GPU at Full Load and Running Hot in Game Menus or Loading Screens, but Normal In-Game? Uncapped Menu Frame Rates, VSync & Load Comparison Diagnosis Guide

Key Takeaways
When a card runs at full load and gets hot in menus or loading screens but behaves normally in gameplay, the comparison itself is the clue: a simple menu scene without a frame-rate limit can demand far more frames than a demanding game scene, and this guide separates uncapped menu rendering, sync settings, and genuine cooling faults with paired tests.
What Exactly Counts as This Symptom
This guide deals with a narrowly shaped symptom, and the shape has to be matched before anything is changed. The machine is idle or quiet on the desktop. A game is launched, and while the game sits in its menus, or while it sits on a loading screen, the graphics card runs at or near full load, its fans spin up loudly, and the card gets noticeably hot. Then the game itself starts, and everything settles: load falls to an ordinary level for that game, fan noise drops, and the card behaves as expected for the rest of the session. Quitting back to the menu brings the high load and the heat straight back.
That contrast is the whole diagnosis. The card is not hot everywhere, and it is not hot at random. It is hot exactly where the picture on screen looks simplest, and normal exactly where the picture looks most demanding. The shape must also repeat: the same game, the same menu, and the same loading screen produce the same pattern each time, and the same game in actual play stays normal each time. A card that is hot in menus, in gameplay, and on the desktop alike is a different shape, and it belongs with general heat and cooling diagnosis rather than this guide. A card that is hot only in gameplay and quiet in menus is the reverse shape, and it also belongs elsewhere. A card that crashes in menus belongs with crash diagnosis, not with this guide, because a crash adds a failure that this symptom does not have. A card whose fans are loud but whose load readings stay low is a different question again, about fan control rather than rendering load.
One more distinction belongs at the start, because it decides which comparisons matter. A menu and a loading screen are not the same kind of scene, even though both sit outside gameplay. A menu is usually a simple, mostly static scene: a background image, a few panels, and some text. A loading screen is often a near-static image with a small animation, shown while the game prepares the next area. Gameplay, by contrast, is a constantly changing three-dimensional scene with far more geometry, lighting, and effects work per frame. Intuition says the demanding scene should load the card more heavily, and the simple scene should load it less. When the observed pattern is the reverse of that intuition, the explanation is almost always about how many frames are being produced, not about how hard each frame is to draw. That is the idea the rest of this guide tests.
The Mechanism: Simple Scenes Can Demand More Frames Than Demanding Ones
A graphics card's total work in a given moment is roughly the work of drawing one frame multiplied by the number of frames drawn in that moment. Gameplay frames are expensive: each one carries a large amount of geometry, shading, and effects work. Menu frames are cheap: each one carries very little. But cheap frames drawn at a very high rate can still add up to more total work than expensive frames drawn at a moderate rate. If nothing limits the frame rate, a simple menu scene can be redrawn hundreds of times per second, because there is almost nothing stopping the card from finishing one frame and immediately starting the next. The card therefore runs at full load drawing pictures that barely change, and it gets hot doing work the viewer never sees, because the monitor can only display a limited number of those frames anyway.
Gameplay, meanwhile, may be limited in ways the menu is not. Many games apply a frame-rate limit, or vertical synchronisation, to gameplay but leave menus and loading screens unlimited. Some games apply the same limit everywhere, and some apply it nowhere, so the pattern varies from game to game, which is why this symptom often appears in one game and not in another on the same machine. Where gameplay is limited, the card draws only as many frames as the limit or the display allows, and then waits between frames. That waiting time is what keeps load and heat at an ordinary level during play. Where menus are unlimited, there is no waiting time, and the card works continuously. The result is the exact pattern this guide describes: full load and heat where the scene looks simple, normal behaviour where the scene looks demanding.
Vertical synchronisation deserves a careful explanation, because it is widely misunderstood in this context. Synchronisation ties the rate at which new frames are presented to the display's own refresh cycle, so the card does not produce frames the display cannot show. When synchronisation is active, the card waits for the display between frames, and that waiting keeps load down. When it is inactive, the card produces frames as fast as it can, regardless of what the display can show. The key diagnostic point is that synchronisation and frame-rate limits are settings, and settings can differ between menus and gameplay within the same game. A setting that is active during play but inactive in menus produces exactly this symptom, with no fault in the card at all. Conversely, a card with a genuine cooling weakness will still show that weakness during sustained gameplay, even if menus also run hot. That is why the menu-versus-gameplay comparison, made carefully, separates a settings problem from a hardware problem.
