USB Keyboard & Mouse Disconnect Under Heavy Load? Load-Linked Dropouts, Hub Power & Controller Diagnosis

| HK GPU Fix

USB Keyboard & Mouse Disconnect Under Heavy Load? Load-Linked Dropouts, Hub Power & Controller Diagnosis

Key Takeaways

When a keyboard and mouse disconnect only under heavy load and reconnect moments later, the cause is usually a hub, port group, cable, power-management setting or controller timeout, not a broken device. This guide classifies the dropout pattern, explains each mechanism, and walks through an isolation sequence that narrows the range before any driver or firmware change is considered.

Three Ways a Dropout Pattern Can Present

Every useful step in this guide depends on a classification made before any setting is touched. The symptom is a keyboard and mouse that disconnect and reconnect: the device may sound its disconnect and reconnect cues, its indicator light may go dark and light again, and input may be unresponsive for a short interval before recovering. When that pattern appears only under heavy load, while ordinary writing and browsing stay normal, the reader is in the territory of this guide. But the same visible dropout can belong to three different patterns, and each pattern points at a different range of causes. Classifying first is not paperwork; it decides which tests mean anything at all.

The first pattern is the main subject of this guide: dropouts that occur only under heavy load, while idle use stays completely normal. The machine is fine for ordinary work, and the keyboard and mouse only disconnect once a game or another heavy load situation begins. This pattern points at load-related triggers: hub power behaviour, controller load handling, power-management behaviour at load transitions, and cable or contact margins that only show themselves under load.

The second pattern is dropouts that occur at any time, with no relationship to load. The keyboard and mouse disconnect while idle just as they do while gaming, and heavy load merely makes the user more likely to notice, because a dropout during play is far more disruptive than a dropout while reading. For this pattern, the load tests in this guide have limited value, and the reader should first follow the logic of a persistent dropout: device, cable, port, and driver checks that do not depend on load at all. Forcing a load explanation onto a pattern that ignores load wastes the investigation.

The third pattern is dropouts tied to one port or one hub: the keyboard and mouse disconnect only when connected through a particular port or a particular hub, and work normally when moved to another port or plugged directly into the machine. Even when the dropout happens to coincide with heavy load, the range evidence takes priority over the load explanation, because the port or hub binding has already narrowed the suspect area. The bypass-the-hub and change-port-group tests in this guide exist precisely for this pattern.

Two further observations must be recorded alongside the classification, because they can change the diagnosis immediately. First, note the range of affected devices when a dropout occurs: only the keyboard, only the mouse, both together, or other devices on the same hub at the same time. A dropout that takes every device on a hub down together points at the hub or its port group, while a dropout confined to one device points more narrowly. Second, note whether the whole machine restarts or the screen goes black at the same time. If it does, stop this guide at once: that pattern belongs to power and restart diagnosis, and this guide does not cover it. The guide that follows assumes the machine itself stays up, the screen stays on, and only the USB devices disconnect and reconnect.

Seven Mechanisms, With the First, Second, and Seventh Given Their Full Weight

A short version of this topic says the hub lacks power and stops there. That summary is wrong in two ways: it misdescribes what keyboards and mice consume, and it hides the controller, cable, power-management, and timeout mechanisms that produce the identical visible dropout. Each mechanism below must be understood before testing, and the first, second, and seventh deserve particular care, because they are the ones most often flattened into a single misleading phrase.

Hub power budget and power method

A hub without its own power supply shares the power of its upstream port among every device connected downstream. A hub with its own power supply powers its downstream devices from its own supply instead. Keyboards and mice themselves draw little power, so the mechanism here is not a keyboard or mouse consuming too much. The mechanism is sharing plus reset: when several devices share one upstream budget, or when downstream devices work simultaneously under heavy load, the hub may reset internally and force its downstream devices to re-enumerate, which is the technical name for the machine rediscovering a device as though it had just been connected. The visible result is a keyboard and mouse that disconnect and reconnect.

The diagnostic consequence is that the hub cannot be judged in isolation from what is connected to it. The same hub may behave perfectly with only a keyboard and mouse attached, and reset when further devices are added, because the combination changed the sharing and reset behaviour. Testing must therefore vary the downstream combination one device at a time, not simply swap the hub and declare the matter settled.

