Bottleneck Calculator — Free PC, CPU & GPU Bottleneck Checker (2026)

Two components decide most of your frame rate, and they only perform as well as the slower of the pair allows. Those two are the central processing unit (CPU) and the graphics processing unit (GPU). When your processor cannot prepare frames as fast as your graphics card can draw them, the graphics card waits.

This bottleneck calculator estimates which of the two runs out of headroom first in your PC, and how large the gap is. It accounts for resolution and workload, because both change the answer substantially — the same pair of parts can be well matched for one job and badly matched for another. The sections below explain how to read the result you get, what each range actually means, and which situations the calculation cannot see.

How to Use the Bottleneck Calculator

Fill the fields above in order. Each one changes the answer, so it is worth knowing why.

1. Processor. Search for your exact model. Suffixes matter, because a K-suffix part and its non-K counterpart reach different sustained clocks.
2. Graphics card. Pick the specific variant. An 8 GB and a 16 GB version of the same card share a name and behave differently once memory fills.
3. Resolution. The single most influential input. Higher resolutions move work onto the graphics card and reduce a processor limit without you changing a single part.
4. Workload. Gaming, streaming, video editing and general desktop use load the two components in different proportions. Streaming while playing adds processor work that gaming alone does not.
5. Memory speed and capacity. Advanced fields. Memory feeds the processor, so a slow or single-module configuration can hold back a fast chip regardless of how the two headline components compare.
6. Overclock offsets. Advanced fields. Raising a clock target lifts that component’s score, which is useful for testing whether tuning closes a gap or whether the gap needs new hardware.

The result appears immediately and updates as you change inputs. Change resolution first before changing anything else — it usually moves the number more than any other field.

How to Read Your Bottleneck Result

The percentage measures the headroom difference between your two components under the workload you selected. It is not a defect score, and it is not the share of frames you are losing. A result of 20% does not mean 20% fewer frames.

The result also names a component, and that name is the more useful half of the output. It tells you which part runs out of capacity first, and therefore which part responds to money or tuning. The component that is not named has spare capacity the named one cannot reach, which is why upgrading it changes nothing.

What each bottleneck range means and what it calls for.

Treat the boundaries as soft. A 19% result and a 21% result describe the same system, and the difference between the two rows is a suggestion about where your effort pays off rather than a threshold something crosses.

The same two components produce a high percentage at 1080p and a low one at 4K. Nothing about the hardware changed between those two results, and neither figure is more correct than the other. The number describes a combination working on a specific job, not a verdict on the parts you own.

That has a direct consequence for how you use it. A figure you would consider unacceptable at one resolution may be irrelevant at the resolution you actually play at, and a pairing described as mismatched in a general discussion may be well matched for your monitor. Check your own resolution and workload before acting on any percentage, including this one. No single threshold separates an acceptable result from a problem, which is why the table above is the answer rather than a signpost to one: as an illustration, 20% at 1080p in a simulation title points at a processor that cannot keep up, while the same 20% at 4K describes a pairing working close to its intended balance. Mild imbalance is the normal state of a working system, not a fault to correct.

What a PC Bottleneck Actually Is

A bottleneck is the point where one component finishes its work and waits for another, so the slower part sets the pace for the whole system.

The closest hardware comparison is a production line where one station takes longer than the rest. Stations either side sit idle, and the line’s output equals the slowest station’s output no matter how quickly the others could work. Adding capacity anywhere except that station changes nothing.

Every system has a slowest station. The question worth asking is not whether you have a bottleneck, but whether it sits somewhere that affects what you actually do.

A bottleneck is a condition, not a fault

Because something has to be slowest, a balanced PC is not one without a bottleneck — it is one whose bottleneck sits where nobody notices it. Nothing is being damaged and nothing is misconfigured. A component that is waiting does less work, draws less power and runs cooler than one held at full load. The percentage describes a gap between two parts, not strain on either of them.

Which bottlenecks you actually feel

The difference is whether the limit binds during something you do. A processor that cannot sustain 240 frames per second is irrelevant behind a 60 Hz display. A graphics card short of memory only matters at the texture settings and resolution you actually select. So the same percentage carries different weight for different people: the figure describes a hardware relationship, while whether it reaches you depends on your resolution, your refresh rate and the titles you play.

The limiting component moves with the workload

No part is permanently the bottleneck. One machine is processor-limited in a strategy title tracking thousands of units and graphics-limited in a linear shooter at 4K, with nothing changed in between. Resolution moves the limit, genre moves it, and settings move it. A single percentage is a snapshot of one workload rather than a fixed property of your build, which is why the resolution you check at matters as much as the parts you enter.

CPU Bottleneck vs GPU Bottleneck

The two limits produce different symptoms, and telling them apart costs nothing.

Signs of a CPU bottleneck

Processor utilisation sits high while the graphics card runs well below maximum. The clearest test is that frame rate barely moves when you lower graphics settings — you are removing work from a component that was not the constraint. Frame rate also fails to improve when you drop resolution, for the same reason.

