Fix "Display Driver Stopped Responding and Recovered"

"Display driver stopped responding and has recovered" does not mean your graphics card is dying. It means Windows caught a driver that stopped responding and forcibly restarted it before the freeze could take down your whole PC — that is the entire message, and it is Windows doing its job, not failing at it. The screen blacks out or flickers for a second, sometimes with a small notification bubble, and then everything comes back like nothing happened. The counterintuitive part: this "error" is actually a safety feature called Timeout Detection and Recovery (TDR), whose entire purpose is to stop a stuck driver from freezing your whole session. In the huge majority of cases, the message you are staring at is proof the safety net worked — not proof your hardware is failing.

Jake found out the hard way. He runs a small phone and PC shop, and two weeks earlier he had sold a customer a rebuilt gaming desktop — nothing exotic, just a mid-range card he had cleaned up and re-sold with a fresh Windows install. The customer came back on a Saturday, annoyed, saying the screen kept going black for a second every time he loaded up a match, then came back with a little notification about a "display driver" doing something. Jake had no idea what that meant, and he was not about to tell a paying customer "I don't know" twice in one shop visit. He called Ethan that night. "Is the card dead?" he asked, half-expecting to eat the cost of a replacement. Ethan's answer was immediate: "Almost certainly not, and don't offer him money back yet. This message is Windows bragging that it caught a problem, not confessing to one."

That is the position most people are in when this message shows up: nothing was obviously wrong five minutes ago, and now the screen keeps flickering mid-game or mid-render. The good news is that this is one of the most diagnosable errors in Windows, because the causes are a short, well-known list, and you can usually work through them in under twenty minutes.

What's actually happening when you see this

Windows expects your graphics driver to keep responding to the commands it sends. When it does not — because a game demanded too much, a driver has a bug, or the GPU itself is struggling — Windows assumes the driver has hung, resets it, and puts your display back together. That reset is the black flicker. The notification bubble that says "Display driver stopped responding and has recovered" is just Windows telling you, after the fact, what it had to do.

The message alone does not tell you why it happened. That is the part that actually matters, and it usually comes down to one of six things: an outdated or corrupted driver, a power plan that is throttling your GPU during heavy tasks, a driver timeout that is too short for what you are asking the GPU to do, an unstable overclock or undervolt, a GPU running hotter than it should, or — less often, but it happens — a graphics card that is genuinely starting to fail. The methods below are ordered the way you should actually try them: cheapest and most likely first, hardware-replacement last.

Fixing it on Windows 11

These steps are identical on Windows 10 as well — the menus look slightly different, which is covered in its own section further down, but the actions themselves do not change. Work through them in order.

Method 1: Update or reinstall the graphics driver

This fixes the largest share of cases, because a good chunk of TDR events trace back to a driver that has a known bug already patched by the manufacturer, or a driver file that got corrupted during a previous update.

Press Win + R to open the Run box — a small dialog that lets you launch a program or system tool directly by its internal name, without hunting through menus. Type devmgmt.msc and press Enter. The ".msc" extension means it is a Microsoft Management Console snap-in, a small administrative tool; devmgmt.msc specifically opens Device Manager, the list of every piece of hardware Windows knows about.

Expand Display adapters, right-click your graphics card, and choose Update driver. Let Windows search automatically, then restart once it finishes. If Windows reports the driver is already up to date — which happens often, because Windows Update tends to lag behind the manufacturer's own releases — go straight to the manufacturer's site (NVIDIA, AMD, or Intel, depending on your card) and download the latest driver package directly. A manufacturer's own installer usually offers a "clean install" checkbox during setup; take it, since it removes leftover settings from the previous driver that a simple overwrite would keep.

Method 2: Roll back the driver instead

If the recoveries started right after a driver update — a new game patch triggered one, or Windows Update silently installed something — the newest driver may be the actual cause, and rolling back is the faster fix than trying yet another update.

Back in Device Manager, right-click your graphics card and choose Properties instead of Update driver. Open the Driver tab. If the Roll Back Driver button is available, click it, confirm the reason if Windows asks, and restart. Windows only keeps one previous driver in reserve, so this button is grayed out if there is nothing to roll back to, or if you have not updated the driver since your last clean install.

Method 3: A clean reinstall with Display Driver Uninstaller

A normal "Update driver" or even the manufacturer's own installer does not always remove every trace of the previous driver — leftover registry entries and configuration files can sit there and quietly interfere with the new one. Display Driver Uninstaller (DDU) is a well-known third-party tool built specifically to strip out every fragment of the old driver before you put a fresh one in.