Loading screens add a second wrinkle. During loading, the game is doing heavy work elsewhere: reading data, building the next area, and preparing textures and geometry. The graphics card may still be redrawing a simple loading image at an unlimited rate at the same time, so high card load during loading does not by itself prove that rendering is the problem. It is also possible for a loading screen to show high load because the card is genuinely busy preparing the next scene. The diagnostic steps below therefore treat loading screens as supporting evidence, not as the primary test. The primary test is always the menu, because the menu is a stable, repeatable, low-complexity scene that can be compared directly against gameplay in the same game.
Step One: Record the Paired Comparison
The first step changes nothing at all. Record the load and the behaviour in each state, as a pair, in the same game, in the same session. Note the card load reading in the menu, the card load reading during gameplay, and whether the fans and the heat follow the same pattern. Note also whether the pattern repeats when the game is quit back to the menu from gameplay, and whether it repeats after the machine has been restarted and the game launched fresh. A single observation of a hot menu is not evidence; a repeated pair, menu high and gameplay normal, in the same game, is evidence.
Also record the frame rate in each state, if the game or an overlay can show it. The frame-rate comparison is the most valuable single fact in this investigation. If the menu frame rate is far higher than the gameplay frame rate, the uncapped-menu explanation gains strong support. If the menu frame rate is similar to the gameplay frame rate, yet the menu load is still far higher, the explanation lies elsewhere, and the later steps matter more. If no frame-rate display is available, the load comparison alone still carries weight, but the frame-rate figure, where it can be obtained, turns a suspicion into a much firmer conclusion. Record both figures side by side, in writing, before changing any setting.
Next, check whether the pattern is limited to one game or appears in several games. Launch a second game, preferably one from a different maker, and compare its menu against its gameplay in the same way. If several games all show hot menus and normal gameplay, the cause is more likely a system-wide setting, such as a driver-level synchronisation or frame-limit setting that applies to menus generally, or a global setting that games inherit. If only one game shows the pattern while others are normal in both states, the cause is more likely that game's own menu settings, and the fix will be found inside that game rather than in system settings. This game-versus-game comparison is cheap, it costs no part-swapping, and it narrows the search more than any other early step.
Step Two: Separate a Loading Screen From a Still Menu
A loading screen and a still menu are not the same test, and they must be recorded separately. A still menu that stays at high load after loading has finished points at the menu render loop and at limit coverage for menus: the menu itself keeps asking for frames without a limit. High load that appears only while loading is in progress, and falls naturally once loading completes, points at the loading work itself, with rendering and resource loading stacked together for a short period. Record the two separately, in the same session, and do not average them: a loading spike that fades on its own is not the same evidence as a still menu that stays high.
Step Three: Apply a Limit and Watch the Menu Alone
With the paired comparison recorded, change one thing: impose a frame-rate limit that covers the menu, and watch the menu alone. The limit can come from the game's own settings, if the game offers a menu-specific or global frame-rate limit, or from the driver control panel, which can usually cap the frame rate for a particular game or for all applications. Set the limit, return to the same menu, in the same game, and record the load, the fan behaviour, and the heat again. Then remove the limit and confirm the menu load returns to its previous high level. That return to the previous state is what proves the limit was responsible, rather than some other change made at the same time.
If imposing a limit brings the menu load down to an ordinary level while the gameplay behaviour stays unchanged, the diagnosis is effectively complete: the menu was uncapped, and capping it removes the unnecessary work. The menu needs no more frames than the display can show, and the extra frames were doing nothing the viewer could see. If imposing a limit changes nothing in the menu, the menu load is not explained by frame count alone, and the investigation moves to the later steps. Either result is progress, provided the limit was the only change made and the result was recorded before and after.
Synchronisation settings should be tested in the same single-variable way, separately from a frame-rate limit. Enable synchronisation for the game, if it was off, and test the menu again. Then, if needed, test with synchronisation on and a limit applied together, but record each stage separately so the effect of each setting is known. Some games override driver-level synchronisation in menus, which is why a driver setting that fixes gameplay may leave the menu untouched. Where that happens, the game's own settings are the place to look, and a game that offers no menu limit at all may need its menu left as it is, with the understanding that the high menu load is the game's own behaviour rather than a card fault. That understanding matters: a card working hard because a game asks it to draw unlimited menu frames is not a card failing.
Step Four: Separate a Settings Problem From a Cooling Problem
A hot menu alone does not prove the cooling system is faulty, but it also does not clear it. The separation test is sustained gameplay. Play the game for a settled session, long enough for the card to reach its steady working state, and watch the load and the heat during play. If gameplay load and heat stay ordinary, and the card's behaviour during play is stable, the cooling system is coping with the game's real workload, and the menu heat is best explained by the extra, unnecessary menu frames. If gameplay also runs hot, or if the card's behaviour during play becomes unstable, the cooling question stays open regardless of what the menu does, and cooling diagnosis becomes the next investigation rather than this guide.
