GPU Fails Through a Standalone KVM/HDMI Switch but Works Direct: EDID Emulation, Bandwidth and Switching Diagnosis Guide

Key Summary
A computer that works perfectly when plugged straight into the monitor can fail through a standalone KVM or HDMI switch, because the switch changes what the computer sees. This guide separates EDID emulation, bandwidth, HDCP relay, power and firmware, and cable faults, using a direct-connection comparison, a one-variable downgrade comparison, and the switching action itself as the test.
The Shape of the Problem: Direct Works, Through the Switch Fails
Only one narrow pattern is covered here, and the pattern itself is the diagnosis. The same computer, the same cable, and the same monitor work completely normally when the cable runs straight from the graphics card to the monitor. The moment a standalone KVM switch or a standalone HDMI switch is placed in the middle — a separate box that selects between two or more computers, not a KVM built into the monitor, and not an adapter cable — the picture fails. The failure can take several forms: no signal, a black screen, a resolution that has dropped back to a lower value, a refresh rate that has disappeared from the settings, windows that have moved from where they were, or protected video content that plays with sound but no picture. The common thread is the change of path: direct works, through the switch fails.
Three facts pin the shape down, and all three have to hold. First, the direct connection works. The same computer, the same cable, the same monitor, and the same display mode produce a normal picture when the switch is bypassed. Second, the failure appears only through the switch. The switch is the variable, and the failure follows the switch rather than following the computer or the monitor. Third, the switch is a standalone device. It is a separate box with its own inputs, its own switching button or hotkey, and its own specification — not a monitor menu, not a cable that converts one connector type to another, and not a dock that shares bandwidth between several displays. A failure through a monitor's built-in KVM, through an adapter, or through a dock belongs with a different guide, because the device in the middle works differently and the tests are different.
That distinction matters because the most frequent mistake with this shape is to blame the graphics card or the monitor. A direct connection that works proves that neither of them is at fault. The card can output the signal, the cable can carry it, and the monitor can display it. Something in the middle is changing the signal, or changing what the computer believes about the monitor, and the diagnosis has to stay in the middle. Reinstalling a driver, changing the card, or replacing the monitor will not change what a switch in the middle does to the signal, and it wastes the evidence that the direct comparison has already given: the two ends work, and the middle does not.
The shape also has to be separated from a failure that appears at switch-on. A machine that needs its cable unplugged and plugged back in every time it is switched on is a cold-start handshake shape, and it belongs with a different guide. This guide covers failures that appear through the switch during normal use, and failures that appear at the moment of switching between computers. The switching action itself is the variable here, not the power-on sequence. A switch that works until the moment it is switched, and fails at that moment, is telling the reader something about the switching action — about what the computer sees when the switch moves, and about what the switch tells the computer when it moves back.
A second limit belongs at the front of the guide. This guide does not teach how to use an adapter, and it does not reproduce the bandwidth table of the direct-connection guide. It uses one step from that subject — lowering one variable at a time as a comparison — and nothing more. It does not teach window arrangement, and it does not reproduce the guide for windows that move when a monitor disappears. It uses the observation — windows moving after a switch — as evidence for one layer, and nothing more. It does not teach HDCP as a subject in its own right. It uses one comparison — ordinary content against protected content — as the fourth fork, and nothing more. It stays within its own pattern by using one step at a time and not rewriting the guides next door.
What EDID Is, and What a Switch Does With It
EDID — Extended Display Identification Data — is the capability record of a monitor. It lists the resolutions the monitor can display, the refresh rates it supports, the colour formats it accepts, the audio modes it handles, and the monitor's name. The graphics card has to read the EDID before it knows what the monitor can do, and Windows only offers the display modes that appear in the EDID. Without the EDID, the card does not know what to send, and the system falls back to a conservative mode or to no picture at all. The EDID is therefore not a detail. It is the document on which the whole display negotiation is built.
A switch sits in the middle of that negotiation, and it can handle the EDID in three ways. The first way is pass-through: the switch passes the monitor's EDID through to the computer unchanged, so the computer reads the monitor's real record. The second way is emulation: the switch presents its own fixed EDID to every computer connected to it, pretending to be the monitor, so that each computer always believes a monitor is present even when the switch is currently showing a different computer. The third way is nothing at all: when the switch moves away from a computer, that computer receives no EDID, because the switch has stopped passing anything to it. The third way is the one that causes the failures in this guide.
When a switch offers no emulation, switching away from a computer leaves that computer with no monitor signal. The system treats the monitor as disconnected: the desktop area shrinks, windows are moved back to the primary monitor, and the resolution and refresh rate are reset to conservative values. Switching back requires the system to enumerate the monitor again, to rebuild the display identity from scratch, and to re-establish the mode. Windows do not return to their previous positions automatically, because the system has already rearranged them for a smaller desktop. The resolution does not return automatically either, because the system has already reset it. The mechanism is the same one described in the guide for windows that move when a monitor disappears — but that guide takes window arrangement as its subject, while this guide takes the switch as its subject. The question here is not how to put the windows back. It is which device made the monitor disappear, and whether the switch was supposed to prevent it.
EDID emulation exists to prevent exactly that disappearance. Its job is to make the computer believe, at all times, that the monitor is still connected, so that the computer never re-enumerates the display and never rearranges the desktop. That job has a strict boundary, and the boundary has to be written plainly: emulation does not add bandwidth, it does not repair a faulty cable, and it does not let a switch support a higher resolution than its hardware allows. A switch that supports only HDMI 2.0 remains an HDMI 2.0 switch when emulation is added, and a 4K 120Hz mode still cannot pass through it. The emulation solves the "monitor disappeared" layer. The bandwidth solves the "mode cannot pass" layer. The two layers are different, and a switch that has emulation can still fail on bandwidth, just as a switch with ample bandwidth can still fail on emulation. Writing that a switch with EDID emulation is therefore guaranteed to work is wrong, and it is one of the errors this guide is written to prevent.
HPD — hot-plug detect — is the second half of the same mechanism. HPD is the signal that tells the computer that a monitor is connected. When a switch moves, the HPD signal changes, and the change tells the computer to re-enumerate the display. A switch with EDID emulation usually holds the HPD signal steady as well, so that the computer never notices the switch at all. A switch without emulation lets the HPD signal drop when it switches away, and the computer re-enumerates. EDID missing and HPD dropped, together, produce the characteristic pattern of this layer: the resolution resets, the windows move, and the refresh rate falls back. Either signal on its own can cause part of the pattern, but the two together are the full explanation for why a switch without emulation rearranges the desktop every time it moves.