Processor limits are worst at low resolution, because that is where graphics work per frame is smallest and processor work is proportionally largest. They are also worst in simulation, strategy, MMO and battle-royale titles, which track large numbers of entities, process server updates and submit many draw calls per frame.

Signs of a GPU bottleneck

The graphics card sits near maximum utilisation while the processor keeps headroom. Frame rate responds directly and predictably to graphics settings and resolution — lower the settings, gain frames.

This is the condition you want. A graphics-limited system is using its most expensive component fully, and it gives you a working control: settings and resolution become dials that trade image quality for frame rate. A processor-limited system does not respond to those dials, which is why an identical percentage feels worse when the processor is the component named.

Which is worse — a CPU or GPU bottleneck?

A graphics bottleneck is the better one to have, and the reason is control rather than severity. Lower the resolution, drop a few expensive effects or enable upscaling, and a graphics-limited frame rate moves. A processor limit ignores every one of those, because none of them reduce the work the processor does to prepare each frame. The levers that remain are the ones that cut simulation and draw-call load — entity counts, draw distance, crowd density — and there are fewer of them, they cost more visually, and plenty of titles do not expose them at all.

The second reason is how each limit feels. A graphics card short of headroom lowers frame rate evenly: every frame takes a little longer, and the result is a lower but steady number. A processor short of headroom tends to stall individual frames while the rest arrive normally, so the average can read acceptably while play feels broken. That is why a processor limit is reported as worse than its percentage suggests — the cost lands in frame-time consistency and 1% low frame rates rather than in the average.

This reverses the instinct most people bring to the result. Being told the graphics card is the limit means the system is behaving as intended and the dials are yours. Being told the processor is the limit means fewer dials and a more expensive way out.

Reading CPU and GPU Utilisation Together

Neither number means much alone. Read them as a pair.

How to read the four combinations of processor and graphics card utilisation. The low-and-low row catches people out most often, because it looks like a hardware problem and almost never is. A frame cap in the game, a driver-level limiter, or a variable refresh ceiling will hold both components below maximum while everything works exactly as configured.

One caveat matters more than the table itself. Utilisation figures are averages, and averages hide short spikes. A processor reporting 60% can still stall individual frames, because game engines load a few threads unevenly and one saturated thread barely moves an overall average. This is why a stuttery 90 frames per second and a smooth 90 frames per second can show identical average utilisation, and why an average alone cannot tell you whether a system feels good to play on.

Frame-time consistency is the honest measure. Averages tell you roughly where load sits; the spread between your typical frame and your slowest frames tells you what the system actually feels like. A pattern worth recognising is a processor sitting comfortably below maximum while frame times spike periodically — that is uneven threading rather than spare capacity, and no average will show it to you.

To capture these figures on your own machine, our measurement method sets out what to record, for how long, and the two tests that settle which component is the limit.

Neither component busy — the limit is elsewhere

Graphics-limited — the state you want for gaming

Both working, close to balanced

Processor-limited — the graphics card is waiting

Processor-limited, graphics card near saturation

Both saturated — no headroom for heavier scenes

Why Does Resolution Change Your Result?

Resolution decides where load falls, and it is the reason one pairing can be described as both balanced and mismatched.

The mechanism is simple. Raising resolution multiplies the pixels the graphics card shades every frame, so graphics work per frame climbs steeply. Processor work per frame stays roughly flat, because game logic, physics and draw call submission do not care how many pixels the result occupies. Push resolution up and the graphics card takes on more while the processor’s job barely changes, so a pairing that was processor-limited becomes graphics-limited.

As an illustration, consider a mid-range processor with a high-end graphics card. At 1080p the graphics card finishes each frame quickly and waits on the processor, so the processor is named and the percentage is high. At 1440p the gap narrows, because the card now has more than twice the pixels to shade while the processor’s workload is unchanged. At 4K the same card has four times the pixels of the 1080p case, the processor’s flat workload is no longer the constraint, and the same pair reports a low figure with the graphics card named instead.

The practical rule: if a result names your processor and you play above 1080p, check the figure at your actual resolution before spending anything.

Components Beyond CPU and GPU

Processor and graphics card set the ceiling in most systems, but they are not the only parts that can hold performance back. The rest tend to produce stutter and inconsistency rather than a lower average frame rate, which is exactly why they get misdiagnosed as a processor or graphics problem.

How each component beyond the processor and graphics card can become the limit. Thermal limits sit across all of these. When a component reaches its temperature target it lowers clocks to stay there, so a system that performs well for two minutes and worse after twenty is thermally limited rather than bottlenecked. The symptom is performance that declines with time under sustained load, which no hardware pairing calculation predicts, because it depends on your case airflow, your cooler and the temperature of the room.

What to Upgrade First

Upgrade the component the calculator names, at the resolution you actually play at. That single sentence resolves most cases.