Also compare the desktop, which is the quietest state of all. A desktop that is quiet while menus are hot strengthens the menu-specific explanation, because it shows the card can idle normally when nothing asks it to work. A desktop that is also loaded or hot points away from menus entirely, and toward something running in the background or a wider load problem. The desktop, the menu, and gameplay form a three-way comparison, and each state constrains the explanation in a different direction. A problem that follows only the menu is a menu problem. A problem that follows every state is not.
Background applications deserve one check here, because they can mimic this symptom. An overlay, a recording tool, or a background application that activates when a game launches could add load that coincides with menus. The check is simple: close overlays and background tools, launch the game, and repeat the menu-versus-gameplay comparison. If the pattern survives with nothing else running, background software is cleared. If the pattern changes, the background application was part of the story, and it should be investigated on its own terms rather than blamed on the card.
| Comparison | What the result proves | What it does not prove |
|---|---|---|
| Menu load high, gameplay load normal, repeatedly | The problem follows the menu state, not the game as a whole | Whether the cause is a limit setting or a game-specific menu behaviour, without the limit test |
| Menu frame rate far higher than gameplay frame rate | The menu is drawing far more frames than gameplay | That the card is faulty; extra frames are work the game asked for |
| Several games show the same menu pattern | The cause is more likely system-wide than game-specific | Which system setting is responsible, without testing settings one at a time |
| Only one game shows the pattern | The cause is more likely that game's own menu settings | That the game's menu cannot be limited; its own settings must be checked |
| A frame-rate limit brings menu load down | The menu was uncapped, and the extra frames caused the load | That the limit is the best permanent setting; it is a diagnostic result first |
| A frame-rate limit changes nothing in the menu | Menu load is not explained by frame count alone | What the remaining cause is, without the cooling and background checks |
| Synchronisation on changes the menu | The menu was presenting frames without waiting for the display | That synchronisation is active in menus for every game; some games override it |
| Sustained gameplay stays ordinary | Cooling copes with the game's real workload | That menus will stay cool; menus still need their own limit |
| Sustained gameplay also runs hot or unstable | The cooling question stays open regardless of menus | That menus caused the gameplay heat; the two states need separate judgement |
| Desktop quiet, menus hot, gameplay normal | The card idles normally and the problem follows the menu state | That no background application is involved, without closing overlays and retesting |
| Still menu stays high after loading has finished | Menu render loop and menu limit coverage | That loading work is the cause, without a still-menu record |
| High only during loading, falls naturally once loading completes | Loading work stacked with rendering | That the menu itself is uncapped, without a still-menu record |
Boundary Cases: When It Is Not an Uncapped Menu
First, a card that is hot in menus, in gameplay, and on the desktop alike is not this symptom, and the menu steps in this guide will mislead if applied to it. Second, a card that crashes, freezes, or shows corrupted pictures in menus is a different problem with a failure added, and crash or corruption diagnosis comes first. Third, a card whose fans are loud while its load readings stay low is a fan-control question, not a rendering-load question. Fourth, a single game whose menus are hot but whose gameplay is also hot is not the menu-versus-gameplay contrast this guide depends on, and it should be judged as a general load or cooling question. Fifth, a menu that is hot only while a large download or installation runs in the background may be reflecting background work rather than menu rendering, and the background work should be finished or paused before the comparison is judged. Sixth, a machine that has only just been switched on, with a game launched before background tasks have settled, can show temporarily high load that fades on its own; the comparison should be repeated in a settled session before conclusions are drawn. Seventh, a menu in a game that is known to render a complex animated background, rather than a simple static menu, may legitimately load the card more heavily than a plain menu, and the menu-versus-gameplay comparison should be judged against that game's own menu complexity rather than against an assumed simple menu.
Eighth, instability that appears only after an overclock belongs to clock and voltage settings, and this guide assumes those settings are at their defaults: restore defaults first, then repeat the stage comparison. Ninth, a fan that is not spinning, a blocked intake, or heavy dust inside the case is a cooling-condition problem that affects menus and gameplay alike, and it should not be explained away as a menu limit problem.
Common Misconceptions
Misconception one: a simple-looking menu cannot possibly load a graphics card. It can, because total work is work per frame multiplied by frames per second, and a cheap frame drawn at a very high rate adds up to more work than an expensive frame drawn at a moderate rate.
Misconception two: if the card gets hot, the card must be faulty. A card working hard because a game asks it to draw unlimited menu frames is doing what it was asked to do. Heat under an unnecessary workload is a settings story first, and a hardware story only if the heat persists under the game's real workload.
Misconception three: a frame-rate limit harms gameplay. A limit set at or above what the display can show removes only frames the viewer could never see. Applied carefully, it changes the menu without changing what gameplay looks like, and the before-and-after comparison proves that rather than assuming it.
Misconception four: synchronisation and a frame-rate limit are the same thing. They are related but distinct: synchronisation ties presentation to the display's refresh cycle, while a limit caps how many frames are produced regardless of the display. Testing them separately matters, because a game may honour one and override the other in menus.
Misconception five: high load during a loading screen proves the same cause as a hot menu. Loading screens involve preparation work beyond rendering, so they are supporting evidence, not the primary test. The menu, as a stable and repeatable scene, is the cleaner comparison.
Misconception six: if one game shows the pattern, every game will. Menu frame handling varies from game to game, so a pattern in one game and normality in another points at that game's own settings rather than at a general card fault.
Misconception seven: turning the fans up fixes the problem. Faster fans remove heat more quickly, but they do not remove the unnecessary work that creates the heat. The fix for uncapped menu rendering is to stop drawing frames nobody can see, not to cool the card harder while it keeps drawing them.
When to Stop and Seek Advice
Stop changing settings if any of the following appears. The card crashes, freezes, or shows corrupted pictures in menus or gameplay, which are failures beyond this guide's symptom. Gameplay itself runs hot or becomes unstable during a settled session, which keeps the cooling question open. The pattern persists after a frame-rate limit, synchronisation, and background-application checks have each been tested one variable at a time with no change. Or any step feels unsafe, in which case the setting should be left as it is and advice sought first.
Bring five items to any inspection: the menu load and gameplay load as a recorded pair in the same game, the menu frame rate and gameplay frame rate side by side, whether other games show the same pattern, the result of applying a frame-rate limit to the menu alone, and whether sustained gameplay stays ordinary. Those five let advice start from the comparison directly, without repeating tests already done.
Readers nearby are welcome to bring their records to HK GPU Fix for advice, at Room 1209, Dali Plaza, Shek Mun, telephone 2186 8010, WhatsApp 6558 6806. There is no Mong Kok shop, and there is no home-visit service. Describing the menu load, the gameplay load, and the frame-rate figures for both states is enough for a quick judgement on the next step.
Frequently Asked Questions
Why does a simple menu load the card more than demanding gameplay?
Because total work is work per frame multiplied by frames per second. A cheap menu frame drawn at a very high rate, with no limit, can add up to more work than an expensive gameplay frame drawn at a moderate, limited rate. The menu looks simple, but the card is drawing it far more often.
How can I tell whether the menu is uncapped?
Compare the frame rate in the menu against the frame rate in gameplay, in the same game. A menu frame rate far higher than gameplay, together with high menu load and normal gameplay load, strongly supports the uncapped-menu explanation. Applying a frame-rate limit to the menu and watching the load fall confirms it.
Does a hot menu mean my graphics card is failing?
Not by itself. A card drawing unlimited menu frames is doing what the game asked it to do. The card is a concern if heat or instability persists during sustained gameplay, which is the game's real workload, after menu settings have been dealt with.
What is the difference between synchronisation and a frame-rate limit?
Synchronisation ties frame presentation to the display's refresh cycle, so the card waits for the display between frames. A frame-rate limit caps how many frames are produced, regardless of the display. A game may honour one in menus and override the other, so they should be tested separately.
Why does only one of my games show this pattern?
Because menu frame handling varies from game to game. Some games limit menus, some leave them unlimited, and some let driver settings govern menus while others override them. A pattern confined to one game points at that game's own settings.
Why are loading screens less reliable as evidence than menus?
Because loading involves preparation work beyond rendering, such as reading data and building the next area, so high load during loading may reflect that preparation rather than menu rendering. Menus are stable, repeatable scenes, which makes them the cleaner comparison.
Will capping the frame rate make my games look worse?
A limit set at or above what the display can show removes only frames the viewer could never see. The before-and-after comparison in the menu, and in gameplay, shows whether anything visible changed, rather than assuming it.
Should I just turn the fans up instead?
Faster fans remove heat more quickly but do not remove the unnecessary work creating it. Capping the menu frame rate stops the extra frames from being drawn at all, which addresses the cause rather than the heat alone.
What should I bring if I seek advice?
Bring the menu load and gameplay load as a recorded pair, the frame rate in each state, whether other games show the same pattern, the result of applying a frame-rate limit to the menu, and whether sustained gameplay stays ordinary. That record lets advice start from the comparison rather than repeating the whole diagnosis.