Bandwidth: The Switch Version Is the Ceiling
The display chain is a chain: the graphics card output, the switch, the cable, and the monitor input. Every link in the chain has to support the mode being sent, and the weakest link sets the ceiling for the whole chain. When the mode exceeds what any one link can carry, the system falls back to a conservative mode rather than failing outright. The monitor does not report that the switch is the limit. It simply shows a lower resolution, a missing refresh rate, or a disabled HDR setting, and the reader has to infer the cause from the pattern of what works and what does not.
The numbers in the sources give the ceiling a concrete shape. An HDMI 2.0 link carries up to 18 Gbps, and a 4K 60Hz signal at 4:4:4 colour sampling in 8-bit already sits close to that ceiling. An HDMI 2.1 link carries 40 or 48 Gbps, and that capacity is what a 4K 120Hz or 144Hz signal with HDR and variable refresh rate requires. An older switch does not announce that it has capped the mode. It quietly drops the mode back: the resolution falls to 1080p, the refresh rate disappears from the settings, or HDR cannot be switched on. All three are bandwidth-layer symptoms, and all three are silent. The switch does not report an error. It reports a lower mode, and the lower mode is the error message.
The phrase "supports 4K" has to be read with the same caution. A switch described only as supporting 4K, with no refresh-rate number beside it, may support 4K at 30Hz or at 60Hz and no higher. Published guidance on switches warns that a 4K claim without a refresh-rate figure should be treated with suspicion, because the claim most often hides a 30Hz or 60Hz ceiling — but the same guidance also warns against the opposite error, which is to assume that every switch without a refresh-rate figure is necessarily a low-ceiling switch. The correct step is to check the specification of the particular model: the HDMI or DisplayPort version, the bandwidth figure in Gbps, the refresh rate at each resolution, and whether HDR and variable refresh rate are passed through. No general claim about all switches can replace that check, and this guide does not make one.
The mode is not one number. It is a combination: resolution, refresh rate, colour depth, HDR, variable refresh rate, and colour sampling, all counted together. A 4K 60Hz signal with HDR and 4:4:4 sampling needs more bandwidth than the same resolution and refresh rate without them. That is why the downgrade comparison in this guide lowers one variable at a time: resolution first, then refresh rate, then HDR off, then variable refresh rate off, then colour sampling from 4:4:4 to 4:2:2. Each step reduces the bandwidth demand by one variable, and the step at which the picture becomes stable identifies the variable that pushed the chain past its ceiling. Lowering several variables at once would show that the picture works, but it would not show which variable mattered, and the diagnosis would end with a working low mode and no layer identified.
HDCP Relay: Only Protected Content Fails
HDCP is a content-protection handshake. The player, the graphics card, the switch, and the monitor all have to support the same HDCP version — HDCP 2.3, for example — before protected content can be played. When the switch does not support HDCP, or when its version does not match the version the other devices expect, protected content fails while ordinary content works normally. The screen goes black, or the picture disappears while the sound continues, but only for protected video: a streaming service playing protected content, for example. The desktop, a game, and an ordinary video play normally through the same switch, on the same cable, at the same mode.
That contrast is the fourth fork, and it is a clean one, because it separates the HDCP layer from every other layer in the guide. A bandwidth failure affects ordinary content too. An EDID failure affects the desktop too. A cable failure affects everything. Only an HDCP relay failure affects protected content alone. The test is therefore to play ordinary content and then protected content through the same chain, on the same machine, and to watch which one fails. Ordinary content normal and protected content black is the HDCP relay layer. Ordinary content black as well is not the HDCP layer, and the diagnosis has to go back to the bandwidth, EDID, or cable layers. The HDCP conclusion must never be applied to a general black screen, because a general black screen has a different cause, and applying the HDCP answer to it sends the reader to the wrong layer.
The HDCP mechanism also explains why the direct comparison matters for this layer. If protected content plays normally when the computer is plugged straight into the monitor, and fails only through the switch, the switch is the device that breaks the relay. The direct connection proves that the player, the card, and the monitor all support the version required. The switch is the one link that does not, and the failure follows the switch. That is the same logic as the first fork applied to one kind of content, and it is the reason the direct comparison is the threshold step for the whole guide: without it, no later comparison means anything.
A Switch Is Not an Adapter, and Not a Dock
Four devices look similar from the outside, and they have to be kept apart, because their diagnosis runs in different directions. A standalone switch selects between several sources and shows one at a time; its key variables are the switching action, the EDID emulation, and the bandwidth ceiling. An adapter converts one protocol to another — DisplayPort to HDMI, for example — in a static connection that never switches; its key variables are passive against active conversion, the protocol, and the bandwidth of the conversion. A dock or an MST hub shares its bandwidth between several monitors, so the specification for one monitor does not apply when two are connected; its key variable is the shared bandwidth. A DisplayLink dock sends its picture over USB as compressed data rather than as a native display signal, and its diagnosis runs through USB bandwidth and compression rather than through a display link at all. The four devices cannot be mixed, and the diagnostic logic of one cannot be applied to another.
The boundary sources state the separation plainly. The guide for adapters takes protocol conversion as its subject, and its tests — passive against active, protocol, bandwidth — do not apply to a switch, because a switch does not convert a protocol. The guide for bandwidth and display modes states in its own boundary that a KVM switch is outside its scope, and its bandwidth table is used here only as a comparison step, not as the subject. The guide for window arrangement states in its own boundary that the signal layer is outside its scope, and its window observations are used here only as evidence for the EDID layer. The guide for HDCP takes the handshake as its subject, and its mechanism is used here only as the fourth fork. A switch is a switch. Treating it as an adapter, a dock, or a hub is a category error, and it is one of the errors this guide is written to prevent.
| Device | What it does | Key variable in diagnosis | Why its logic does not transfer |
|---|---|---|---|
| Standalone KVM or HDMI switch | Selects one of several sources, shows one at a time | Switching action, EDID emulation, bandwidth ceiling | The switching action itself is the variable; no other device has one |
| Adapter or adapter cable | Converts one protocol to another, static connection | Passive against active, protocol, conversion bandwidth | No switching action; the connection never moves |
| Dock or MST hub | Shares bandwidth between several monitors | Shared bandwidth across all connected monitors | One monitor's specification does not apply when two share the link |
| DisplayLink dock | Sends picture over USB as compressed data | USB bandwidth, compression, driver | Not a native display signal; the link is USB, not HDMI or DisplayPort |
The Deciding Forks: Four Comparisons That Narrow the Field
Four comparisons carry the diagnosis here, and every comparison ends in one of two answers. They belong in sequence, since each answer narrows the next question, and the opening comparison — the direct comparison — is the gate for the whole guide. Without it, every later comparison is meaningless, because the fault could sit in the computer or the monitor rather than in the switch.
