Skip to content

VersusCPU

CPU Bottleneck vs GPU Bottleneck: How to Tell Which Is Holding Your Games Back

CPU bottleneck vs GPU bottleneck: how to tell which limits your games using GPU usage, a resolution test and Intel's PresentMon GPU Busy metric, what to rule out first, and whether a new graphics card or CPU will actually help.

By , Senior Hardware Editor

4 min read

CPU Bottleneck vs GPU Bottleneck: How to Tell Which Is Holding Your Games Back

"Bottleneck" is one of the most misunderstood words in PC gaming. Every PC has one: in any given moment, either the processor (CPU) or the graphics card (GPU) is the part that limits how many frames per second you get. That's normal. The useful question is which one it is in the games you play, because that tells you whether a new graphics card, a new CPU or a settings change will help. This guide explains how to check with free tools.

What each bottleneck means

GPU-bound

CPU-bound

What's happening

The graphics card is working flat out; the CPU is waiting for it

The CPU can't prepare frames fast enough; the GPU is waiting

Typical signs

GPU usage near 100%

GPU usage well below 100% while frame rate stays low

Lowering resolution or graphics settings

Frame rate goes up

Frame rate barely changes

Where it's common

High resolutions and heavy graphics settings

Low resolutions and very high frame-rate targets

What helps

Lower settings, upscaling, a faster GPU

A faster CPU, faster memory, lowering CPU-heavy settings

Being GPU-bound is usually what you want in a gaming PC: it means you're getting full use of the most expensive part.

How to check: three simple tests

1. Watch GPU usage while you play

Use an on-screen overlay that shows GPU usage and frame rate, such as MSI Afterburner with RivaTuner; our guide to monitoring temperatures and FPS with HWiNFO and Afterburner shows how to set one up. Play normally for a few minutes. If GPU usage sits close to 100%, you're GPU-bound. If it sits well below that and your frame rate is lower than you'd like, the CPU (or something else, such as a frame cap) is likely the limit.

2. Change the resolution

Drop the game's resolution, or set an upscaler such as DLSS or FSR to a more aggressive mode, without changing anything else. If frame rate rises noticeably, you were GPU-bound. If it hardly moves, the CPU is the limit at that point. Our DLSS, FSR and frame generation explainer covers how upscaling changes the load.

3. Use a tool built for the job: Intel PresentMon

Intel's free PresentMon tool combines performance data and GPU telemetry in one overlay and capture utility. Intel highlights its "GPU Busy" metric, which shows in real time how much of each frame the graphics card is actually working, so you can see the balance between CPU and GPU. Intel says PresentMon works with other vendors' hardware as well as Intel's, supports DirectX 12, 11 and 9, OpenGL and Vulkan games, runs on Windows 10 and 11, and is open source. In practice, if the GPU is busy for nearly the whole frame you're GPU-bound; if it's busy for much less, the GPU is spending time waiting, usually on the CPU.

Nvidia's free FrameView app is another option for measuring frame rates and frame times, along with power and performance per watt, on a wide range of graphics cards.

Rule out the other limits first

Low GPU usage isn't always the CPU's fault. Check these before buying anything:

  • Frame caps and V-Sync. If you've capped your frame rate, GPU usage will drop once the cap is reached. That's expected. See how to cap FPS on Nvidia and AMD.

  • Memory running at default speed. DDR5 runs at its slower default unless you enable XMP or EXPO; see our XMP and EXPO guide.

  • Background apps and power settings. A busy background app or a power-saving mode can hold the CPU back.

  • Heat. A CPU or GPU that's overheating will slow itself down. Check temperatures with the monitoring guide above.

What to upgrade

  • GPU-bound and unhappy with performance? A faster graphics card is the upgrade that will help. Our best GPU for 1440p guide is a good starting point.

  • CPU-bound in the games you care about? A faster CPU helps, but on many platforms that means a new motherboard too. Our gaming PC upgrade guide covers the options. Before upgrading, try raising graphics settings or resolution: if the CPU is the limit, you can often turn up visual quality for free.

Our verdict

Every PC is either CPU-bound or GPU-bound at any moment; what matters is which one limits the games you play. If GPU usage is near 100% and lowering resolution raises your frame rate, the graphics card is the limit. If GPU usage is low and resolution changes do little, the CPU is. Check with an overlay or Intel's PresentMon and rule out frame caps, memory settings and heat before you spend money on an upgrade.

The bottom line

Every PC is either CPU-bound or GPU-bound at any moment; what matters is which one limits the games you play. If GPU usage is near 100% and lowering resolution raises your frame rate, the graphics card is the limit. If GPU usage is low and resolution changes do little, the CPU is. Check with an overlay or Intel's PresentMon and rule out frame caps, memory settings and heat before you spend money on an upgrade.

Sources

  1. 01Intel, PresentMon
  2. 02Nvidia, FrameView app

About the author

Callum Reid

Senior Hardware Editor

Callum leads GPU, CPU and platform architecture coverage at Forge PC. Before joining the editorial desk, he spent nine years specifying, assembling and stress-testing custom gaming and workstation systems in the UK, with a focus on PCIe bandwidth, VRM thermal design, DDR5 memory tuning and 1440p/4K frame-time consistency.

More from Callum Reid

Read next

LoRA vs QLoRA: Which Fine-Tuning Method Fits Your GPU?

Versus

LoRA vs QLoRA: Which Fine-Tuning Method Fits Your GPU?

LoRA vs QLoRA for local LLM fine-tuning: both train small adapters on a frozen model, but QLoRA loads the base model in 4-bit, which Unsloth says cuts VRAM use about four times for a marginal accuracy cost. QLoRA suits most home GPUs.

Callum Reid3 min

RAG vs Fine-Tuning for Local AI: Which Should You Use on Your PC?

Versus

RAG vs Fine-Tuning for Local AI: Which Should You Use on Your PC?

RAG vs fine-tuning for local AI: RAG feeds a model relevant chunks of your documents for each question, while fine-tuning changes the model's weights. For most home PCs, start with RAG; fine-tune for style, format or specialised tasks.

Callum Reid4 min