The quickest honest test takes five minutes and needs nothing you do not already have. Open a game you play, turn on an overlay that shows CPU and GPU usage, and look at the graphics card figure while you are actually playing — not in a menu, not in a cutscene. If your GPU sits at 95–100%, nothing is bottlenecking it. If it hovers around 60–80% while your frame rate is lower than you would like, something else is setting the pace, and the processor is the most likely candidate.
That is the whole diagnostic. Everything below is about reading it correctly, because the simple version has edge cases that matter.
Why GPU usage is the number that tells you
A frame has to be built before it can be drawn. The processor works out where everything is — physics, AI, what the game logic decided this tick — and hands a list of draw calls to the graphics card, which turns that list into pixels.
If the card finishes drawing before the processor has the next frame ready, it waits. Waiting shows up as utilisation below 100%, because utilisation is simply the proportion of time the card had work to do. A graphics card at 65% spent roughly a third of every second idle, holding out its hand for instructions that had not arrived yet.
This is why GPU usage is a better signal than CPU usage, and the reason trips a lot of people up: a processor can be the bottleneck while showing 40% total CPU usage. Games lean hard on a small number of threads. If one core is pinned at 100% and the other eleven are nearly idle, Task Manager averages that into a reassuring low number while the game is completely limited by the one busy core. Total CPU usage hides exactly the thing you are looking for. GPU usage does not.
Reading the overlay properly
Any of these will show you both figures. None of them cost anything.
- Steam — in-game overlay, performance monitor set to Detailed.
- Xbox Game Bar on Windows — Windows+G, Performance widget.
- The driver overlay — NVIDIA app or AMD Adrenalin, both have a performance tab.
- MSI Afterburner with RivaTuner — the most detailed, and the one worth learning if you intend to do this more than once, because it will show per-core usage.
Four things to get right when you look:
Play, do not pause. Menus and pause screens often run uncapped or at a fixed low cost, and they tell you nothing about the game.
Remove any frame cap first. This is the mistake that invalidates most self-diagnoses. If you have V-Sync on, or a frame limiter, or your monitor is 60 Hz and the game is hitting 60, your GPU will show low usage and nothing is wrong — the system is deliberately idling because you asked it to. A frame cap looks exactly like a CPU bottleneck on an overlay. Turn it off before you measure.
Watch for a minute, not a second. Usage swings. An open-world game dropping into a crowded town will shift the balance toward the processor for a few seconds and back out again. One glance can catch either extreme.
Check more than one game. This is the step most people skip, and it is the one that produces a real answer rather than a trivia fact. A simulation-heavy game and a corridor shooter will give you genuinely different results on identical hardware, and the difference is the useful information.
The test that actually proves it
The overlay tells you which component is working. This tells you whether it matters.
Drop your resolution to the lowest the game offers and watch the frame rate.
Lowering resolution removes work from the graphics card and almost none from the processor — the card draws fewer pixels, but the game still has to simulate the same world, run the same AI, issue the same draw calls. So:
- Frame rate rises a lot — the graphics card was the limit. It was working flat out and you just gave it less to do. Your processor has headroom.
- Frame rate barely moves — the processor was the limit all along. You removed the card’s workload and the frame rate did not care, because the card was never the thing holding you back.
That ceiling you just found, the number the frame rate refuses to climb past, is roughly your processor’s limit in that game. It is the most useful figure you can measure at home, and it is far more informative than any percentage, because it is in the same units as the thing you care about.
Now run it the other way. Raise the resolution and the quality preset. Frame rate should fall, and as it falls the GPU usage should climb toward 100%. If a CPU-limited build moves to 1440p or 4K, it often stops being CPU-limited entirely — not because the processor got faster, but because the card now has enough work to stay busy.
What the answer means for your money
This is where the diagnosis turns into a decision, and the two directions are not symmetrical.
Graphics card is the limit (usage near 100%). This is the comfortable position. One component is working flat out and the fix, if you want one, is a single part. Your processor will still be useful after a graphics upgrade.
Processor is the limit (usage well below 100%, resolution test flat). Less comfortable, because the fix is rarely just the chip. A new processor usually means a new motherboard, and new memory with it. Before pricing that out, three things are worth trying, and they are all free:
- Play at a higher resolution. It sounds backwards, but if you are CPU-limited at 1080p with a card that has headroom, moving to 1440p costs less frame rate than you expect and uses hardware you already own.
- Check memory is running at its rated speed. XMP or EXPO is off by default on many boards, which can leave memory running at 2133 MT/s instead of 6000. That is a real and recoverable loss in CPU-limited scenes, and it costs one setting in the BIOS.
- Turn down the settings that cost the processor rather than the card. Crowd density, draw distance, shadow distance and physics detail are processor work. Texture quality and anti-aliasing are not. Most “low preset” advice turns down the wrong half.
Neither, and the frame rate is fine. Then there is nothing to fix. Which brings us to the number everyone starts with.
About that percentage
Most bottleneck calculators, this site’s included, will give you a percentage. It is a reasonable summary and it is also the least useful output of the exercise, for two reasons.
It has no agreed definition. Different calculators mean different things by “23% bottleneck” and none of them publish the arithmetic, so the figure is not comparable between sites and cannot be checked.
And it is not actionable. Knowing you have a 23% bottleneck tells you nothing about whether to spend money, what to spend it on, or whether you would notice the difference. A 20% imbalance on a build running at 160 fps is invisible. The same 20% on a build running at 38 fps is the difference between playable and not.
The questions worth answering are: what frame rate am I getting, is it enough for my monitor, and what is the cheapest change that moves it? A percentage answers none of those.
What this site can and cannot tell you
Our bottleneck calculator models both ceilings — what your processor can sustain and what your card can render — from published specifications, and shows you where they cross at each resolution. That is useful for a question you cannot test at home, such as what a part you do not own yet would do in your system.
What it is not is a measurement. The figures are derived from specifications using a published model, not from benchmark runs, so treat them as well-reasoned estimates with a margin, not as results. Our methodology page sets out exactly how they are calculated and where the model is weakest.
For the machine in front of you, the overlay and the resolution test beat any calculator, including ours. They are measuring your actual hardware with your actual games, and nothing modelled can compete with that.