The first fork is the direct comparison. The same computer, the same cable, the same monitor, and the same display mode are connected directly, with the switch bypassed entirely. If the direct connection works normally, the switch layer — or the switch-and-cable combination layer — is the working answer, and the diagnosis continues. If the direct connection fails as well, the switch is not the problem, and the diagnosis must not continue circling the switch. The direct comparison is the gate. It has to be run before any other test, and its result decides whether the rest of the guide applies at all. A reader who skips it and starts changing switch settings is diagnosing a device that may never have been at fault.
The second fork is the downgrade comparison. Through the switch, one variable is lowered at a time: the resolution from 4K to 1440p to 1080p, the refresh rate from 144Hz to 120Hz to 60Hz, HDR switched off, variable refresh rate switched off, and the colour sampling changed from 4:4:4 to 4:2:2. Each change is tested on its own, and the result is recorded before the next change is made. If the picture becomes stable after one of those changes, the bandwidth layer is the working answer: the switch version is the ceiling, and the mode that failed exceeded it. If no mode is stable — if every resolution, every refresh rate, and every combination fails — the fault is not bandwidth alone, and the diagnosis moves back to the EDID, cable, or power layers. The downgrade comparison is a test, not a repair. Lowering the mode to make the picture work proves where the ceiling sits. It does not repair the ceiling, and it is not a suggestion to use the lower mode permanently.
The third fork is the switching-action comparison. The question is whether the failure appears only at the moment of switching, or only after switching. The test is to switch to another computer and then switch back, and to record exactly when the failure appears: at the moment of switching away, at the moment of switching back, or only after the switch has settled. A brief black screen at the moment of switching, with the picture returning after a few seconds, is a re-handshake: the system is re-enumerating the display and re-establishing the mode, and the black screen is part of that process. A resolution reset, windows moved, or a refresh rate fallen back after the switch has settled is the EDID emulation layer: the computer believed the monitor had been disconnected, and it rearranged the desktop accordingly. A long black screen after switching, with no return, is the link or bandwidth layer. The switching action itself is the variable, not the background. Two identical setups can behave differently at the moment of switching, and the difference is the evidence.
The fourth fork is the content comparison. Ordinary content — the desktop, a game, an ordinary video — is played through the switch, and then protected content — a streaming service playing HDCP-protected video — is played through the same chain, on the same machine. If ordinary content is normal and protected content is black, with sound but no picture, the HDCP relay layer is the working answer. If ordinary content is black as well, the HDCP layer is not the answer, and the diagnosis returns to the bandwidth, EDID, or cable layers. This fork uses the HDCP mechanism from its own guide as one comparison among four, without rewriting the HDCP guide. The content comparison is clean because it separates one kind of content from another on the same chain, and the separation is the evidence.
| Fork | Comparison made | Result pointing at the switch layer | Result pointing elsewhere |
|---|---|---|---|
| Direct connection | Same computer, cable, monitor, and mode, switch bypassed | Direct works, through the switch fails | Direct fails as well — the switch is not the problem |
| Downgrade | One variable lowered at a time through the switch | Picture becomes stable after one downgrade | No mode is stable — not bandwidth alone |
| Switching action | Switch away and switch back, record when the failure appears | Failure follows the switch: reset or moved windows after switching | Failure appears without any switching — the switch action is not the variable |
| Content | Ordinary content against protected content on the same chain | Ordinary normal, protected black with sound | Ordinary content black as well — not the HDCP layer |
Why a Brief Black Screen at Switching Is Not Always a Fault
A switch that goes black for a few seconds at the moment of switching, and then returns with a normal picture, is often behaving normally. The switch has to re-establish the link: the computer has to re-enumerate the display, the EDID has to be read again, the mode has to be negotiated again, and the HDCP handshake has to be run again. That process takes time, and during that time there is no picture. A few seconds of black, followed by a full return to the correct resolution, the correct refresh rate, and the windows in their places, is a re-handshake. It is not a failure.
The boundary is duration and recovery, not whether the screen went black at all. A brief black screen that returns on its own is a handshake. A long black screen that does not return, a screen that returns only after the cable is unplugged and plugged back in, a resolution that returns reset, or windows that return moved, is a fault. The distinction has to be written plainly, because the two patterns look identical for the first few seconds and only separate when the picture either returns or does not. A reader who treats every black screen at switching as a fault will chase a problem that does not exist. A reader who treats every black screen as normal will miss a fault that does exist. The test is to watch what happens after the black screen, and for how long, and to record it.
Vendor marketing has to be read with the same caution. One switch maker's product page describes switching in one to three seconds with no black screen, and presents that as a selling point. That description is a vendor claim about one product, not a universal guarantee about all switches. A switch that takes a few seconds and shows a brief black screen is not necessarily faulty, and a switch that is advertised as having no black screen is not necessarily free of one in every setup. The sources for this guide state the distinction directly: a short black screen can be a normal re-handshake, while a long black screen, a fall back to 60Hz, or a picture that only returns after the cable is re-plugged is not. The vendor claim is quoted here as a vendor claim, and it is not presented as a standard that every switch must meet.
The Five Suspicion Layers
A failure through a standalone switch, where the direct connection works, always lands in one of the five layers below. They are listed by the shape of the evidence, yet the listing order is not proof: a layer is named only after a test confirms it or excludes it. Which layer is most often responsible is not settled by the sources, and this guide invents no ranking. Each layer is given with its own evidence, and the tests decide.
| Layer | What it means | Evidence that points at it |
|---|---|---|
| First layer: EDID and HPD layer | Switch has no EDID emulation, the emulated EDID does not match, or the HPD signal drops at switching | Resolution resets, windows move, or refresh rate falls back after switching; the switch manual shows no emulation or learning function; testing each input port in turn shows different results |
| Second layer: bandwidth layer | Switch HDMI or DisplayPort version ceiling, cable certification and length, or the mode combination of resolution, refresh rate, colour depth, HDR, and variable refresh rate | Downgrade comparison restores the picture; 4K 60Hz works and 4K 120Hz does not; a certified short cable brings an improvement |
| Third layer: HDCP relay layer | Switch HDCP version, or the contrast between protected content and ordinary content | Only protected content is black while ordinary content is normal; protected content is normal on a direct connection |
| Fourth layer: switch power and firmware layer | Powered switch with an unstable power supply, firmware version, or the switching button and hotkey setting | Intermittent disconnection when the power is unstable; an improvement after a firmware check carried out as the manual describes; the fault follows one switching method and not another |
| Fifth layer: cable and connector layer | Cable certification (Ultra High Speed HDMI or VESA-certified DisplayPort cable), length, a loose connector, or a DisplayPort pin 20 fault | A cable-swap comparison shows the fault following the cable; an expensive cable that does not match the specification does not help; a short certified cable restores the picture |
The first layer is the EDID and HPD layer. The switch either has no EDID emulation, or its emulated EDID does not match the monitor, or the HPD signal drops when the switch moves. The evidence is the pattern after switching: the resolution resets, the windows move, or the refresh rate falls back. The test is to check the switch manual for an emulation or learning function, and to run the learning or reset procedure exactly as the manual describes for that model — never by copying the reset steps of a different model. Each input port is tested in turn, because one port behaving badly does not prove that every port behaves the same way. If the problem disappears when emulation is present and correctly learned, the EDID layer is confirmed. If the problem continues with emulation present, the EDID layer is excluded, and the diagnosis moves on. Emulation is not a guarantee. It solves the disappearance layer, and it solves nothing else.