Motherboard USB controllers and port groups

The USB ports on a machine are not all wired to the same controller. Front ports, rear ports, and different port groups may pass through different internal hubs or controllers. Devices on the same controller group share that controller processing and power path. When that group controller, under heavy load, accumulates timeouts or errors and resets a port, the devices in that group disconnect and reconnect together, while devices on another group may be entirely unaffected.

This grouping mechanism is the explanation behind the otherwise puzzling observation that moving to another port cures the problem. The cure is not magic, and it is not evidence that the first port is broken in any absolute sense. It is evidence that the devices left a controller group whose processing or power path was resetting, and joined a group that was not. Any tester who does not understand grouping will treat a port change as a random trick rather than as a range-narrowing test, and will fail to record which group the devices came from and which group they moved to.

Power management: selective suspend and device power control

The operating system provides USB selective suspend, and a setting that allows the system to switch a device off to save power. Motherboard and processor power state transitions also affect controller and port power states. The common misunderstanding is that power saving only matters while idle. Load transitions are themselves state change points: moving from idle into heavy load, or from heavy load back to idle, changes power states, and a controller or hub may enter or leave a power-saving state at exactly that moment, causing a device to re-enumerate.

The practical rule is simple: record the original power-management settings before changing any of them, change only the power-management layer in its step, and restore the settings if the step does not confirm a cause. Changing power settings together with a port change or a driver change destroys the comparison, because the tester will never know which layer mattered.

Controller timeout and reset under heavy load

A gaming mouse with a high polling rate generates dense interrupt traffic. When the system is simultaneously heavily loaded, with processor, graphics, and storage all busy, the controller driver handling of USB traffic may be delayed, a device may fail to complete its handshake in the expected time, and the host may judge the transfer timed out and reset the port. The device then re-enumerates and recovers. The signature of this mechanism is a dropout paired with a reconnection, recovery within a short interval, and a keyboard and mouse that are themselves undamaged.

Timeout reset must be kept conceptually separate from device failure. A device that reconnects cleanly and works normally between dropouts has demonstrated that it can operate; the failure is in the timing of the host-device exchange under load, not in the device hardware. Treating every timeout reset as a broken device leads to replacing hardware that was never faulty, while the controller timing problem remains.

Cables, contact margins, and signal integrity

Extension cables, adapter cables, ageing or marginal-quality cables, loose connectors, and the internal wiring of front ports all reduce signal margin. Under light load and stable conditions, a marginal connection may be masked: transfers succeed, and nothing visibly goes wrong. Under heavy load, fan speed rises and vibration increases, internal temperature changes, and several devices work at once, and the marginal connection can cross into visible dropout. The correct language for this layer is margin, not certainty. A cable that looks fine can be marginal, and a cable that looks worn can still pass a controlled comparison.

Proof in this layer comes only from substitution comparisons, not from appearance: change the cable alone, keeping device and port identical, and re-test; or change the port alone, keeping cable and device identical. A dropout that follows the cable across ports implicates the cable. A dropout that stays with the port across cables implicates the port. Appearance alone implicates nothing.

Driver and firmware layers

Chipset drivers, USB controller drivers, and motherboard firmware USB settings affect how a controller handles mixed devices on the same controller, such as high-speed devices alongside a low-speed keyboard and mouse, how power state transitions are handled, and whether a port is reset when errors accumulate. The signature of a driver or firmware layer problem is negative: port changes, hub changes, cable changes, and device changes have all failed to isolate a range, while the problem remains bound to one controller group.

This layer belongs last in the sequence, because driver changes have a wide effect and can mask the evidence that earlier steps would have produced. Updating a driver before range isolation, and watching the problem disappear, leaves the true cause unknown; the next time a port or hub changes, the problem may return, and the tester will have no record of what actually mattered.

