An integrated system has no dedicated graphics card, which changes the question. There is no second component to compare against, so asking which part bottlenecks the other stops being useful, and an ordinary bottleneck calculator has nothing to compare. This integrated graphics bottleneck calculator asks a different question instead: which resolutions and settings are actually viable without a GPU — and what your panel can show sets the target you are aiming at. The answer usually depends on your memory configuration rather than on the graphics block itself.
Why Are Integrated Graphics Memory-Bound?
Integrated graphics have no dedicated memory, so they share system memory bandwidth with the processor.
Dedicated memory against shared memory
- Who shares the bus?
- Nobody — the card carries its own
- What limits graphics?
- The graphics chip itself
Two separate memory systems
Illustration of the mechanism, not measured data.
A dedicated card carries its own memory, physically attached and reserved entirely for graphics work. Integrated graphics have nothing of the sort. Every texture read, every framebuffer write, every geometry fetch travels over the same memory bus the processor is using for game logic at the same moment. The two compete for a single shared resource.
Three consequences follow, and all of them are actionable.
Memory speed affects graphics performance directly. On a system with a dedicated card, faster system memory helps the processor and barely touches the graphics side. Here it raises the graphics ceiling itself, because the graphics block is frequently waiting on memory rather than on its own capability. This is the one configuration where buying a faster memory kit reliably buys frames.
Single-channel memory is close to catastrophic. One module halves the available bandwidth, and it halves a resource two components were already dividing between them. The graphics side takes the worse end of that, being the more bandwidth-hungry of the two. This is the single most common reason a budget machine underperforms what its specification sheet suggests, and it is entirely free to fix.
Capacity counts twice. A portion of system memory is reserved for graphics and stops being available to anything else. So 8 GB total is not 8 GB for your game, and the shortfall arrives on top of a graphics limit rather than instead of one.
Put together, these make memory configuration the highest-value change on an integrated system — ahead of settings, ahead of drivers, ahead of almost anything else you could do. The RAM bottleneck calculator covers what speed is worth targeting on your platform.
The Bottleneck Question Inverts
On a system with two graphics components, asking which one limits the other is sensible. On an integrated system it is close to meaningless, because the graphics side is almost always the limit. You already know the answer before you run any calculation.
So the tool above reports something else. Instead of naming a limiting component and a percentage, it tells you whether a specific target — a resolution and a settings tier, for a type of game — is viable on your configuration. That is the decision actually in front of you.
This reframing has a practical benefit. A percentage would invite you to try to fix the imbalance, and there is no fixing it: you cannot add graphics capability to an integrated system without adding a card. A viability verdict invites you to choose a target you can hit, which is a decision you can act on today.
The one genuine exception is the memory constraint described above. When bandwidth is capping the graphics block, that is fixable, and the tool says so explicitly rather than folding it into a single number.
What Can Integrated Graphics Actually Run?
Esports titles and older releases run well at 1080p. Current demanding releases do not.
Esports titles and older releases run well. Competitive shooters are built to run on modest hardware because their audience demands high frame rates, and games from previous console generations are comfortably within reach at 1080p with reduced settings.
Mid-weight modern titles are usually playable at 1080p low, sometimes needing 720p or upscaling depending on the engine. This is the band where your memory configuration decides the outcome, and where the difference between single and dual channel is the difference between playable and not.
Current demanding releases are the honest limit. Some will run at 720p with everything reduced and upscaling enabled. Others will not be worth playing at any setting, and no configuration change fixes that.
| Title type | Realistic target | What decides the outcome |
|---|---|---|
| Esports titles and older releases | 1080p with reduced settings | Comfortably within reach |
| Mid-weight modern titles | 1080p low, sometimes 720p or upscaling | Your memory configuration |
| Current demanding releases | 720p reduced with upscaling, or not worth playing | The engine, and no setting fixes it |
What integrated graphics can realistically run, matching the three bands described above.
Two things have moved this picture. Recent processors with wider memory interfaces and larger graphics blocks — the designs also used in handheld gaming devices — perform substantially better than desktop integrated graphics of a few years ago. And upscaling changes the arithmetic meaningfully, because rendering internally at a lower resolution reduces both the bandwidth demand and the shading work at once. Enabling it is usually a better first move than dropping your display resolution outright.
Reducing an Integrated Graphics Limit
Cheapest first, and the order matters more here than on any other page:
Why capacity counts twice on an integrated system
Capacity and channel count both matter here, because the graphics block reads from the same memory.
- Populate two memory channels. This is the single largest gain available to integrated graphics and it is usually free. The graphics portion reads from the same memory as the processor, so one module leaves it working with half the bandwidth. Nothing else on this list comes close.
- Enable the memory profile. With two channels in place, running the modules at their rated speed rather than the conservative firmware default raises that same bandwidth ceiling again.
- Drop resolution before you drop settings. Integrated graphics are bandwidth-bound, and resolution moves bandwidth demand further than any individual quality setting will.
- Use upscaling wherever a title offers it. Rendering internally lower and reconstructing pulls the same lever as resolution at less visible cost.
- Then add a dedicated card. That is the subject of the next section.
Which component to spend on first, when several are candidates, is covered by the general upgrade guidance on the homepage.
When Should You Add a Dedicated Card?
Add one when the targets you want are consistently outside what the tool above calls viable, when you have already fixed the memory configuration and still fall short, or when you want settings above low at 1080p in current titles.
Two practical constraints catch people out in small prebuilt systems. Physical clearance is the first — many compact cases and slim desktops cannot fit a full-size card, and the low-profile options are limited. Power delivery is the second: prebuilt machines frequently ship with a supply sized precisely for a system without a graphics card, sometimes lacking the connectors entirely. Check both before buying, since a card that does not fit or cannot be powered is an expensive lesson. The PSU bottleneck calculator will tell you whether your supply has the headroom.
If you do add a card, the balance question becomes relevant for the first time, and the PC bottleneck calculator will take both parts — a dedicated card brings its own compute and memory limits with it, which integrated graphics never had to answer for. How this page models the bandwidth ceiling is documented alongside the reason and why the graphics classes are described by generation rather than by model name. To check which component is the limit on a machine you already own, two settings changes will tell you.