The second layer is the bandwidth layer. The switch carries an HDMI or DisplayPort version ceiling, the cable carries a certification and a length, and the mode itself piles several demands together: resolution and refresh rate, colour depth and HDR, and whether variable refresh rate is switched on. The evidence is the downgrade comparison: the picture becomes stable after one variable is lowered, or 4K 60Hz works while 4K 120Hz does not, or a certified short cable brings an improvement. The test lowers one variable at a time and records which variable restores the picture. A switch that passes 1080p at 60Hz but not 4K at 144Hz is not broken. It is a switch whose ceiling sits between those two modes, and the ceiling is the diagnosis. The phrase "supports 4K" without a refresh-rate number has to be checked against the specification of the particular model, because the claim can hide a 30Hz or 60Hz ceiling — but the same caution runs the other way, and no switch should be assumed to have a low ceiling only because its headline claim omits the refresh rate. The specification is the evidence.
The third layer is the HDCP relay layer. The switch has an HDCP version, and the player, the card, the switch, and the monitor all have to support the same version before protected content can play. The evidence is the content comparison: only protected content is black, while ordinary content is normal, and protected content is normal on a direct connection. The test is to play both kinds of content through the same chain and to watch which one fails. The HDCP conclusion must never be applied to a general black screen, because a general black screen affects ordinary content too, and the HDCP layer cannot explain it. This layer uses the HDCP mechanism from its own guide as one comparison among four, and it does not rewrite that guide's handshake diagnosis.
The fourth layer is the switch power and firmware layer. An active switch needs power — an external power adapter or USB power — and an unstable supply can cause intermittent disconnection. The firmware version can matter too, and the switching button or hotkey setting can be misconfigured. The evidence is intermittency that follows the power supply, or an improvement after a firmware check carried out as the manual describes. Firmware is handled with a strict limit in this guide: the manual is consulted, and the firmware is checked only as the manual describes. The reader is not instructed to flash firmware freely, and no firmware version is presented as a solution. A firmware update carried out without the manual is a risk, not a test, and this guide does not present it as one.
The fifth layer is the cable and connector layer. The cable has a certification — Ultra High Speed HDMI for HDMI 2.1, or a VESA-certified cable for DisplayPort — and a length, and both matter. High-bandwidth modes need a short cable: an 8K-capable switch, for example, is specified by its maker to need a certified cable within two metres. A loose connector, or a DisplayPort cable with a pin 20 fault that feeds power back into the chain, can produce the same intermittent failure. The evidence is a cable-swap comparison: the fault follows the cable, or it does not. An expensive cable is not evidence. A cable that matches the specification is evidence, and the test is whether the fault follows the cable when the cable is changed. The price of the cable proves nothing about whether it matches the mode being sent.
The Seven Diagnostic Steps
Follow the seven steps in the order given. Each step ends with a result that decides the next step. Jumping to a cable purchase, or to a firmware update, before the first steps are done usually wastes the attempt: without a baseline, a direct comparison, and a downgrade comparison, a later change proves nothing, even when the picture returns.
| Step | Action | What the result means |
|---|---|---|
| Step 1 | Record the baseline: the native resolution, refresh rate, HDR, variable refresh rate, and colour sampling on a direct connection; the switch model, HDMI or DisplayPort version, whether EDID emulation is present, and the power supply method | Without a baseline, no later comparison means anything |
| Step 2 | Run the direct comparison: same computer, same cable, same monitor, same mode, switch bypassed | Direct normal means the switch layer — continue; direct fails as well means the switch is not the problem — stop and change direction |
| Step 3 | Run the downgrade comparison through the switch, one variable at a time: resolution, then refresh rate, then HDR off, then variable refresh rate off, then colour sampling from 4:4:4 to 4:2:2 | Picture stable after one downgrade means the bandwidth layer; no mode stable means the EDID or cable layers |
| Step 4 | Run the switching-action comparison: switch to another computer and switch back, record exactly when the failure appears; test each input port in turn | Brief black screen returning in seconds is a re-handshake; resolution reset or windows moved after switching is the EDID layer; long black screen is the link or bandwidth layer |
| Step 5 | Run the EDID emulation comparison: check the switch manual for an emulation or learning function, and carry out one learning or reset exactly as the manual describes for that model | Problem gone with emulation means the EDID layer; problem unchanged with emulation means the EDID layer is excluded |
| Step 6 | Run the content comparison: same machine, same chain, play ordinary content and then protected content | Ordinary normal and protected black means the HDCP relay layer; ordinary black as well means the HDCP layer is excluded |
| Step 7 | Run the cable, power, and firmware comparison: change to a certified short cable (Ultra High Speed HDMI within two metres for HDMI 2.1, VESA-certified cable for DisplayPort), check that the switch power supply is stable, and check the firmware version as the manual describes | Picture normal after a cable change means the cable layer; power and firmware are checked only as the manual describes, never by flashing freely |
Step one is the groundwork for all seven steps, and it is done before any test is run. Record the baseline on a direct connection: the native resolution, the refresh rate, HDR on or off, variable refresh rate on or off, and the colour sampling. Then record the switch itself: the model, the HDMI or DisplayPort version, whether EDID emulation is present, and how the switch is powered. Without that baseline, no later comparison means anything, because there is no known-good state to compare the switched state against. That baseline is the yardstick for every step after it, and it has to be written down before the switch goes back into the chain.
Step two is the direct comparison, and it is the threshold for the whole guide. The same computer, the same cable, the same monitor, and the same display mode are connected directly, with the switch bypassed entirely. If the direct connection is normal, the switch layer is the working answer, and the diagnosis continues. If the direct connection fails as well, the switch is not the problem, and the diagnosis must stop circling the switch and change direction. The direct comparison has to be run before any other test, and its result decides whether the rest of the guide applies at all. A reader who skips it is diagnosing a device that may never have been at fault.
Step three is the downgrade comparison, and it has to be run one variable at a time. Through the switch, the resolution is lowered first, then the refresh rate, then HDR is switched off, then variable refresh rate is switched off, then the colour sampling is changed from 4:4:4 to 4:2:2. Only one variable is changed at a time, and the result is recorded before the next change is made. If the picture becomes stable after one of those changes, the bandwidth layer is confirmed, and the variable that restored the picture identifies the ceiling. If no mode is stable — if every resolution, every refresh rate, and every combination fails — the fault is not bandwidth alone, and the diagnosis moves back to the EDID or cable layers. The downgrade is a comparison test only. It is not a suggestion to use the lower mode permanently, and it is not a repair.
