Tool AT-W23REV. 2026-A
Blue screen help: what does the stop code mean, and what helps?
A blue screen shows a stop code at the bottom, such as CRITICAL_PROCESS_DIED or 0x000000EF. It tells you where Windows stopped, not always why. According to Microsoft, third-party drivers are behind around 70 per cent of cases and hardware behind 10 per cent. The blue screen help recognises all 379 codes in the Microsoft reference, even with typos, and ranks causes and steps according to your answers. It also tells you when it is better to stop.
- free
- no sign-up
- runs in the browser
01 / TOOL
Enter the stop code, read off causes and steps
Tool · Look up a stop code
The stop code names the place where Windows stopped, not always the cause. The ranking below is a reasoned narrowing-down; only an analysis of the crash logs gives certainty.
It runs in your browser. Nothing is sent or stored.
02 / HOW TO READ IT
How to read the blue screen
Windows stops when it detects a situation in which it can no longer safely carry on, for example because data would otherwise be damaged. Microsoft calls this a bug check. Three things on the screen then matter.
- The stop code. At the bottom of the screen, as a name in capitals with underscores. This is the value you enter above.
- The module, if shown. Below that, Windows sometimes names a file, usually ending in .sys. It was running when Windows stopped and is the best clue to a driver.
- The progress percentage. Windows is writing the memory dump from which the cause can later be read. Let it finish.
From Windows 11 version 24H2 the message can be black instead of blue; the code is shown further down. If the screen disappeared too quickly, you will find the crash in Event Viewer under Windows Logs > System, with the code as a hex number. The memory dumps are in the folder C:\Windows\Minidump.
03 / CAUSES
Where blue screens come from
Microsoft has analysed what is behind the crashes that reach it. The result is clear: most of them trace back to drivers that do not come from Microsoft. A driver runs in the core of Windows, and an error there cannot simply be caught the way it can in a program.
Research has confirmed this for years. Under Windows XP, 85 per cent of reported failures were caused by drivers Swift et al. 2005. In operating system kernels, drivers contained three to seven times as many errors as the rest of the code, and fresh code was more error-prone than older code Chou et al. 2001. Ten years later, drivers still contained the most errors, simply because they make up most of the code Palix et al. 2011. That is why, with no other information, the tool puts a driver first, and a freshly updated driver counts as the prime suspect.
| Share | Cause |
|---|---|
| 70 % | Third-party driver code |
| 10 % | Hardware problems |
| 5 % | Microsoft code |
| 15 % | unknown, because the memory was too badly damaged for analysis |
Source: Microsoft Learn, “Advanced troubleshooting for Stop error or blue screen error issue”, as of 12 February 2026.
Programs with their own driver, and updates
Not every driver belongs to a device. Third-party antivirus, VPN, backup and tuning programs, or the anti-cheat software of some games, bring their own drivers. Under Windows XP, crashes caused by drivers fell from 2004 to 2006 in all classes except two: antivirus and storage. Antivirus drivers are constantly being reworked, and that keeps producing waves of new errors Glerum et al. 2009. The same analysis also shows why updates help: one manufacturer distributed a fix for a processor error via Windows Update, and within two days the crash reports fell to a fifth.
Updates can, however, also be the trigger themselves. In a survey of 50 administrators, two thirds estimated that 5 to 10 per cent of the updates they installed cause problems Crameri et al. 2007. If you say the blue screen began after an update, you therefore get as the first step after the restart: uninstall the update. For every code, Microsoft advises checking the last change first; a change you name therefore outranks every other assessment in the tool.
When hardware is the cause
Hardware accounts for ten per cent, and such errors are rarely one-offs. In the crash data from around 950,000 consumer PCs, the chance of a crash caused by the processor was 1 in 190. After the first such crash it was 1 in 3.3. Overclocked processors crashed 4 to 20 times as often, laptops less often than desktops Nightingale et al. 2011. Memory is similar: over 8 per cent of the memory modules in Google’s data centres had errors every year, and permanent faults predominated, not one-off glitches Schroeder et al. 2009 Hwang et al. 2012 Sridharan and Liberty 2012. In Facebook’s servers, most memory errors came not from the chips but from the memory controller and channel Meza et al. 2015; that is why the tool advises testing modules in another slot as well.