Boot into Safe Mode first, since DDU works best when the driver is not actively loaded — hold Shift while clicking Restart, then go to Troubleshoot → Advanced options → Startup Settings → Restart, and pick the Safe Mode option once the PC comes back up. Run DDU, select your GPU manufacturer, and let it fully clean the old driver. Restart normally, then install the latest driver from the manufacturer's site from scratch.

This is the method to reach for after Methods 1 and 2 have not helped, not the one to start with — it takes longer, and it is not necessary for the majority of people who see this error.

Method 4: Set your power plan to High performance

On the default Balanced power plan, Windows sometimes throttles back a discrete graphics card during a demanding task to save power, and the resulting stall can look exactly like a frozen driver from the GPU's point of view — enough for TDR to trigger a recovery that has nothing to do with the driver itself.

Open Settings → System → Power & battery, and under Power mode, choose Best performance. If you prefer the older Control Panel route, it is still there: Control Panel → Hardware and Sound → Power Options → Change plan settings → Change advanced power settings, then find your graphics card's entry in the list and set it to Maximum performance instead of Balanced. Apply, then restart.

Method 5: Give the driver more time with TdrDelay

TdrDelay is the registry value that controls how many seconds Windows waits before deciding a driver has genuinely hung. Windows ships with a short factory timeout — short enough that a perfectly healthy driver can occasionally get flagged during a long, legitimate GPU task, such as heavy shader compilation in a new game, a large render job, or a demanding compute workload. Raising the value gives real work more room to finish before Windows assumes the worst.

Press Win + R, type regedit, and press Enter — this opens the Registry Editor, the tool Windows uses to store its own low-level settings. Navigate to:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers

Right-click an empty area in the right-hand pane, choose New → DWORD (32-bit) Value, and name it exactly TdrDelay. Double-click it, switch the base to Decimal, and set the value to 8 — that gives the driver eight full seconds before Windows decides it has hung, which is enough headroom for the vast majority of legitimate long-running GPU tasks without masking a genuinely frozen driver forever. Restart for it to take effect.

Method 6: Rule out an overclock, undervolt, or overheating GPU

If you or the PC's previous owner ever pushed the graphics card past its factory clock speeds, or applied an undervolt to run it cooler and quieter, that is worth suspecting before anything else on this list. Even a "stable" overclock that passed a stress test can misbehave under the specific, unusual load a particular game or app creates. Open your GPU manufacturer's utility — NVIDIA App, AMD Software, or Intel Graphics Software — and reset every clock, memory, and voltage setting back to its factory default. If the recoveries stop, the overclock or undervolt was the cause.

Heat causes the same symptom through a different door. Open Task Manager (Ctrl + Shift + Esc), open the Performance tab, and click your GPU to watch its temperature live while you run whatever triggers the error. A card sitting in the high 80s Celsius or above under sustained load, dust-clogged fans, or a case that is hot to the touch all point toward thermal throttling rather than a software fault. Cleaning dust out of the case and fans, and making sure nothing is blocking intake vents, is free and worth doing before anything more involved.

Windows 11 traps that trip people up mid-fix

None of these change the fix itself, but they trip people up while they are trying to follow it. Device Manager, Run, and regedit all still open exactly the way they always have — that part has not moved. What has changed is the right-click menu: Windows 11's default context menu is condensed down to a short list of icons, and some of the options you might expect, depending on where you right-click, are tucked under Show more options at the bottom (or you can skip straight to the full menu with Shift+F10). If a menu you are looking for seems to have disappeared, that is almost always where it went.

File Explorer also changed its default landing page. It now opens to Home rather than the old Quick Access view, which matters if a driver installer you downloaded is not showing up where you expect — check the Downloads folder directly rather than hunting through Home.

Fixing it on Windows 10

Every method above works exactly the same way on Windows 10 — devmgmt.msc, regedit, and the TdrDelay value all live in the same places, and DDU works identically. The only real differences are cosmetic:

Device Manager opens the same way, but the right-click menu is the classic, uncondensed version — no "Show more options" hunting required. For the power plan, go to Control Panel → Hardware and Sound → Power Options → Change plan settings → Change advanced power settings, find your graphics card in the list, and set it to Maximum performance. Windows 10 does not have the newer Settings > Power & battery page that Windows 11 introduced, so Control Panel is the only route.

Which fix matches your symptom

If none of that fixed it

This is the point where the honest answer changes. If you have updated or rolled back the driver, tried a clean reinstall, set the power plan to maximum performance, ruled out an overclock, and confirmed temperatures are normal, and the recoveries are still happening, you are very likely looking at failing hardware rather than a software fix you have not tried yet.

A few patterns are worth naming specifically. If the recoveries happen constantly, even during light desktop use with nothing demanding running, suspect the GPU itself or its power delivery rather than anything driver-related. If they only happen under heavy load but happen every single time without fail, that also leans toward the card struggling rather than a driver bug, since a driver issue is usually inconsistent. If you have a second, known-good graphics card you can swap in and the recoveries stop, that confirms the original card. If you can put the suspect card into a different, known-good PC and it still recovers there, the card itself — not the rest of your system — is the fault.

Ethan put it more bluntly when we talked through this section: "People want there to be a magic setting for everything, and most of the time there is one — for this error there usually is. But if you've genuinely ruled out the driver, the power plan, the timeout, the overclock, and the heat, and it's still happening, stop looking for a seventh registry tweak. You're not going to find one. Test the hardware instead."

Laptops, multiple GPUs, and other edge cases

Laptops with two graphics chips

Most gaming and business laptops built in the last several years have two GPUs: a weaker integrated one built into the CPU for everyday tasks and battery life, and a stronger discrete one for demanding work. This error shows up more often on these machines, usually because a game or app got assigned to the wrong GPU, or Windows switched between the two mid-task. Open Settings → System → Display → Graphics, find the problem app in the list (add it manually if it is not there), and set it to use the High performance GPU explicitly rather than leaving it on "Let Windows decide."

Multiple monitors

Driving several displays, especially at high refresh rates or mixed resolutions, asks more of the GPU and the driver simultaneously than a single-monitor setup does. If the error only started after adding a second or third monitor, try dropping one display's refresh rate closer to the others, and confirm every cable is one that is rated for the resolution and refresh rate you are actually running — an older cable quietly bottlenecking a 144Hz signal is a surprisingly common, and surprisingly cheap, cause.

Remote Desktop and virtual machines

If you are seeing this inside a Remote Desktop session or a virtual machine rather than on the physical PC in front of you, the cause is usually different from everything above. Remote sessions and most VMs render through a software or virtualized graphics driver rather than talking to a physical GPU directly, so a recovery there more often points at the host machine's own driver, a resource cap on the virtual machine, or the remote connection itself rather than a piece of failing hardware.

Docking stations and external GPUs

If your setup runs through a USB-C dock or an external GPU enclosure rather than plugging straight into the graphics card, add that connection to your suspect list before you touch a single driver setting. A dock that cannot reliably push the bandwidth your monitor needs, or a loose external-GPU connection, produces the exact same black-flicker-and-recover symptom as a driver fault, and no amount of driver troubleshooting fixes a cable or connector problem. Plug the monitor directly into the GPU (or the laptop, if it has its own port) for a day and see whether the recoveries follow the dock or stay behind with the direct connection — that single test tells you which side of the setup to focus on.

For power users: keeping this from coming back

Once the immediate error is gone, the more useful question is how to stop finding out about a problem the hard way, mid-game or mid-render. NVIDIA App, AMD Software, and Intel Graphics Software can all be set to check for driver updates automatically and simply notify you rather than requiring you to remember to look — worth turning on once, so the next known driver bug never gets the chance to reach your PC in the first place.

If you run an overclock or an undervolt on purpose, treat "stable" as an ongoing claim, not a one-time result. A curve that survived a twenty-minute stress test can still misbehave months later under a specific new game's shader load, a hotter room in summer, or thermal paste that has started to dry out. Re-running a short stress test after any major driver update, and after any season where ambient temperatures shift noticeably, catches drift before it turns into a recovery notification during something you actually care about.

Before you make any registry edit, including the TdrDelay change above, it costs nothing to export the GraphicsDrivers key first: in Registry Editor, right-click the key, choose Export, and save the file somewhere you will remember. If anything about the change does not sit right, double-clicking that exported file puts everything back exactly as it was, no guesswork involved.

A note on where to get drivers, and when to reach for extra tools

Only ever download a graphics driver from Windows Update or the manufacturer's own site — NVIDIA, AMD, or Intel directly. "Driver update" utilities from unfamiliar third-party sites are a genuinely common way people end up with adware or worse bundled alongside a driver they did not need to go looking for in the first place, and they add no real value over the manufacturer's own tool. Display Driver Uninstaller is the one third-party exception worth using, precisely because it does not install anything — it only removes.

If you troubleshoot GPU issues often, most manufacturer utilities (NVIDIA App, AMD Software) will show live temperature and clock-speed readouts without any extra download, which is usually enough for the kind of monitoring described in Method 6. Dedicated monitoring tools go further, with logging and alerts, but they are a "later" tool, not a "first fix" tool.

One more popular piece of advice worth naming and setting aside: some corners of the internet suggest disabling TDR entirely rather than adjusting its timing. Do not do this. Turning it off does not fix whatever is causing the freeze — it just removes Windows's ability to recover from one, so instead of a one-second flicker you get a full system hang that needs a hard reset. Raising TdrDelay, as covered above, is the version of this fix that actually helps without giving up the safety net.

Frequently asked questions

Does this error mean my graphics card is dying?

Almost never on its own. A single recovery event is Windows catching a driver that froze and restarting it. Repeated recoveries within minutes, especially alongside visible artifacts, coil whine, or a hot case, point toward a hardware or thermal problem worth checking separately.

What does TDR actually stand for?

Timeout Detection and Recovery. It is the Windows feature that watches for a graphics driver that stops responding, resets it, and shows this message instead of freezing or crashing the whole system.

Is it safe to raise TdrDelay to 8 or higher?

Yes, for most gaming and creative workloads. It gives the driver more time before Windows decides it has frozen, stopping false-positive resets during long GPU tasks. It does not fix a genuinely broken driver or failing hardware — it just gives the real fix room to work.

Should I use DDU or just update the driver normally?

Try a normal driver update or rollback first. Reach for Display Driver Uninstaller only when the normal update did not help, since it strips every trace of the old driver before you install a clean one, and it takes longer.

Can a power plan really cause this error?

Yes. On a Balanced power plan, Windows sometimes throttles a discrete GPU during a demanding task to save power, and the resulting stall can look identical to a driver freeze. Setting the plan to High performance for that GPU removes this specific cause.

Does this happen on laptops with two graphics chips?

Yes, and it is more common there. If Windows or the app is using the weaker integrated GPU for a demanding task, or switching between the integrated and discrete GPU mid-task, the handoff can trigger a recovery. Forcing the app to use the discrete GPU in Graphics settings usually resolves it.

Why does it happen more in one specific game or app?

That program is pushing the GPU harder or longer than anything else you run, which is exactly the condition TDR watches for. It usually points at that title's shader compilation, VRAM usage, or a driver bug specific to that game's engine rather than a system-wide fault.

Will reinstalling Windows fix this?

Rarely, and it should not be the first, second, or third thing you try. Almost every cause of this error lives in the driver, the power settings, or the hardware itself, none of which a Windows reinstall changes — a fresh install will pull a fresh driver with the same underlying issue.

How do I know if it's overheating and not the driver?

Check the GPU temperature during the task that triggers the recovery, using the Performance tab in Task Manager or the manufacturer's own utility. Temperatures pushing into the high 80s Celsius or above under sustained load, a case that is hot to the touch, or noticeably louder fans all point toward heat rather than software.

Can an overclock or undervolt cause this?

Yes, this is one of the most common self-inflicted causes. A GPU or memory overclock that is not fully stable, or an undervolt pushed too aggressively, can cause exactly this kind of freeze-and-recover under load. Reset the clocks to their factory defaults and see if the error stops.

Does this affect only gaming, or also everyday use?

It shows up most during games and other GPU-heavy tasks because that is when a driver is under the most sustained pressure, but the same message can appear during video editing, 3D rendering, or even ordinary browsing if the driver itself is faulty or the GPU is genuinely failing.

Is my monitor cable ever the actual cause?

It is an underrated one. A loose or damaged HDMI or DisplayPort cable, or a cable that cannot reliably carry your monitor's refresh rate, can cause the display to briefly drop signal in a way that looks identical to a driver recovery. Reseating the cable, or trying a different one, rules this out in under a minute.

Does this happen over Remote Desktop or in a virtual machine?

It can, though the cause is usually different from a physical PC. Remote Desktop sessions and most virtual machines use a software or virtualized graphics driver rather than your actual GPU, so a recovery there points at the virtual driver, the host's own GPU driver, or a resource limit on the VM rather than a failing graphics card.

What if none of these steps fix it?

At that point you are very likely looking at failing hardware, most often the GPU itself, its power delivery, or the PSU feeding it. If a different, known-good GPU in the same PC still recovers, the fault is upstream of the card; if the same GPU still recovers in a different PC, the card itself is the problem.

Revision note. Originally published November 28, 2016. Rewritten for Windows 11 and Windows 10. The original driver and TdrDelay steps held up well over the years; what changed is mainly where the power plan settings live, since Windows 11 moved the everyday toggle into Settings while Windows 10 keeps it in Control Panel. If your screen just went black mid-game for the third time tonight, take a breath — this one almost always has a straightforward answer, and you'll likely have it sorted before your coffee gets cold.