Step four is the switching-action comparison. The switch is moved to another computer and then moved back, and the exact moment of the failure is recorded: at the moment of switching away, at the moment of switching back, or only after the switch has settled. A brief black screen that returns within a few seconds is a re-handshake. A resolution reset, windows moved, or a refresh rate fallen back after the switch has settled is the EDID emulation layer. A long black screen after switching is the link or bandwidth layer. Each input port is tested in turn, because one port behaving badly does not prove that every port behaves the same way, and the fault may follow one port rather than the switch as a whole. The switching action itself is the variable, and the test has to treat it as one.
Step five is the EDID emulation comparison. The switch manual is checked for an emulation or learning function, and for whether the emulated EDID is learned from the monitor or supplied as a fixed file. One learning or reset is carried out exactly as the manual describes for that model, and never by copying the reset steps of a different model. If the problem disappears when emulation is present and correctly learned, the EDID layer is confirmed. If the problem continues with emulation present, the EDID layer is excluded. The manual is the authority for this step, because the learning procedure differs between models, and a procedure copied from another model can make the problem worse rather than better.
Step six is the content comparison. On the same machine, through the same chain, ordinary content is played first and protected content second. Ordinary content normal and protected content black, with sound but no picture, is the HDCP relay layer. Ordinary content black as well is not the HDCP layer, and the diagnosis returns to the bandwidth, EDID, or cable layers. The content comparison is clean because it changes only the kind of content, on the same chain, and the difference in the result is the evidence. The HDCP conclusion is applied only to the protected-content pattern, and never to a general black screen.
Step seven is the cable, power, and firmware comparison. The cable is changed to a certified short cable — Ultra High Speed HDMI within two metres for HDMI 2.1, or a VESA-certified cable for DisplayPort — and the picture is tested again. The switch power supply is checked for stability, and the firmware version is checked as the manual describes. If the picture becomes normal after the cable change, the cable layer is confirmed, and the fault followed the cable. Power and firmware are checked only as the manual describes. Firmware is never flashed freely, and no firmware version is presented as a solution. After every comparison, restore the settings that were changed for it, where that is possible, and record the outcome. A comparison with no record behind it cannot be used, because the step after it needs to know what changed and what the result was. The monitor's own sleep, deep sleep, and automatic input detection settings are worth checking as the monitor manual describes; they are mentioned here only and not expanded.
Six Misconceptions That Mislead This Diagnosis
Misconception one: a more expensive cable will solve it. This is wrong. The cable is only one of five layers, and price is not evidence. A cable has to match the specification — the certification, the version, and the length for the mode being sent — and an expensive cable that does not match the specification will not help. The cable test is a swap comparison: the fault follows the cable, or it does not. Paying more for a cable and counting that purchase as the repair bypasses the comparison that would have shown whether the cable was involved at all. A certified short cable serves as a way to rule the cable layer out. It is not a cure for every fault.
Misconception two: a switch with EDID emulation is guaranteed to work. This is wrong. Emulation only makes the computer believe that the monitor is always present, so that the computer never re-enumerates the display and never rearranges the desktop. It does not add bandwidth, it does not repair a faulty cable, and it does not let an older switch support a higher resolution than its hardware allows. A switch that supports only HDMI 2.0 remains an HDMI 2.0 switch when emulation is added, and a 4K 120Hz mode still cannot pass through it. Emulation solves the disappearance layer. It solves nothing else, and a switch with emulation can still fail on bandwidth, on HDCP relay, on power, or on the cable.
Misconception three: "supports 4K" means the switch supports 4K at a high refresh rate. This is wrong. A 4K claim with no refresh-rate number beside it can hide a 30Hz or 60Hz ceiling, and the mode that fails may be the refresh rate rather than the resolution. The ceiling has to be checked against the specification of the particular model: the HDMI or DisplayPort version, the bandwidth figure, the refresh rate at each resolution, and whether HDR and variable refresh rate are passed through. At the same time, the opposite error has to be avoided: a switch without a refresh-rate figure in its headline claim is not necessarily a low-ceiling switch. The specification is the evidence, and no general claim about all switches can replace it.
Misconception four: a direct connection that works proves the graphics card or the monitor is broken. This is wrong, and the direction is reversed. A direct connection that works proves that the card and the monitor are fine: the card can output the signal, the cable can carry it, and the monitor can display it. The failure appears only when the switch is placed in the middle, so the problem sits in the middle — in the switch, or in the switch-and-cable combination. Blaming the card or the monitor after a direct connection has already proved them working is a category error, and it sends the reader towards replacing hardware that has already been shown to work.
Misconception five: a black screen at the moment of switching is always a fault. This is wrong. A brief black screen at switching can be a normal re-handshake: the system re-enumerates the display, reads the EDID again, renegotiates the mode, and runs the HDCP handshake again, and that process takes a few seconds. The boundary is duration and recovery: a black screen that returns on its own within a few seconds, with the correct resolution, the correct refresh rate, and the windows in their places, is a handshake. A long black screen, a resolution that returns reset, windows that return moved, or a picture that returns only after the cable is re-plugged, is a fault. The test is to watch what happens after the black screen, not whether the screen went black at all.
Misconception six: a switch is the same as an adapter. This is wrong. An adapter converts one protocol to another in a static connection that never switches. A switch selects between several sources, and the switching action itself is a variable: the EDID handling at the moment of switching, the HPD signal at the moment of switching, and the re-handshake after switching are all part of the diagnosis. The two devices have different key variables, different failure patterns, and different tests, and the diagnostic logic of one cannot be applied to the other. The boundary sources state the separation plainly, and this guide keeps it.
