes nothing. Knowing which side has headroom left tells you where money makes a difference and where it does not.
Diagnose a Machine You Already Own
The diagnose mode needs no component data at all. Run the same benchmark twice, changing only the resolution, and the difference between the two runs tells you which part was holding you back.
Understand Why Resolution Changes the Answer
Graphics work scales with pixel count and processor work largely does not, so the same pair can be graphics-limited at 4K and processor-limited at 1080p. The resolution table shows exactly how much more work each step up is.
Every Step Is Shown
The full derivation is on the page: both inputs, the minimum, the difference and the share. There is no weighting, no penalty and no fitted constant to take on trust.
Learn to Measure It Directly
For the exact answer rather than an estimate, free frame-timing tools report how long the graphics card was actually busy each frame. The page explains which metrics to read and what the gap between them means.
Frequently Asked Questions
A bottleneck is one component setting a limit the rest of the system cannot get past. Drawing a frame needs the processor to prepare the work and the graphics card to render it. Whichever takes longer decides the frame time, and the other waits. If the processor is the slower of the two you have a CPU bottleneck; if the graphics card is, you have a GPU bottleneck.
Because doing so honestly would need one score for processors and one for graphics cards on a single comparable scale, and no such scale exists. PassMark's CPU Mark averages eight processor tests and its 3D Mark averages four graphics tests plus GPU compute — different tests of different work. Subtracting one from the other gives a number that looks precise and is not measuring anything. Frames per second is the unit both parts share, so that is what this tool uses.
From a processor review. Reviewers test processors at a low resolution with a very fast graphics card precisely so the card is never the limit, which makes the resulting frame rate a clean measure of what the processor can feed. Use the figure for the game and preset you care about.
From a graphics card review, at the resolution you actually play at. Graphics performance changes a great deal between 1080p and 4K, so a figure taken at the wrong resolution will give you the wrong answer. Reviewers pair the card with a top processor so the card is the constraint.
Run your game's benchmark at your normal resolution, then drop the resolution and run it again with nothing else changed. Lowering resolution removes graphics work. If the frame rate barely moves, the graphics card was never your limit and the processor is. If the frame rate rises noticeably, the graphics card was the limit at the higher resolution.
No. Some difference between the two parts is normal, and a small one is inside the variation you get between two runs of the same benchmark. What matters is how many frames you are actually giving up: ten percent of a 40 fps ceiling is four frames, while ten percent of a 200 fps ceiling is twenty. Read the fps figure, not just the share.
There is no threshold that applies to everyone, which is why this page shows the frame rate alongside the share rather than grading you. A useful test is whether the frames you are losing would change your experience. Below the refresh rate of your monitor they usually do; above it they usually do not.
Reduce processor work or increase graphics work. Close background applications, turn down settings that are processor-heavy such as draw distance, crowd density and physics, and make sure memory is running at its rated speed. Raising the resolution shifts load to the graphics card, which raises fidelity at the same frame rate. If none of that is enough, the processor is the part to replace.
Reduce graphics work. Lower the resolution or the settings that scale with pixels — shadows, anti-aliasing, ray tracing, texture quality — or enable an upscaler such as DLSS or FSR, which renders at a lower internal resolution. If you want the fidelity and the frames, the graphics card is the part to replace.
No. One part waiting on another is ordinary operation, not stress. Nothing runs hotter or wears faster because of it. A bottleneck costs you performance you paid for, which is a value problem, not a safety one.
Whichever one is setting your ceiling. If your processor caps you below what your card could draw, a faster card changes almost nothing, and the reverse is equally true. Measure first: the diagnose mode on this page needs nothing more than two runs of the same benchmark.
Most are not, because most invent a comparable score for two components that have never been measured on the same scale, and none of them publish how the number is produced. A frame rate comparison is different in kind: both figures are measurements in the same unit, and the arithmetic on top is a minimum and a ratio you can check by hand. It is still specific to the game and settings the figures came from.
Processor limits show up at 1080p and graphics limits at 4K. Graphics work scales with pixel count, so 4K asks four times as much of the card as 1080p while asking the processor for much the same. That is why reviewers test processors at low resolutions and graphics cards at high ones.
No, it makes it more visible. Lowering resolution removes graphics work, so if the processor was already the limit you gain nothing and the card idles further. That is exactly why it works as a diagnostic. To get more out of a processor-limited system, reduce processor work or raise fidelity rather than frame rate.
Yes, and it shows up on the processor side. Too little memory forces the system to page to storage, and memory running below its rated speed or in a single channel slows the processor's frame preparation. Both appear in this test as a processor limit, so check that memory is at its rated speed before concluding you need a new processor.
GPU Busy is how long the graphics card actually spent working on a frame, reported per frame by free tools such as Intel PresentMon. Compared against the total frame time it settles the question directly: when the two are close, the card is pacing you; when the frame time is much longer, the extra is processor-side. It is a measurement rather than an estimate, which is why this page points at it.