For the assessment this means: if you say “for weeks” or “daily”, memory, drive and heat move further up. A faulty memory module often produces quite different codes, incidentally, because it corrupts data in changing places. If you keep getting different codes, that is itself a clue.
04 / STOP CODES
The 61 most common stop codes in plain language
The tool describes these codes individually, based on the relevant Microsoft page. It also recognises the other codes among the 379 and assigns them to a group by their name. Each name links to the Microsoft page; it is written for developers, but it gives the parameters in case someone analyses the memory dumps.
| Code | Name | Meaning | Group |
|---|---|---|---|
| 0x00000001 | APC_INDEX_MISMATCH | A driver or file system did not cleanly finish internal processes in the Windows kernel. | Drivers |
| 0x0000000A | IRQL_NOT_LESS_OR_EQUAL | Windows or a driver accessed an invalid memory address at a moment when that is not allowed. | Drivers |
| 0x00000018 | REFERENCE_BY_POINTER | A driver released a system object more often than it requested it. | Drivers |
| 0x00000019 | BAD_POOL_HEADER | A memory area of the Windows kernel (pool) is damaged. | Memory management |
| 0x0000001A | MEMORY_MANAGEMENT | A severe error occurred in the memory management of Windows. | Memory management |
| 0x0000001E | KMODE_EXCEPTION_NOT_HANDLED | A program in the Windows kernel, usually a driver, triggered an error that nobody caught. | Drivers |
| 0x00000024 | NTFS_FILE_SYSTEM | Error in ntfs.sys, the driver through which Windows reads and writes NTFS drives. Back up your data first. | Hard drive and file system |
| 0x0000003B | SYSTEM_SERVICE_EXCEPTION | An error occurred when a program passed into the Windows kernel, usually an invalid memory access. | Windows kernel |
| 0x00000044 | MULTIPLE_IRP_COMPLETE_REQUESTS | A driver tried to complete a request that had already been completed. | Drivers |
| 0x0000004E | PFN_LIST_CORRUPT | The list used to manage RAM is damaged. | Memory management |
| 0x00000050 | PAGE_FAULT_IN_NONPAGED_AREA | Windows accessed invalid system memory, often memory that had already been freed. | Memory management |
| 0x0000005A | CRITICAL_SERVICE_FAILED | A service that Windows needs in order to start has failed. | Windows startup |
| 0x0000006B | PROCESS1_INITIALIZATION_FAILED | Windows could not initialise at startup. Back up your data first. | Windows startup |
| 0x00000074 | BAD_SYSTEM_CONFIG_INFO | The registry is faulty: the system part is damaged, or important entries are missing. | Windows startup |
| 0x00000077 | KERNEL_STACK_INPAGE_ERROR | A page of kernel data could not be read from the page file into memory. Back up your data first. | Hard drive and file system |
| 0x0000007A | KERNEL_DATA_INPAGE_ERROR | Kernel data could not be read from the drive into memory. Back up your data first. | Hard drive and file system |
| 0x0000007B | INACCESSIBLE_BOOT_DEVICE | At startup, Windows lost access to the system partition. Back up your data first. | Windows startup |
| 0x0000007E | SYSTEM_THREAD_EXCEPTION_NOT_HANDLED | A system process in the Windows kernel triggered an error that was not caught. | Drivers |
| 0x0000007F | UNEXPECTED_KERNEL_MODE_TRAP | The processor reported an exception that the kernel could not catch. | Hardware |
| 0x00000080 | NMI_HARDWARE_FAILURE | The hardware has reported a severe error. | Hardware |
| 0x0000008E | KERNEL_MODE_EXCEPTION_NOT_HANDLED | A program in kernel mode triggered an error that was not caught. Microsoft lists this code as very common. | Drivers |
| 0x0000009C | MACHINE_CHECK_EXCEPTION | The processor has reported a severe hardware error (machine check). | Hardware |
| 0x0000009F | DRIVER_POWER_STATE_FAILURE | A driver did not respond in time, or responded incorrectly, when switching to a power-saving state. | Power management |
| 0x000000A0 | INTERNAL_POWER_ERROR | The power management of Windows has detected a severe error. | Power management |