Comparison Table: What Each Result Points At
The comparisons are collected in the table below. A row begins with what the reader can actually see — a direct-connection result, a downgrade result, a switching result, a content result — and finishes with the layer it points to. Use it once the seven steps are done, as a way of checking the diagnosis, not as a shortcut instead of the steps.
| Result seen during diagnosis | Layer it points at | What to do next |
|---|---|---|
| Direct works, through the switch fails | Switch layer, or switch-and-cable combination layer | Continue with the downgrade and switching-action comparisons |
| Direct fails as well | Not the switch layer | Stop circling the switch; change direction |
| Picture stable after one downgrade (resolution, refresh rate, HDR, variable refresh rate, or colour sampling) | Second layer: bandwidth layer | Note which variable restored the picture; the switch version is the ceiling |
| No mode stable through the switch, at any resolution or refresh rate | Not bandwidth alone | Move back to the EDID, cable, or power layers |
| Resolution reset, windows moved, or refresh rate fallen back after switching | First layer: EDID and HPD layer | Check the manual for emulation or learning; test each input port in turn |
| Brief black screen at switching, picture returning in seconds with the correct mode | Re-handshake, not a fault | Record the duration and the recovery; do not treat it as a failure |
| Ordinary content normal, protected content black with sound | Third layer: HDCP relay layer | Note the content contrast; the relay is the layer |
| Fault follows the cable when the cable is changed | Fifth layer: cable and connector layer | The cable is the faulty link identified by the swap comparison |
| Intermittent disconnection, fault following the power supply | Fourth layer: switch power and firmware layer | Check the power supply and the firmware version as the manual describes |
The final three rows deserve a close read, because they state plainly what each comparison establishes and what it leaves unresolved. A fault that follows the cable proves the cable layer, because only the cable changed and the result changed with it. A brief black screen that returns on its own proves a re-handshake, because the recovery is part of the result and not an afterthought. An intermittent disconnection that follows the power supply proves the power layer, because the fault appears and disappears with the supply. Every one of those rows is an answer in itself, and each one leads to a different next step.
What the Sources Disagree About
The sources behind this guide pull in different directions, and those differences have to be written out in full rather than smoothed away. The first disagreement is about whether EDID emulation is essential or auxiliary. Switch makers' manuals present EDID emulation as a standard feature that prevents handshake delay, and they describe it as part of the product. Independent guides describe emulation as solving one layer only: it does not add bandwidth, it does not repair a faulty cable, and it cannot add a resolution the switch does not support. The two positions are not contradicting each other once the layers are separated. Emulation solves the "monitor disappeared" layer. Bandwidth solves the "mode cannot pass" layer. The two layers are different, and a switch can need one, or the other, or both. This guide writes the two layers separately, and it does not present emulation as a universal answer.
The second disagreement is about a black screen at switching: normal or fault. One independent guide states that a short black screen can be a normal re-handshake. A switch maker's product page presents "no black screen" as a selling point, and describes switching in one to three seconds. The boundary is duration and recovery: a black screen that returns within a few seconds is a handshake process, while a long black screen, or a picture that returns only after the cable is re-plugged, is a fault. This guide does not present the vendor claim as a standard that every switch must meet. The claim is a vendor claim about one product, and it is quoted here as one, not as a universal guarantee.
The third disagreement concerns which cause is most often responsible. One guide puts EDID at the centre. Another lists the signal path, bandwidth, wake behaviour, and EDID as four parallel categories without ranking them. A third puts EDID and bandwidth side by side. The sources do not settle the question, and this guide invents no answer. It lays out five layers, each carrying its own evidence, and it never names one of them the most frequent cause. Someone told that one cause is the most frequent will look at that cause first and may stop there; someone given five layers and five tests will look at each one and let the evidence decide.
The fourth disagreement is about how to read a "supports 4K" claim. One published guide warns that a 4K claim without a refresh-rate figure should be treated with suspicion, because the claim most often hides a 30Hz or 60Hz ceiling. The same guidance also warns that not every switch without a refresh-rate figure is necessarily a low-ceiling switch, and that the specification of the particular model has to be checked. This guide follows both halves of that warning: a headline claim without a refresh-rate number is a reason to check the specification, and it is not, on its own, proof of a low ceiling. The specification — the version, the bandwidth figure, the refresh rate at each resolution — is the evidence, and the headline claim is not.
Boundary: Shapes This Guide Does Not Cover
The boundary table below gathers the lookalikes: shapes that could be mistaken for this one but belong with other guides. Each row states the lookalike, the guide it belongs to, and the test that tells it apart. Use the table as a check on the diagnosis itself: when what the reader sees matches a row here more closely than it matches the opening pattern, the reader has the wrong guide and should move to the right one.
| Shape | Which guide it belongs with | Why it is not this guide |
|---|---|---|
| Monitor's built-in KVM or USB hub switching fails | #236, monitor built-in KVM | The switch is inside the monitor and controlled by the monitor menu, not a standalone box |
| Failure through an adapter or adapter cable, direct works | #342, adapters | The device converts a protocol in a static connection; there is no switching action |
| 4K 144Hz cannot be reached, or the resolution and refresh rate are locked at 60Hz on a direct connection | #110, bandwidth, EDID, and DSC on a direct connection | This guide uses only the downgrade comparison; the direct-connection bandwidth subject belongs there |
| Windows move after the monitor wakes or reconnects, with no switch in the chain | #267, window arrangement | The subject is arrangement, and its boundary states that the signal layer is outside its scope |
| Protected video black with sound, on a direct connection | #438, HDCP handshake | This guide uses only the content comparison as its fourth fork; the HDCP subject belongs there |
| Every switch-on needs the DisplayPort cable unplugged and plugged back in before a picture appears | #471, cold-start handshake | The trigger is switching on, not the switching action between computers |
| Clone mode to a television or projector pulls the main screen down | #269, clone mode | Clone mode shares a mode intersection between two displays; no switch is involved |
| Dock, MST hub, or DisplayLink dock fails | #110 boundary, docks and hubs | The diagnostic logic is different: shared bandwidth, or USB compressed transport, not a standalone switch |
| Monitor sleep, deep sleep, or automatic input detection | Monitor settings layer | This guide mentions it only in the seventh step and does not expand it |
Two rows in that table deserve a closer look, because they are the ones readers most often mistake for this pattern. The first is the monitor's built-in KVM. A built-in KVM is controlled by the monitor menu, and its switching is part of the monitor's own behaviour. A standalone switch is a separate box with its own inputs, its own button or hotkey, and its own specification. The two devices look similar — both let one monitor serve two computers — but their controls, their EDID handling, and their failure patterns are different, and their guides are different. The second is the adapter. An adapter converts one protocol to another and never switches. A switch selects between sources and switches. The adapter's key variables are passive against active conversion and protocol. The switch's key variables are the switching action, EDID emulation, and the bandwidth ceiling. Treating a switch as an adapter, or an adapter as a switch, sends the reader to the wrong set of tests.
When to Stop Testing
Stop once the seven steps have named no layer. A switch that has been recorded at baseline, tested on a direct connection, run through a downgrade comparison one variable at a time, tested on its switching action and on each input port in turn, checked for EDID emulation as its manual describes, tested on ordinary against protected content, and run through a cable, power, and firmware comparison — and still returns no clear answer — has told the reader everything the home tests can tell. More attempts after that are guesswork, and guesswork with settings that were never wrong makes the diagnosis harder, not easier. The settings changed for comparisons should be put back before stopping, so that the chain is left in a known state.