Whole-machine power change under heavy load, as an exclusion hypothesis only

Under heavy load the whole machine draws more power, and the margin on the motherboard USB power path may become tighter, making a marginal hub or controller more likely to reset. This may be used only as a hypothesis to be tested by port and hub isolation, and it must never be written as a conclusion that the power supply is inadequate and must be replaced. The counter-evidence standard must be stated alongside the hypothesis: if the machine does not restart, the screen does not go black, and other power paths behave normally during the dropout, there is no evidence pointing at whole-machine power. If the whole machine restarts or the screen goes black at the same time, the problem belongs to power and restart diagnosis, and this guide stops. This guide is not a power supply buying guide, and it contains no power recommendation and no brand recommendation.

The wireless receiver as its own variable

A wireless keyboard and mouse receiver plugged into a hub carries three layers of variables at once: the hub power behaviour, the controller grouping of the port the hub uses, and the signal environment around the receiver. The same receiver plugged directly into a rear motherboard port faces a different combination of all three. Receiver position must therefore be treated as an independent variable in testing, not folded into a general claim about interference or batteries. Moving the receiver from a hub to a direct rear port is a controlled comparison in its own right, and it must be performed before any claim about interference or battery condition is made.

Record the Baseline Before Changing Anything

Before any change, record how the keyboard and mouse are connected: directly or through a hub, whether the hub has its own power supply, whether a front or rear port is used, and what else is connected to the same hub. Record the wireless receiver position if the devices are wireless. Record the affected range during a dropout: only the keyboard, only the mouse, both together, or other devices together. Record the relationship to load: only during games or heavy load, or also while idle. And record, as a diversion-critical fact, whether the whole machine restarts or the screen goes black. Without this baseline, later results cannot be read, because the tester will not know the starting arrangement that produced the problem.

The Isolation Sequence, In Order

The sequence runs from classification, to range isolation, to settings, and only then to drivers and firmware. Each step changes one variable, re-tests under the same heavy load situation, and records the result. Any step that makes the problem disappear must be followed by a restore confirmation: return to the original arrangement and confirm the problem returns. Without that confirmation, the step yields only a suspect, never a confirmed cause.

First, classify. Using the same devices, observe over a period in an idle situation and in a heavy load situation, and record whether dropouts occur in each. Dropouts while idle as well point to the second pattern; dropouts only under load point to the first; dropouts bound to one port or hub point to the third. The classification decides how the later steps are read, and it must not be skipped.

Second, bypass the hub. Move the keyboard and mouse from the hub to a direct rear motherboard port, changing nothing else, and re-test under the same heavy load situation. If the problem disappears, the range has narrowed to the hub or the original port group. If the problem persists, the hub is not the main cause, and the next step begins without also changing the cable or the driver.

Third, test the hub alone, but only if the bypass step implicated the hub. Connect the keyboard and mouse alone to that hub, with no other device on it, and re-test. Then add the other devices back one at a time, re-testing after each addition. This separates a hub that is itself problematic from a combination of devices that only fails together. Adding devices back in a group defeats the purpose; one device at a time is the method.

Fourth, change port group. Move the keyboard and mouse to a different port group, for example from a front port to a rear port, or from one port group to another, and re-test. If the problem disappears, the original port group or its internal wiring is implicated. If it persists, the range is not the port group. The port must be changed without changing the cable at the same time, or the tester cannot tell whether the port or the cable mattered.

Fifth, change cable and device, one item at a time. First change only the cable, keeping device and port identical, and re-test. Then change only one device, the keyboard or the mouse, keeping port and cable identical, and re-test. For wireless devices, move the receiver position, from a hub to a direct motherboard port, as the equivalent comparison, since receiver position is itself a variable.

Sixth, compare load types. Re-test under different kinds of heavy load, such as different games and different heavy load applications, to establish whether the dropout follows any heavy load or only one application. If it follows only one application, the problem may lie in that application device interaction or overlay behaviour rather than in USB itself, and the reader must be told that diversion plainly, instead of having an individual application problem written up as a general USB problem.

Seventh, test power-management settings, changing only that layer. Record the original settings, then switch off USB selective suspend and switch off the setting that allows the system to switch the relevant devices off to save power, and re-test. If the problem disappears, restore the settings one at a time for confirmation, to identify which setting mattered. If it persists, restore all power-management settings and move on. This step must not be merged with the driver step.

