What GPU bottlenecking actually means

Why being limited by your graphics card is the desirable state, which settings to lower when it is not, and how a memory limit differs from a speed limit.

By
Published
3 minute read

Being limited by your graphics card while gaming is the normal and desirable state. The term gets used as though it were a fault to be corrected.

Why this is what you want

If the graphics card is at 100 percent utilisation, every component is being used efficiently and you are getting the performance you paid for. Nothing is being wasted.

The alternative, a card sitting at 60 percent while the processor struggles, means you have paid for card capability you cannot use.

So a graphics card limit is a sign of a well-balanced system, not a problem to solve.

When it is worth acting on

Only when the frame rate you get is below the frame rate you want. If you are at 100 frames per second on a 60 Hz display, you have frames to spare and nothing to fix.

When it does matter, the options in order of cost are: lower the settings that cost most for least visual return, which is usually shadows, reflections and ambient occlusion; enable upscaling, which renders at a lower resolution and reconstructs; or buy a faster card.

Video memory limits look different

Worth distinguishing from a straightforward card limit. When a card runs out of video memory, the symptom is not a lower steady frame rate but stutter: sudden pauses as assets are fetched across the PCIe bus.

If the frame rate is acceptable but the game hitches periodically, particularly after changing area or raising texture settings, memory capacity rather than card speed is the likely cause. Lowering texture quality specifically, rather than overall settings, is the test.

This distinction matters commercially. A card with a fast processor and 8 GB of memory can be a worse buy than a slower card with 12 GB for the same money, if the games you play exceed 8 GB. The benchmark average frame rate will favour the faster card while the actual experience favours the other one.

Which settings to lower first

Not all graphics settings cost the same for what they give back. When you need frames, these are usually the ones to reduce first.

Shadow quality is frequently the single most expensive setting relative to how much you notice it in motion. Dropping it one or two steps often recovers a meaningful amount.

Ray traced effects cost the most of anything on the list. Turning reflections or global illumination down, rather than off, is often a reasonable compromise.

Ambient occlusion and volumetric effects are moderately expensive and subtle in motion.

Texture quality is the exception and should usually be left high. It costs video memory rather than processing, so if you have the capacity it is close to free, and it is the setting you notice most.

Upscaling changes the arithmetic

Upscaling renders the game at a lower resolution and reconstructs a higher-resolution image. At the higher quality presets it looks close to native, and it reduces the card workload substantially.

This is the most effective single change available when a card is the limit, and it is usually better than lowering several settings. Try it before concluding you need new hardware.

Frame generation is different and worth understanding separately. It inserts synthesised frames between rendered ones, which raises the frame rate figure without reducing input latency the way real frames do. It helps a smooth-but-not-fast experience feel smoother, and does not help a game that already feels sluggish to respond.

Spotted something wrong on this page? Tell us and we will check it.

How to choose a graphics card

Start from your display, understand what video memory actually does, and make the three physical checks that stop a card fitting.

3 minute read

Integrated versus dedicated graphics

What integrated graphics genuinely handle, why memory configuration matters so much more for them, and two good reasons to want them even with a card fitted.

2 minute read