Stop when the direct comparison fails as well. That result is an answer in itself: the switch is not the problem, and continuing to test the switch wastes the evidence the direct comparison has already given. Stop also when a cable change, a downgrade, and an emulation check have each been run once with the result unmoved, and one more variable has been altered with the result still unmoved. When two changed variables leave the result unchanged, that is an answer: the layer is not one those variables control. Changing a third, fourth, and fifth setting one after another adds no information. It only builds a list of changes that have to be reversed before any inspection can start from a known state.
Stop at once, and run no further switching tests, if the switch shows a fresh symptom while testing: a burning smell, a power adapter that is hot to the touch, a connector that is loose or damaged, or a monitor that shows a distorted picture that was not there before. Those symptoms belong with another family of problems, and going on switching while they are present can turn a diagnostic question into a hardware fault. Record what showed up, when it showed up, and what was being tested then, and bring that record along. That note is part of the evidence, and it carries more weight than a list of settings that were changed.
Do not throw the baseline notes away when testing stops. The native resolution, refresh rate, HDR, variable refresh rate, and colour sampling on a direct connection, the switch model and version, the result of the direct comparison, the variable that restored the picture in the downgrade comparison, the moment at which the failure appeared in the switching-action comparison, the emulation result, the content comparison result, and the cable result, together say more than any one of them alone. An inspection starts from those notes — the reason the first step insists on writing the baseline down before any test is run.
What to Bring for Inspection
Bring the notes from the seven steps: the baseline on a direct connection, including the native resolution, refresh rate, HDR, variable refresh rate, and colour sampling; the switch model, its HDMI or DisplayPort version, whether EDID emulation is present, and how it is powered; the result of the direct comparison; the downgrade comparison result, including which variable, when lowered, restored the picture; the switching-action result, including the exact moment the failure appeared and whether it followed one input port or all of them; the EDID emulation result, including what the manual says about emulation or learning and what happened when the procedure was carried out; the content comparison result, including whether ordinary content was normal while protected content was black; and the cable, power, and firmware results, including which cable was tried and whether the fault followed it.
Bring the switch itself, if it can be brought, together with the cable that was in use, because the switch and the cable are the chain, and an inspection that has only one of them has only half the evidence. Bring the monitor model and the graphics card model and driver version, because the combination is part of the evidence, and the mode that failed has to be checked against the specification of each link in the chain. Bring the settings in their current state, after every comparison setting has been restored, and leave out any list of every setting that was changed and changed back. The evidence is the current settings together with the seven comparison results. The history of changes is useful only where a change restored the picture, because that change identifies the layer.
For inspection, HK GPU Fix is at Dali Plaza 1209, Shek Mun. The phone number is 2186 8010 and WhatsApp is 6558 6806. The published inspection fee range is HK$400 to HK$2,800, and no price is estimated before inspection. There is no Mong Kok shop and no on-site service.
Frequently Asked Questions
My computer works fine plugged straight into the monitor. Why does it fail through the switch?
Because the switch changes what the computer sees, or what the signal has to pass through. A direct connection proves that the card can output the signal, the cable can carry it, and the monitor can display it. A switch in the middle can break that chain in several ways: it can fail to pass the monitor's EDID to the computer, so the computer believes the monitor has been disconnected; it can cap the bandwidth, so a high mode cannot pass through it; it can break the HDCP relay, so protected content fails; or it can be let down by its power supply, its firmware, or the cable. The direct comparison is the threshold test because it proves the two ends work and the middle does not.
The picture goes black for a few seconds when I switch, then comes back. Is the switch faulty?
Not necessarily. A brief black screen at switching can be a normal re-handshake: the system re-enumerates the display, reads the EDID again, renegotiates the mode, and runs the HDCP handshake again, and that process takes a few seconds. The boundary is duration and recovery: a black screen that returns on its own within a few seconds, with the correct resolution, the correct refresh rate, and the windows in their places, is a handshake. A long black screen, a resolution that returns reset, windows that return moved, or a picture that returns only after the cable is re-plugged, is a fault. Record how long the black screen lasts and what returns with it.
After switching back, my resolution has dropped and my windows have moved. What happened?
The computer believed the monitor had been disconnected while the switch was showing another computer, and it rearranged the desktop for a smaller display. When a switch offers no EDID emulation, switching away leaves the computer with no EDID and no hot-plug signal, so the system treats the monitor as removed: the desktop shrinks, windows are moved back to the primary monitor, and the resolution and refresh rate are reset. Switching back requires the system to enumerate the monitor again, and the windows do not return to their previous positions automatically. The test is to check the switch manual for an EDID emulation or learning function and to carry out one learning or reset exactly as the manual describes for that model.
The switch says it supports 4K, but 4K at 120Hz does not work through it. Why?
Because "supports 4K" on its own says nothing about the refresh rate. A switch described only as supporting 4K may support 4K at 30Hz or at 60Hz and no higher, and a 4K 120Hz mode with HDR and variable refresh rate needs far more bandwidth than a 4K 60Hz mode. An HDMI 2.0 link carries up to 18 Gbps, and a 4K 60Hz signal at 4:4:4 in 8-bit already sits close to that ceiling, while an HDMI 2.1 link at 40 or 48 Gbps is what a 4K 120Hz mode requires. Check the specification of the particular model — the version, the bandwidth figure, and the refresh rate at each resolution — rather than the headline claim. The downgrade comparison, lowering one variable at a time, shows which variable pushes the chain past its ceiling.
Ordinary videos play fine through the switch, but protected streaming content is black. What does that mean?
It points at the HDCP relay layer. Protected content needs the player, the graphics card, the switch, and the monitor all to support the same HDCP version, and a switch that does not support that version, or whose version does not match, will show protected content as black, with sound but no picture, while ordinary content plays normally. The test is the content comparison: play ordinary content and then protected content through the same chain, on the same machine. Ordinary normal and protected black is the HDCP relay layer. Ordinary content black as well is not the HDCP layer, and the diagnosis has to return to the bandwidth, EDID, or cable layers.
Will a more expensive cable fix the problem through the switch?
Not necessarily, and price is not evidence. The cable is only one of five layers, and a cable has to match the specification — the certification, the version, and the length for the mode being sent. An expensive cable that does not match the specification will not help, and a certified short cable is a tool for excluding the cable layer, not a universal repair. The test is a swap comparison: change the cable, keep everything else the same, and watch whether the fault follows the cable. For HDMI 2.1, the maker's specification is an Ultra High Speed certified cable within two metres. If the fault stays after the cable is changed, the cable was not the layer.
My switch has EDID emulation. Does that mean it cannot be the problem?
No. EDID emulation only makes the computer believe that the monitor is always present, so that the computer never re-enumerates the display and never rearranges the desktop. It does not add bandwidth, it does not repair a faulty cable, and it does not let an older switch support a higher resolution than its hardware allows. A switch that supports only HDMI 2.0 remains an HDMI 2.0 switch when emulation is added, and a 4K 120Hz mode still cannot pass through it. If the problem continues with emulation present, the EDID layer is excluded, and the diagnosis moves to the bandwidth, HDCP relay, power, or cable layers.
