By Walter Jobs, Technology writer and editor · Published 9 October 2026
Short answer: the clearest single sign is a graphics card sitting well below 95% while the frame rate is lower than you want. Everything else on this list is a variation of that, seen from a different angle. What follows is seven observable symptoms, what each one proves, and the fault each is most often mistaken for, because several of these have an innocent twin that costs nothing to fix.
This page is about recognising the symptom. If you want the mechanism behind it, what causes a bottleneck in a gaming PC explains why the two halves of the pipeline fall out of step in the first place.
Sign 1: the graphics card is not busy
During a demanding scene, with no frame cap, your card should be at 95% or above. Below that it is waiting between frames.
Confirms: something upstream is not feeding it. That something is often but not always the processor.
Mistaken for: a faulty card. It is not; an idle card is a healthy card being underused. Work through the nine causes of low GPU usage before concluding it is the processor, because six of them are not.
Sign 2: one processor core is pinned while the rest idle
This is the most reliable sign on the list and the one most people never check, because they look at total processor usage instead. Switch your monitoring to per core logical processors and watch during play.
Confirms: a processor limit, almost definitively. Game engines run a main thread that cannot be split across cores, so when that thread saturates one core, nothing else on the chip can help. On a sixteen thread processor this reads as roughly 25% overall while you are completely out of headroom.
Mistaken for: a healthy machine. Total usage is the single most misleading number in this whole diagnosis.
Sign 3: raising graphics settings costs you nothing
Go from medium to high, or from 1080p to 1440p, and watch the frame rate. If it barely moves, the card had capacity spare all along.
Confirms: the card is not the limit, so the processor almost certainly is. This is the most useful test on the list because it requires no monitoring software at all, and it hands you something immediately: free visual quality you were not using.
Mistaken for: nothing, which is why it is such a good test. Few other faults produce this result.
Sign 4: the frame rate ignores resolution
A card limited machine loses frames steeply as resolution climbs, because every step up multiplies the pixels it has to fill. A processor limited one returns nearly the same number at 1080p and 1440p, because simulating the game world does not get harder when you add pixels.
Confirms: a processor ceiling you are sitting against at both resolutions.
Mistaken for: a monitor or scaling problem. It is neither. The full version of this behaviour is worked through in why a CPU bottleneck disappears at 1440p.
Sign 5: busy scenes collapse while empty ones are fine
Standing in a quiet corridor at 140 frames per second and dropping to 55 in a crowded town square, a large battle, or a base with hundreds of objects in it.
Confirms: a processor limit, because what changed between those two places is the number of things being simulated and submitted, not the number of pixels. The view got no harder to draw; it got harder to work out.
Mistaken for: poor optimisation in the game. Sometimes that is fair, but the same scene on a faster processor runs better, which tells you where the limit sits.
Sign 6: strategy and simulation games run worse than they look
A grand strategy title, a city builder deep into a save, a survival game with a large automated base, or an open world with dense traffic. Visually simple, and slower than games that look far more impressive.
Confirms: these genres are processor bound by design. Thousands of entities each need updating every tick, and that work cannot be handed to a graphics card.
Mistaken for: an aging machine in general. It is specific: the same machine may be perfectly adequate for visually heavy, simulation light shooters.
Sign 7: competitive titles will not reach your monitor’s refresh rate
You bought a 165 Hz or 240 Hz screen, the game is deliberately light on graphics, your card is barely awake, and you still cannot fill the panel.
Confirms: a processor ceiling, and the most expensive version of one, because high refresh play is where processors are asked for the most and where the shortfall is most visible. There is more on this in CPU bottlenecks in esports games.
Mistaken for: a monitor fault or a cable problem. Check that the refresh rate is actually set in Windows first, since a 165 Hz panel ships running at 60 Hz more often than anyone would like.
How many signs do you need?
| What you found | What it means |
|---|---|
| Sign 2 alone | Enough on its own. A pinned core with others idle is conclusive |
| Signs 1 and 3 together | Enough. Idle card plus free settings is the classic pair |
| Signs 4, 5, 6 or 7 alone | Strong indication, worth confirming with sign 2 |
| Sign 1 alone | Not enough. Six non processor causes produce it |
| None of them, but low frame rate | The card is your limit, which is the normal and correct state |
The last row deserves emphasis. Most gaming PCs are limited by their graphics card, and that is what a well balanced machine is supposed to look like. Finding none of these signs does not mean you failed the test; it means you have nothing to fix on the processor side and your money belongs elsewhere.
You found some. What now?
Before replacing anything, find out how large the gap is, because the answer changes what you should do. A processor that is 8% behind your card is not worth spending on. One that is 35% behind is costing you a third of what you already paid for.
Put your two parts into the bottleneck calculator. It reports which part is the limit at each resolution, the frame rate the pair should produce, and the specific upgrade worth making. If you would rather start from your card and see which processors suit it, best CPU for every graphics card names the cheapest one that stops being the limit and the one that still will not be after you replace the card.
Two things worth trying before you buy anything, because both are free and both regularly turn out to be the real answer: confirm your memory is running in dual channel rather than on a single stick, and raise your graphics settings, since sign 3 means the quality is there for nothing.
Frequently asked questions
What is the quickest test for a CPU bottleneck?
Raise your resolution one step and watch the frame rate. If it stays roughly the same, your card had headroom spare and the processor is setting the pace. It takes thirty seconds, needs no software, and is harder to misread than a usage percentage.
Is 100% CPU usage in games always a bottleneck?
No, and the reverse matters more. A chip at 100% across every core while the game runs well is simply a busy processor. A chip at 30% overall with one core saturated is the one limiting you. The total figure tells you far less than the per core picture, which is covered in 100% CPU usage while gaming.
Can a CPU bottleneck appear only in some games?
Yes, and that is the normal pattern rather than the exception. Simulation heavy and competitive titles lean hard on the processor while visually rich, simulation light games lean on the card. The same machine can be processor limited in one and card limited in another on the same afternoon.
How much of a bottleneck is acceptable?
Any figure where the frame rate you get still exceeds what your monitor can display is acceptable, whatever percentage a calculator attaches to it. A percentage with no frame rate next to it cannot tell you whether to act, which is the argument made in does a bottleneck percentage really matter.
Will more RAM fix a CPU bottleneck?
It can, when the real constraint was memory bandwidth rather than the processor itself. Moving from one stick to two matched sticks doubles the width feeding the chip and produces a genuine gain in processor limited games. Adding capacity beyond what you are using does nothing. Check whether you are on a single stick before assuming the chip is at fault.
Every sign on this page is something you can observe on your own machine with tools you already have. Where this article refers to the size of a gap between two parts, that figure comes from our own model, and how we calculate this sets out how it is produced and where it stops being reliable.