| 0x000000A5 | ACPI_BIOS_ERROR | The BIOS/UEFI of the computer does not fully comply with the ACPI standard for power and device management. | Power management |
| 0x000000BE | ATTEMPTED_WRITE_TO_READONLY_MEMORY | A driver tried to write to write-protected memory. | Drivers |
| 0x000000C2 | BAD_POOL_CALLER | A driver requested kernel memory incorrectly or released it twice. | Memory management |
| 0x000000C4 | DRIVER_VERIFIER_DETECTED_VIOLATION | Driver Verifier has detected a driver error. | Drivers |
| 0x000000C5 | DRIVER_CORRUPTED_EXPOOL | A driver has damaged the system memory pool. | Memory management |
| 0x000000CA | PNP_DETECTED_FATAL_ERROR | Plug and Play management has detected a severe error, probably caused by a device driver. | Devices and USB |
| 0x000000CE | DRIVER_UNLOADED_WITHOUT_CANCELLING_PENDING_OPERATIONS | A driver was unloaded without ending its pending operations. | Drivers |
| 0x000000D1 | DRIVER_IRQL_NOT_LESS_OR_EQUAL | A driver accessed pageable or invalid memory at the wrong time. | Drivers |
| 0x000000D5 | DRIVER_PAGE_FAULT_IN_FREED_SPECIAL_POOL | Driver Verifier caught a driver accessing memory that had already been freed. | Drivers |
| 0x000000E2 | MANUALLY_INITIATED_CRASH | The crash was triggered deliberately, using a debugger or a key combination. | triggered deliberately |
| 0x000000EA | THREAD_STUCK_IN_DEVICE_DRIVER | A device driver is stuck in an endless loop, waiting for the hardware. | Image |
| 0x000000ED | UNMOUNTABLE_BOOT_VOLUME | Windows could not mount the boot drive at startup. Back up your data first. | Windows startup |
| 0x000000EF | CRITICAL_PROCESS_DIED | A process that Windows cannot do without was ended, such as csrss.exe, wininit.exe or winlogon.exe. | Windows kernel |
| 0x000000F4 | CRITICAL_OBJECT_TERMINATION | A process or thread needed for operation ended unexpectedly. | Windows kernel |
| 0x000000F5 | FLTMGR_FILE_SYSTEM | Severe error in the filter manager, through which programs hook into every file access. | Hard drive and file system |
| 0x000000F7 | DRIVER_OVERRAN_STACK_BUFFER | A driver overwrote a buffer area; Windows stops before damage results. | Drivers |
| 0x000000FC | ATTEMPTED_EXECUTE_OF_NOEXECUTE_MEMORY | Code was supposed to run in a memory area that is locked for that purpose. | Drivers |
| 0x000000FE | BUGCODE_USB_DRIVER | Error in a USB driver. | Devices and USB |
| 0x00000101 | CLOCK_WATCHDOG_TIMEOUT | A processor core stopped responding. | Hardware |
| 0x00000109 | CRITICAL_STRUCTURE_CORRUPTION | Windows has detected modified kernel code or kernel data. | Windows kernel |
| 0x0000010D | WDF_VIOLATION | A driver built on the Windows driver framework made an error. | Drivers |
| 0x0000010E | VIDEO_MEMORY_MANAGEMENT_INTERNAL | The graphics memory management cannot recover from an error. | Image |
| 0x00000113 | VIDEO_DXGKRNL_FATAL_ERROR | The DirectX graphics kernel of Windows has detected a violation. | Image |
| 0x00000116 | VIDEO_TDR_FAILURE | The graphics card stopped responding, and restarting the graphics driver failed. | Image |
| 0x00000119 | VIDEO_SCHEDULER_INTERNAL_ERROR | The graphics scheduler of Windows has detected a severe error. | Image |
| 0x00000124 | WHEA_UNCORRECTABLE_ERROR | The hardware has reported a severe error that could not be corrected. | Hardware |
| 0x0000012B | FAULTY_HARDWARE_CORRUPTED_PAGE | Windows has detected damaged memory that only direct hardware access can have caused. | Memory management |
| 0x00000133 | DPC_WATCHDOG_VIOLATION | A driver took too long and blocked the system as a result. | Drivers |
| 0x00000139 | KERNEL_SECURITY_CHECK_FAILURE | The Windows kernel has detected a damaged internal data structure. | Windows kernel |
| 0x0000013A | KERNEL_MODE_HEAP_CORRUPTION | The kernel’s memory management has detected damaged memory blocks. | Memory management |
| 0x00000144 | BUGCODE_USB3_DRIVER | Error in a USB 3 driver. | Devices and USB |
| 0x00000154 | UNEXPECTED_STORE_EXCEPTION | A component of the Windows memory management caught an unexpected error. | Memory management |
| 0x00000164 | WIN32K_CRITICAL_FAILURE | Win32k, a core part of Windows for the user interface, has encountered a critical error. | Windows kernel |
| 0x00000192 | KERNEL_AUTO_BOOST_LOCK_ACQUISITION_WITH_RAISED_IRQL | A driver requested a lock at a time when that is not allowed. | Drivers |
| 0x000001C8 | MANUALLY_INITIATED_POWER_BUTTON_HOLD | The power button was held down for a long time; Windows wrote a memory dump first. | triggered deliberately |
| 0x000001D5 | DRIVER_PNP_WATCHDOG | A driver did not complete a Plug and Play operation within the time allowed. | Devices and USB |
| 0xC000021A | WINLOGON_FATAL_ERROR | Winlogon or CSRSS has crashed; Windows cannot run without these processes. | Windows startup |
Numbers and names from Microsoft’s Bug Check Code Reference, retrieved on 14 November 2026. According to Microsoft, 0x1000007E (SYSTEM_THREAD_EXCEPTION_NOT_HANDLED_M), 0x1000007F (UNEXPECTED_KERNEL_MODE_TRAP_M), 0x1000008E (KERNEL_MODE_EXCEPTION_NOT_HANDLED_M) and 0x100000EA (THREAD_STUCK_IN_DEVICE_DRIVER_M) have the same meaning as 0x0000007E, 0x0000007F, 0x0000008E and 0x000000EA.
05 / STEPS
The steps at a glance, starting with the smallest intervention
The order follows Microsoft’s advice to start with the least disruptive option. The tool moves forward what fits your code and your answers and leaves out what does not; for a hardware fault, for example, resetting Windows. The restart still always comes first. It fixes many faults, but not all: on Windows servers in an early field study, restarts did not always solve the problem, and failures occurred in clusters Kalyanakrishnam et al. 1999.
| Step | helps with | Effort | Your files |
|---|---|---|---|
| Restart and note the code | one-off glitches | 5 minutes | kept |
| Install Windows updates | known driver and processor errors | 15 to 30 minutes | kept |
| Uninstall the latest update | blue screen since an update | 10 to 20 minutes | kept (Microsoft) |
| Roll back or renew the driver | drivers, including the one named on the screen | 15 to 30 minutes | kept |
| Remove recently installed programs with a system driver | antivirus, VPN, tuning, anti-cheat | 15 to 30 minutes | kept |
| Unplug new hardware again | new parts and devices | 10 minutes | kept |
| Undo overclocking and memory profile | overclocking, memory profile, BIOS change | 10 minutes | kept |
| Test the memory | memory and slots | 30 to 60 minutes | kept |
| Check and repair system files | damaged Windows files | 20 to 40 minutes | kept |
| Check the hard drive or SSD, after backing up | read errors, cables, drive | 20 minutes to 2 hours | back up first |
| Test in Safe Mode | Narrowing down: driver or Windows? | 15 to 30 minutes | kept |
| System Restore to a point before the problem | changes since the restore point | 30 to 60 minutes | kept, programs installed afterwards are gone |
| Reset Windows, as the last software step | stubborn software faults | 1 to 3 hours | kept with “Keep my files”, programs are gone |
Effort is a rule of thumb. Menu paths follow Windows 11, as of 14 November 2026; under Windows 10 some items have different names. Whether your Windows 10 still receives updates is shown by the Windows 10 check.
Two traps when restoring
- Point-in-time restore. From Windows 11 24H2, Microsoft lists it first for a PC that no longer starts. According to Microsoft, it rolls back the whole PC, including personal files. If you have saved anything since the restore point, you are better off using the classic System Restore or backing up beforehand.
- BitLocker. For most options in the recovery environment, Windows asks for the 48-digit recovery key if the drive is encrypted. It is often held in the Microsoft account. Without it, the data is lost after a reset, and Microsoft cannot recover it.
06 / LIMITS
When to stop and get help
Most of the steps above are harmless. With a drive that already has read errors, that is different. After a first crash caused by the disk system, around one in three or four PCs crashed again, most within ten days of running time; the researchers advise replacing the drive straight away rather than waiting Nightingale et al. 2011. Even one unreadable sector can cost data without a backup Schroeder et al. 2010, and a rising number of reallocated sectors signals a failure Ma et al. 2015. The hard drive check shows how your drive is doing.
- The drive clicks, grinds or is not recognised. Switch off. Every further start reads from the failing drive, and chkdsk /r even more so. If the data exists nowhere else, this is a case for data recovery; the guide Hard drive failed gives first pointers.
- BitLocker asks for the key, and it cannot be found. Do not reset and do not reinstall.
- The memory diagnostic reports errors, or the crash persists despite clean fans and BIOS default settings. Then a component is faulty; software steps no longer help. The Repair or replace? tool calculates whether the repair is worthwhile.
- There is a smell of burning, the device gets unusually hot, the battery is swelling. Switch off immediately and disconnect from the power. Why laptops get hot is explained in the guide Laptop loud and hot.
- The blue screen comes back after all the steps. Then the memory dumps need analysing. Even Microsoft does not group crash reports by code alone, but compares the call stacks in the dump Dang et al. 2012; only they show which driver was involved.
If the computer no longer even gets as far as the blue screen, the guide PC won’t start can help. If it runs but is sluggish and keeps stuttering, take a look at PC slow: what to do. We analyse the memory dumps by remote support, without you having to bring the computer in; if a part has to be replaced, see Getting a computer repaired.
07 / RESEARCH
What the research says about crashes
The order in the tool is based on Microsoft’s information and on 14 vetted papers. Six findings support it; the full list with verification notes is further down.
- Drivers are the most common cause. 85 per cent of failures under Windows XP Swift et al. 2005, three to seven times as many errors as in the rest of the kernel Chou et al. 2001.
- New code is more error-prone. The newest files had up to twice as many errors as the oldest Chou et al. 2001; after a driver update, the driver therefore comes first.
- Hardware faults come back. After the first crash, a second one from the same cause is up to a hundred times more likely Nightingale et al. 2011. Frequency is therefore a question of its own in the tool.
- Memory errors are mostly permanent. That is why the memory test is worthwhile for recurring crashes Schroeder et al. 2009 Hwang et al. 2012.
- Updates fix problems and cause them. Fixes arrive via Windows Update Glerum et al. 2009, and some updates cause trouble themselves Crameri et al. 2007.
- The code is not a diagnosis. Even Microsoft does not classify crashes by code alone, but compares the call stacks in the dump Dang et al. 2012.
08 / FOR AI ASSISTANTS
Using this tool from an AI assistant
Assistants such as Claude or ChatGPT that speak the Model Context Protocol (MCP) can call this page’s calculation directly instead of rebuilding it. The server is public, needs no sign-in and stores nothing. The answer always includes the link to this page, the date of the data and the limits of the result.
- Endpoint
https://aureum-tech.com/mcp- Tool
werkzeug_bluescreen_hilfe- Transport
- Streamable HTTP, read-only
| Name | Required | Meaning |
|---|---|---|
stoppcode | no | Stop code as a name (CRITICAL_PROCESS_DIED, also with spaces or typos) or as a hex number (0x000000EF, EF, 0xef). Pasted text from the blue screen or from Event Viewer (“The bugcheck was: 0x0000009f …”) also works. Leave empty if the code is unknown. |
modul | no | Optional: file that appeared on the blue screen below the code, e.g. nvlddmkm.sys or ntoskrnl.exe. |
ausloeser | no | What changed before the first blue screen: nichts = Nothing, as far as I know; update = A Windows update; treiber = A new or updated driver; hardware = New hardware or a new device; programm = A new program; bios = BIOS setting or overclocking. If omitted: nichts. |
seit | no | Since when: heute = For the first time today; tage = For a few days; wochen = For weeks or longer. If omitted: tage. |
haeufigkeit | no | How often: einmal = Once; mehrmals = Several times; taeglich = Daily or more often; startet_nicht = Windows no longer starts. If omitted: mehrmals. |
Questions where an assistant should call it
- “My PC shows a blue screen with CRITICAL_PROCESS_DIED, what should I do?”
- “What does the stop code 0x00000124 mean, and is my hardware faulty?”
- “Since the last Windows update I get DRIVER_IRQL_NOT_LESS_OR_EQUAL with nvlddmkm.sys, what helps?”
- “Windows no longer starts, blue screen INACCESSIBLE_BOOT_DEVICE. Is my data at risk?”
How the connection works and what MCP can do in your own business is explained on the page about MCP integrations.
09 / COMMON QUESTIONS
Common questions about blue screens and stop codes
What does the stop code on the blue screen mean?
The stop code tells you where Windows stopped so that nothing worse could happen. It appears at the bottom of the screen as a name such as CRITICAL_PROCESS_DIED, and in Event Viewer as a hex number such as 0x000000EF. Microsoft lists 379 such codes. The code narrows things down; it does not always name the cause. According to Microsoft, third-party drivers are behind 70 per cent of cases and hardware behind 10 per cent.
Is a blue screen serious if it only happened once?
Usually not. According to Microsoft, a restart fixes the problem in most cases. Note down the code and the date, and install the updates. If the same code comes back, it is worth looking into: after a crash caused by hardware, a second one is far more likely than the first was. The exception is codes that point to the hard drive, such as KERNEL_DATA_INPAGE_ERROR. In that case, back up your data after the very first time.
Where do I find the stop code if the screen disappeared too quickly?
In Event Viewer under Windows Logs > System, at the entry for the time of the crash. The code appears there as a hex number, and you can enter that here. The memory dumps are in the folder C:\Windows\Minidump. If the computer keeps restarting, you can choose “Disable automatic restart on system failure” in the startup settings; the blue screen then stays on screen.
What does CRITICAL_PROCESS_DIED mean?
A process that Windows cannot do without was ended, such as csrss.exe, wininit.exe or winlogon.exe. As a number, the code is 0x000000EF. With no further information, the tool gives as the most likely causes “Damaged system files or registry”, “A device driver” and “A Windows update”. If it began after an update, “A Windows update” comes first, and the first step after the restart is to uninstall it.
Is the hardware faulty if I get WHEA_UNCORRECTABLE_ERROR?
Not necessarily, but the hardware has reported an error it could not correct. Microsoft names heat, overclocking, memory and a failing processor; a driver is behind it less often. If the code has been appearing daily for weeks, the tool puts “Heat or power supply” and “Memory (RAM) or slot” first. Check the fans and overclocking first, then the memory. If it carries on, a component has to be replaced.
Can a blue screen destroy my data?
Anything not yet saved at the time of the crash is lost. It becomes dangerous when the code points to the drive: in an analysis of around 950,000 PCs, roughly one in three or four with disk-system errors crashed again after the first time, mostly within ten days of uptime. In that case, back up first rather than spending hours on repairs. If Windows asks for the BitLocker recovery key, do not reset the PC without that key.
Why is my screen black instead of blue?
From Windows 11 version 24H2 the message can be black; Microsoft speaks of blue screen and black screen errors, and the stop code is still shown lower down. A black screen asking you to release the power button belongs to MANUALLY_INITIATED_POWER_BUTTON_HOLD. A black screen with no code at all is something else, for which Microsoft has its own guide.
Where does the blue screen help stop?
It recognises every code in the Microsoft reference and ranks causes and steps according to Microsoft’s information, studies and your answers. Which driver or component it really was is only shown by the memory dump. We analyse that, remotely or on site in Rodewisch and the Vogtland, and fix the cause.
10 / STUDIES AND SOURCES
What the tool is based on
The guide values and notes above are based on these 14 scientific papers. Each one has been checked via its DOI or the publisher’s page (most recently on 14 November 2026): it exists, it has not been retracted, and its findings still hold for this question today.
Swift, M. M.; Bershad, B. N.; Levy, H. M. (2005): Improving the reliability of commodity operating systems. ACM Transactions on Computer Systems 23 (1), pp. 77–110
Finding Device drivers remain a main cause of system failures: under Windows XP, 85 per cent of the failures reported at the time were caused by drivers. The paper proposes isolating drivers in the kernel so that their errors do not take the whole system down with them.
In the tool With no other information, a driver comes first among the causes in the tool.
doi.org/10.1145/1047915.1047919 · Journal article
Chou, A.; Yang, J.; