Average frame rate is the number on every box, every benchmark chart and every calculator, this one included. It is also a poor description of how a game feels, because an average hides its own worst moments by design.
The number that tracks how smooth a game feels is the 1% low: take every frame rendered in a benchmark run, sort them, and look at the slowest one percent. That figure is where stutter lives.
What the two numbers look like together
Consider two machines, both averaging 120 frames a second:
| Machine | Average FPS | 1% low | How it feels |
|---|---|---|---|
| A | 120 | 98 | Smooth, consistent, unremarkable in the best way |
| B | 120 | 47 | Visibly hitching every few seconds |
Same headline. Completely different experience. Machine B is the one people describe as “my PC feels laggy even though the FPS counter says it’s fine”, and they are not imagining it.
A rough guide: if your 1% low is above roughly two thirds of your average, the game will feel consistent. Below half, you will notice.
Why the processor usually owns this number
Average frame rate is mostly a graphics card story. Frame consistency is mostly a processor story, and the reason is in what each part does.
The graphics card’s work is fairly uniform. It shades roughly the same number of pixels every frame, so its frame times are stable.
The processor’s work is not uniform at all. Most frames are cheap, and then a handful are expensive: an area transition streams in new assets, a script fires, physics resolves an explosion, twenty AI agents all decide something at once. Those spikes land on one or two threads, and a spike of 25 milliseconds on a frame budget of 8 milliseconds is a visible stutter.
This is why a machine can be GPU limited on average and still stutter. The card sets your maximum frame rate; the processor sets the floor.
What this means for a bottleneck reading
The calculator models averages, because published benchmark data is dominated by averages. That is an honest limit and it is written up in the methodology. It means a small CPU bottleneck on paper can matter more than the percentage suggests:
- A 10% CPU bottleneck in a clean, evenly loaded game is genuinely minor.
- A 10% CPU bottleneck in a dense open world, a big strategy game or a heavily modded save is often the visible difference between smooth and hitchy, because the same processor that is 10% short on average is 40% short in the expensive moments.
Simulation and strategy titles are the extreme case. Late game city builders, grand strategy and large multiplayer maps are almost entirely processor bound, and they punish weak frame consistency more than any shooter does. This is one of the four things a bottleneck calculator cannot tell you.
Where cache comes in
This is the clearest reason AMD’s X3D processors outperform their clock speeds in games. A Ryzen 7 9800X3D or Ryzen 7 7800X3D carries a much larger cache, so more of the data a game needs on an expensive frame is already close to the processor instead of out in memory.
That barely moves a synthetic benchmark. It moves 1% lows a lot, which is why those parts feel quicker than their averages imply — and why they are worth their premium at 1080p, as the budget split guide covers.
Measuring it on your own machine
Any overlay that shows frame times will do. MSI Afterburner with RivaTuner gives you a frame time graph; CapFrameX or the built-in benchmark in games like Cyberpunk 2077 will report 1% lows directly.
Watch the frame time line rather than the FPS counter. A flat line is smooth. Regular spikes are the thing you are chasing.
The fixes, in order of how often they work
- Enable XMP or EXPO. Memory running at default speed hurts 1% lows more than it hurts averages. See does RAM speed cause bottlenecks.
- Get 32 GB, in two sticks. Running out of memory forces paging to disk, which is the ugliest stutter there is.
- Check VRAM. If the card is out of video memory, frames spike hard. Lower textures one step and see if the spikes stop — the full test is in running out of VRAM is not the same as a bottleneck.
- Cap your frame rate slightly below what the machine can sustain. A steady 100 feels better than an average of 120 that keeps dipping to 45.
- Then, and only then, consider a processor with more cache.
The takeaway
Average frame rate tells you what your PC can do on a good frame. The 1% low tells you what it does on a bad one, and bad frames are the ones you notice. Treat the calculator’s percentage as the first question and frame consistency as the second.
FAQs
What are 1% lows in gaming?
Take every frame rendered during a run, sort them by how long each one took, and look at the slowest one percent. Expressed as a frame rate, that is your 1% low. It describes your worst moments rather than your typical ones.
What is a good 1% low?
As a rule, anything above two thirds of your average frame rate feels consistent. Below half, you will see the hitching. A 120 FPS average with a 98 FPS 1% low is healthy; the same average with a 47 FPS low is not.
Why does my game stutter even though my FPS is high?
Because an average hides its own worst frames. A high average with poor 1% lows is exactly what stutter looks like on paper. Watch a frame time graph instead of the FPS counter and the spikes become obvious.
Are 1% lows a CPU or a GPU thing?
Usually the CPU. Graphics work is fairly consistent frame to frame, while processor work arrives in bursts — asset streaming, physics, AI, scripts. That is why a system can be GPU limited on average and still stutter.
Is the 1% low or the 0.1% low more useful?
The 1% low is the more dependable everyday measure. The 0.1% low captures very rare spikes, so it is noisier between runs, though it is useful when hunting down severe hitching.
Will more RAM improve my 1% lows?
It can, in two ways. Running memory at its rated XMP or EXPO speed improves frame consistency more than it improves averages, and having enough capacity avoids the paging to disk that causes the worst stutters.
Does a bottleneck calculator account for 1% lows?
No. Calculators model average performance, because published benchmark data is dominated by averages. A small CPU bottleneck on paper can still produce visible stutter in demanding titles.