Bottleneck Calculator: Is Your CPU or Your GPU Holding You Back?
A bottleneck is not a property of two parts; it changes with every game, resolution and setting. In TechSpot's tests the same two CPUs were 0% apart in one setup and 43% in another. So instead of a calculator that guesses from part names, here is a two-minute check of your own PC, and a worked example of what the numbers mean.
CPU gap, 1080p medium, RTX 5090
37%
Same CPUs, 4K ultra, RTX 5080
0%
Largest gap, Space Marine 2
43%
Time for the check
2 min
Is your CPU or GPU the bottleneck? Answer from your own readings
Play the game you care about with an overlay on (steps below), in a busy scene, and answer three questions.
Not enough to tell yet
Run the resolution test to be sure.
Note your fps, then lower the resolution or render scale a lot (for example to 720p) without changing anything else. If fps jumps, the GPU was the limit. If it barely moves, the CPU is.
How to check for a CPU or GPU bottleneck
- Turn on an overlay. The Nvidia App overlay (Alt + Z), AMD's Adrenalin overlay, MSI Afterburner or Intel's free PresentMon can show fps and GPU usage; Afterburner can also show usage per CPU thread.
- Turn off frame caps for the test. V-Sync, frame limiters and your monitor's refresh cap all hold fps down and make the GPU look underused.
- Play a demanding scene for a minute and note fps and GPU usage. Menus and loading screens tell you nothing.
- Lower the resolution a lot (720p, or render scale 50%) and play the same scene. If fps jumps, the GPU was the limit; if it barely moves, the CPU was.
The precise version: frame time vs "GPU Busy"
PresentMon reports how long each frame takes and how long the GPU actually worked on it ("GPU Busy"). TechSpot's guide to the tool: "If the Frame time is much longer than the GPU Busy time, then the game's performance is being limited by factors such as the CPU's speed." When the two are almost equal, the GPU is the limit. In one of TechSpot's examples, a GPU was busy 1.42 ms of a 3.46 ms frame: clearly CPU- or engine-bound.
Worked example: when does a faster CPU actually help?
TechSpot paired a Ryzen 7 9800X3D and a Ryzen 5 7600X with four graphics cards and tested four games at three resolutions and two presets. Each cell shows how much faster the 9800X3D was. Darker means a bigger CPU effect; 0% means the graphics card was the limit for both.
TechSpot, "Ryzen 7 9800X3D vs. Ryzen 5 7600X: CPU and GPU Scaling Benchmark", average of Rainbow Six Siege, Kingdom Come: Deliverance II, Marvel Rivals and Space Marine 2. Published Jul 24, 2025.
What the example shows
- The CPU matters when the GPU is fast and the resolution is low. 1080p medium with an RTX 5090: 37% on average.
- With a mid-range card, the gap mostly disappears. With the RX 9060 XT, the 9800X3D was 5% faster at 1080p medium and 0% everywhere else.
- The game matters as much as the hardware. In CPU-heavy Space Marine 2, the 9800X3D stayed 43% ahead at 4K with an RTX 5090. In Kingdom Come: Deliverance II at 1440p ultra, the results were "entirely GPU limited", even with an RTX 5090.
- Drivers can matter too. In Rainbow Six Siege at 1080p medium, the 7600X with an RX 9070 was 12% faster than the same CPU with an RTX 5090, which TechSpot attributes to Nvidia's DirectX 12 driver overhead on slower CPUs.
Are bottleneck calculators accurate?
Online bottleneck calculators ask for a CPU and a GPU and return one percentage. The worked example shows why that cannot be right: the same pair of CPUs is 0% apart at 4K ultra and 37% apart at 1080p medium, and a single game can push the gap to 43%. A number without the game, resolution and settings says very little. Our reading
Your own measurements answer the question for the games you actually play, and they take two minutes.
What to do about a CPU or GPU bottleneck
If the GPU is the limit
- Turn on upscaling in Quality mode (DLSS, FSR or XeSS): fewer pixels per frame, more fps.
- Lower the settings that cost the GPU most: ray tracing, shadows, volumetrics.
- If you upgrade, upgrade the graphics card; check the power supply first.
If the CPU is the limit
- Raise the resolution or graphics settings: the GPU has room, so the picture gets better for free.
- Lower CPU-heavy settings: crowd density, view distance, physics and simulation.
- A frame cap a little below your average fps smooths out the dips.
- Frame generation does not fix a CPU limit; see our frame generation guide for why.
FAQ
How do I know if my CPU is bottlenecking my GPU?
Watch GPU usage while you play. If it stays well below 95% and your frame rate barely changes when you lower the resolution, the CPU is the limit. A tool like PresentMon shows it directly: frame time much longer than "GPU Busy" time means the CPU is holding the GPU back.
Are bottleneck calculators accurate?
No. They give one percentage for a CPU and GPU pair, but the real limit depends on the game, resolution and settings. In TechSpot's tests, the same Ryzen 7 9800X3D vs Ryzen 5 7600X gap ranged from 0% to 43% across games and settings.
Is a GPU at 100% usage bad?
No. A graphics card near 100% means the CPU keeps it fully fed, which is what you want in a gaming PC. It only becomes a problem if the frame rate is too low; then lower settings or upscale.
Does a CPU bottleneck matter at 4K?
Rarely. At 4K ultra, TechSpot measured only an 8% average gap between a Ryzen 7 9800X3D and a Ryzen 5 7600X with an RTX 5090, and none with an RTX 5080 or slower. CPU-heavy games like Space Marine 2 are the exception.
Will a CPU upgrade increase my fps?
Only if you are CPU-bound. At 1080p medium with an RTX 5090, TechSpot measured the 9800X3D 37% faster than the 7600X on average; with an RX 9070 the gap was 8%, and with an RX 9060 XT 5%.
How we researched this
The worked example uses TechSpot's published results; the method follows TechSpot's PresentMon guide. The checker turns your readings into a verdict with simple rules; it does not know your hardware.
First version.
Where the figures come from
- Ryzen 7 9800X3D vs. Ryzen 5 7600X: CPU and GPU Scaling Benchmark, TechSpot.
- How to Use Intel PresentMon for Benchmarking and Testing, TechSpot (frame time vs GPU Busy).
- GPU Benchmarks Hierarchy 2026, Tom's Hardware (fast cards become CPU-bound without a high-resolution monitor).