Is a KVM switch the same thing as an adapter?
No. An adapter converts one protocol to another — DisplayPort to HDMI, for example — in a static connection that never switches. A switch selects between several sources and shows one at a time, and the switching action itself is a variable: the EDID handling at the moment of switching, the hot-plug signal at the moment of switching, and the re-handshake after switching are all part of the diagnosis. The two devices have different key variables, different failure patterns, and different tests, and the diagnostic logic of one cannot be applied to the other. A dock or a DisplayLink device is different again, because it shares bandwidth between monitors or sends its picture over USB as compressed data.
Should I update the switch firmware to fix the problem?
Firmware should be checked only as the switch manual describes, and the reader is not instructed to flash firmware freely in this guide. An unstable power supply, a firmware version, or a switching button or hotkey setting can all sit in the power and firmware layer, and the evidence is intermittency that follows the power supply or an improvement after a firmware check carried out as the manual describes. A firmware update carried out without the manual is a risk, not a test. Check the manual first, note the firmware version, and treat the firmware as one comparison among seven, not as the first step.
When should I stop testing and bring the switch in for inspection?
Stop when the seven steps have given no clear layer, when the direct comparison fails as well, or when a new symptom — a burning smell, a hot power adapter, a loose or damaged connector, or a distorted picture — appears during testing. Bring the notes from the seven steps: the baseline on a direct connection, the switch model and version, the direct-comparison result, the downgrade result and which variable restored the picture, the switching-action result, the emulation result, the content result, and the cable, power, and firmware results — together with the switch itself and the cable that was in use, the monitor model, and the graphics card model and driver version. Bring those notes, and the settings as they stand after every comparison setting has been put back — that is where an inspection starts.
Sources
The sources below were consulted for this guide. Vendor manuals and product pages are cited as sources for the specification and the emulation mechanism, not as recommendations, and vendor marketing claims are quoted as vendor claims where the text says so.
- laptopjudge.com, "Kvm Switch Resolution Issues: Prevent Display Drops" — https://laptopjudge.com/kvm-switch-resolution-issues-prevent-display-drops/ — an EDID handshake that disappears at switching drops the resolution back; an EDID emulator gives each KVM port a stable display identity; the diagnosis runs direct connection first, then each port in turn, then an EDID emulator, then an EDID checksum check; HDMI 2.0 carries up to 18 Gbps, and 4K 60Hz at 4:4:4 in 8-bit already sits close to that ceiling; a passive adapter is not an emulator.
- tesmart.com, "EDID Emulation Explained" — https://www.tesmart.com/blogs/news/edid-emulation-explained-how-it-makes-kvm-multi-monitor-switching-more-stable — a KVM that handles EDID badly produces a long black screen, windows moved, the monitor re-detected, and the refresh rate reset; the computer believes the monitor has been disconnected and re-enumerates it; EDID emulation does not make the monitor faster and does not improve picture quality — its work is only to make the computer keep recognising the monitor; for several monitors, emulation is the key to a stable desktop layout.
- laptopjudge.com, "Kvm Switch Display Output: Fix Blank Screen" — https://laptopjudge.com/kvm-switch-display-output-fix-blank-screen/ — a black screen after switching is most often an EDID or hot-plug detect break, not a broken monitor; EDID is the capability record and HPD tells the computer that a monitor is connected, and a change in either can make the computer abandon the monitor or choose an unsupported mode; a normal EDID with a black screen points at the link, bandwidth, or mode layer, while a missing or unreadable EDID points at the KVM EDID or HPD layer; a monitor that is detected but shows no picture is a different pattern from a monitor that has disappeared.
- soldermag.com, "KVM Switch Black Screen? Fix It in the Right Order" — https://www.soldermag.com/posts/kvm-switch-black-screen-fix — a short black screen at switching can be normal, while a long black screen, a fall back to 60Hz, or a picture that returns only after the cable is re-plugged is not; the diagnosis runs direct connection first, because a direct connection that fails means the switch is not the problem; a KVM can pass 1080p at 60Hz but not 4K at 144Hz, and resolution, refresh rate, colour depth, HDR, colour sampling, and variable refresh rate all consume bandwidth; EDID emulation cannot add bandwidth, cannot add an unsupported resolution, and cannot repair a faulty cable; the reset steps of a different model should not be copied.
- realbuyerexperiences.com, "KVM Switches, Explained" — https://realbuyerexperiences.com/explained/kvm-switches/ — EDID handling is the key to success, and a stable EDID stops black-screen flicker and window rearrangement; the connectors and the bandwidth have to match the equipment, with HDMI 2.1 or DisplayPort 1.4 for 4K at a high refresh rate; a "supports 4K" claim without a refresh-rate figure should be treated with suspicion, because it most often hides a 30Hz or 60Hz ceiling; variable refresh rate has to be confirmed as passed through by the switch, not assumed from the cable; a Level1Techs DisplayPort 1.4 KVM is given as an example of a dual-monitor DisplayPort solution, not as a recommendation.
- KanexPro, KVM-2X1HDMI8K product manual — https://www.kanexpro.com/uploads/KVM-2X1HDMI8K-manual.pdf — a 2-by-1 HDMI 2.1 KVM at 40 Gbps, with 8K at 60Hz, 4K at 120Hz and 144Hz, and 1080p at 240Hz, variable refresh rate and ALLM, HDCP 2.3, and HDR, HDR10, HDR10+, Dolby Vision, and HLG pass-through; both inputs have EDID emulation to prevent handshake delay at switching; CEC pass-through; 8K requires an Ultra High Speed certified cable within two metres.
- KanexPro, KVM-2X1DUAL8K product manual — https://www.kanexpro.com/uploads/KVM-2X1DUAL8K-manual.pdf — a dual-display HDMI 2.1 KVM with HDCP 2.3 and EDID emulation on every input; where a black screen continues, the manual's advice is to restart the power to the switch and to check the HDMI cable connection.
- TESmart, HKS801-M24 product page — https://www.tesmart.com/products/hks801-m24 — an HDMI 2.1 8-by-1 KVM at 48 Gbps, with an EDID emulator on every input, variable refresh rate and HDR, and HDCP 2.3; DSC is required for compressed modes such as 8K at 60Hz and 4K at 240Hz; the maker describes switching in one to three seconds with no black screen — quoted here as a vendor claim, not as a universal guarantee.
A computer that works directly and fails through a switch is not a verdict on the computer. Run the direct comparison first, lower one variable at a time, watch the switching action itself, and compare ordinary content against protected content. Those comparisons say which layer the switch is failing on — and they say it before any cable is bought or any firmware is touched.