Three conditions change the answer. If the percentage sits below roughly 10%, neither part is worth replacing yet, because settings and memory configuration will return more than new hardware will. If the named component is your processor while you play above 1440p, verify the result at your own resolution first, since the figure that named it may have come from a lower one. And if your symptom is stutter rather than a low average, look at memory configuration and storage before either headline component — those produce inconsistency, and replacing a processor will not fix a single-module memory configuration.

Resolution belongs in the decision before price does, because it decides whether a graphics card upgrade shows up at all. The same card that lifts 4K play can land on a 1080p system that was never waiting on graphics.

The error that wastes the most money follows from that: buying a faster graphics card while the processor is the limit. At 1080p that purchase returns almost nothing, because the new card waits exactly as the old one did — graphics work was never what held the frames back. It reads as a faulty upgrade when it was a working one aimed at the wrong part.

Weigh the cost of the path as well as the part. A graphics card is usually a straight swap; a processor often is not, because socket and memory generation decide whether it brings a motherboard and memory with it. That routinely makes the cheaper-looking component the more expensive upgrade, and where the two parts are close it is often what settles the choice.

Tuning before buying is almost always correct. It costs nothing, and it tells you whether the limit responds to being asked to do less — which is the same question an upgrade answers, for money.

Popular CPU and GPU Pairings

Some combinations get checked far more than others, usually because they sit at a price point where the balance is uncertain. Rather than trust a general verdict on a specific pair, run it through the calculator above at your own resolution and workload. The answer moves enough between 1080p and 4K that any single verdict would be wrong for roughly half the people reading it.

Will it bottleneck? is the question behind most of those searches, and on its own it has no answer. The same pair that is processor-limited at 1080p can be graphics-limited at 4K, so a pairing is only ever balanced or unbalanced for a given resolution and job. Supply those two and the question becomes answerable.

If you are still choosing parts, work in this order: fix your resolution first, then pick a graphics card suited to it, then find the cheapest processor that does not limit that card. Choosing in that sequence avoids the most common and most expensive mistake in PC building, which is buying a processor far beyond what your graphics card can ever keep busy.

Bottleneck Calculator by Use Case

The same hardware suits different jobs differently, because each job loads the two components in different proportions.

Gaming leans on the graphics card at higher resolutions and on the processor at lower ones. Streaming while playing adds processor work on top of the game, so a pairing that is balanced for gaming alone can become processor-limited the moment recording starts. Video editing and 3D rendering reverse the usual advice entirely: core count keeps helping well past the point where games stop caring, so a processor that is overkill for gaming can be exactly right for production work. Virtual reality is demanding in a different way again, since it renders two views at a high refresh rate and punishes frame-time inconsistency far more harshly than a flat-panel game does. AI and machine-learning work is different again: it leans on graphics memory capacity and raw compute throughput rather than on frame preparation, so a card with more memory frequently matters more than a faster processor.

Laptops deserve separate treatment, because a shared power and thermal budget means laptop parts rarely behave like their desktop namesakes.

How Accurate Is This Bottleneck Calculator?

Accurate enough to guide a purchase, not accurate enough to replace measurement. It compares hardware capability, and there are three things no hardware comparison can know.

Your software environment. Background processes, overlays, recording software and your memory configuration all change how much processor capacity a game actually receives.

Your thermals and power limits. Case airflow, cooler quality and the power limit set by your board or laptop manufacturer decide the clocks your parts sustain, as opposed to the clocks they can reach briefly.

Per-title driver overhead. Driver cost differs between games and between vendors, so two titles on identical hardware can name different limiting components.

It is worth separating a bottleneck checker from a bottleneck test, because they answer different questions. A checker like this one compares two parts on paper and predicts which should run out first, which is what you need before buying. A bottleneck test measures the machine you already own, in the games you actually play, and is the only thing that can confirm the prediction. The first is an estimate; the second is evidence.

No calculator replaces recording your own frame times in the games you actually play. Treat the percentage as a starting hypothesis, then confirm it on your machine before spending money. Our methodology sets out where the scores come from, how the resolution weighting works, and what the model deliberately leaves out.

Where to Go From Here

If your result named a component and the percentage sat above 20%, the next step is not a purchase. Load the game that bothers you most, record frame times for a few minutes of real play rather than a benchmark scene, and check whether the slow frames line up with what the calculator predicted. A processor limit shows up as periodic spikes; a graphics limit shows up as a consistently lower line that responds when you drop settings. Two settings changes will confirm which one you have before you record anything, and the two limits behave very differently once you know which is which.

If the two disagree, trust your own measurement. The calculator compares parts on paper, and your machine is the only one running your software, in your case, at your ambient temperature. If the result puzzles you rather than contradicting you, what bottlenecking is and is not covers the four readings people most often get wrong, and balanced pairings by budget covers what to buy once you know which part is the constraint.