Eighth, and only after ports, hub, cable, device, and power management have all been isolated and the problem still points at one controller group, consider the driver and firmware layer: updating chipset or USB controller drivers through the motherboard maker official channel, and checking motherboard firmware USB settings. Record the original settings and driver state before changing them. Updating a driver is not an opening move; without range isolation first, a disappearance after an update proves nothing about the true cause, and the problem may return the next time a port or hub changes.

Reading the Classification Table

Dropout patternPoints towardNext test
Only under heavy load, normal while idleLoad-related trigger: hub, controller, power transition, or timeoutBypass the hub, then change port group
At any time, including while idlePersistent device, cable, port, or driver problemPersistent dropout logic: device, cable, and port checks without load
Only through one port or one hub, normal when movedThat port group, that hub, or its internal wiringBypass the hub, change port group, test the hub alone
Every device on the same hub drops togetherThe hub or its port groupTest the hub alone, adding devices back one at a time
Only one device drops, others on the same hub unaffectedThat device, its cable, or its receiver positionChange cable, change device, move the receiver
Whole machine restarts or screen goes black at the same timeOutside this guide: power and restart diagnosisStop this guide and follow the power and restart diversion

Eight Misconceptions, Each Taken Apart

A dropout does not mean the keyboard or mouse is broken and must be replaced at once. Disconnect and reconnect is the visible form of the host and device re-enumerating, and its causes include hub behaviour, port grouping, cable margins, power management, and controller timing. Replacing the device without a direct-connection comparison and a device-swap comparison may leave the real cause untouched, and the new device may disconnect in exactly the same way.

A dropout during play does not mean the graphics card is faulty. Play is only a time marker for heavy load; the overlap in time carries no causal force. Graphics card faults typically present as picture abnormalities, black screens, or restarts, which are different from a clean USB disconnect and reconnect. Inferring a graphics fault from timing overlap alone reverses the burden of proof, and this guide states the point explicitly so that readers do not carry it forward.

A dropout does not mean the power supply is inadequate and must be replaced with a larger one. This guide draws no whole-machine power conclusion, and it offers no power supply buying advice. A clean USB dropout with no machine restart and no black screen does not prove whole-machine power inadequacy. Only a pattern that includes whole-machine restart or black screen belongs in power and restart diagnosis, and even there the conclusion follows its own evidence, not this guide.

Hubs are not all alike, and powered and bus-powered hubs do not behave the same way. A bus-powered hub draws its downstream power from the upstream port and shares it, while a powered hub supplies its downstream devices from its own supply. The number and combination of devices on the hub also change its behaviour. Advice that says only to try another hub, without first establishing the hub power method and the downstream device list, skips the two facts that decide what the hub comparison means.

Updating drivers does not solve every dropout. The driver layer only becomes meaningful once range isolation points at a controller group. If the root cause is hub power behaviour or port grouping, a driver update does not change the physical range of the hub or the port group. Updating before isolation is a way of masking evidence, not of finding a cause.

A wireless dropout is not automatically interference or a battery problem. Receiver position is a variable in its own right: a receiver on a hub and the same receiver plugged directly into the machine face different power behaviour, controller grouping, and signal environments. The receiver position comparison must come before any interference or battery conclusion, or the tester will chase an explanation that the position change alone would have excluded.

Front ports and rear ports are not the same group, and changing ports is not meaningless. Front ports pass through internal case wiring and an internal connection, and may belong to a different controller group or pass through an extra internal hub, while rear ports connect directly to the motherboard. Changing port groups is a genuine isolation test, and dismissing it discards one of the cheapest range-narrowing comparisons available.

Changing cables, hub, and driver all at once and watching the problem vanish does not equal a repair. With several variables changed together, no variable can be credited, and restoring any one of them later may bring the problem back. The discipline of one variable at a time plus a restore confirmation is not pedantry; it is the only thing that separates a confirmed cause from a lucky run.

Boundaries: What This Guide Does Not Cover

Virtual reality headset cable power, bandwidth, and tracking problems belong to a separate guide, and this guide covers only ordinary keyboards and mice. USB-C video output, alternate mode display output, high-speed interconnect standards, and docking station problems belong to their own guide, and this guide does not discuss video over USB. Display built-in keyboard-video-mouse or hub switching between two machines belongs to a separate guide; this guide covers disconnection and reconnection on one machine only, and a dropout that occurs only when switching machines should be diverted there. Graphics transient power behaviour triggering power supply protection, whole-machine trips, or restarts belongs to its own guide, and whole-machine power inadequacy causing restarts or black screens belongs to another; a dropout accompanied by a restart or black screen must be diverted to those guides at once. Display no-signal, cable, and interface problems belong to a display guide, and this guide does not discuss picture signals. An individual device already proven faulty by a device-swap comparison, where the same device disconnects on every port under every load, leads only to the conclusion that the device should be replaced as a comparison result, and this guide does not expand that conclusion into a product recommendation, nor does it name any particular model. This guide also draws no graphics card repair conclusion.

When to Stop Testing and Seek Inspection

Stop the isolation sequence and consider inspection when every step has been completed without isolating a range, when the dropout is joined by symptoms outside this guide such as a whole-machine restart, a black screen, corrupted graphics, or failure to start, or when any step produces a freeze, a restart, or an abnormal smell or sound. Bring the test record to the inspection: what was changed, what the original arrangement was, what happened after the change, and what happened after the restore. That record lets an inspection start from the isolated evidence instead of from a vague report.

HK GPU Fix publishes an inspection and repair range of HK$400–2,800, subject to the published pricing, with the final quote after inspection. Inspections are handled at the shop: Room 1209, Dali Plaza, Shek Mun. For enquiries, call 2186 8010 or WhatsApp 6558 6806. There is no on-site service, so the machine should be brought to the shop for inspection.

FAQ

Why do my keyboard and mouse work fine normally, but disconnect when I start gaming?

Ordinary use keeps load low and conditions stable, which can mask a marginal hub, controller group, cable, or power-management behaviour. Heavy load changes power states, increases controller traffic, and stresses shared paths, and those changes can expose a margin that idle use never reaches. The load does not prove any one part is faulty; it reveals which layer needs isolation testing.

Plugging directly into the machine fixed the dropouts. What does that prove?

It narrows the range to the hub or the original port group, provided nothing else changed at the same time. It does not yet prove which of the two is responsible. The next steps are to test the hub alone with only the keyboard and mouse attached, then add devices back one at a time, and to change port group separately, so that the hub and the port group are distinguished.

What is the difference between a powered hub and a bus-powered hub?

A bus-powered hub shares the power of its upstream port among all downstream devices, while a powered hub supplies its downstream devices from its own power supply. The downstream device list matters in both cases, because the combination of devices changes sharing and reset behaviour. Before judging a hub, record its power method and everything connected to it.

Should I update drivers before doing the port and hub tests?

No. Driver and firmware changes belong last, after ports, hub, cable, device, and power management have been isolated. Updating first can mask the evidence that range isolation would have produced, and a disappearance after an update proves nothing about the true cause. If the problem still points at one controller group after isolation, then the driver layer becomes the meaningful next step.

My wireless mouse keeps disconnecting. Is it interference or the battery?

Neither conclusion should come first. Receiver position is an independent variable: a receiver on a hub faces different hub power behaviour, controller grouping, and signal environment from the same receiver plugged directly into a rear motherboard port. Move the receiver and re-test under the same load before drawing any conclusion about interference or battery condition.

When does this stop being a USB problem and become a power or restart problem?

The moment the whole machine restarts or the screen goes black at the same time as the dropout. A clean USB disconnect and reconnect, with the machine staying up and the screen staying on, belongs in this guide. A restart or black screen belongs to power and restart diagnosis, and this guide must be stopped in favour of that diversion.

I replaced the cable, hub, and driver together and the dropouts stopped. Am I done?

No finding has been made. With several variables changed together, no variable can be credited, and restoring any one of them later may bring the problem back. Restore the changes one at a time, re-testing after each restore, until the disappearance and the return on restore point at the same source. Only then is the cause confirmed.

Related articles

Related services

If you need Power connector repair, see these service pages: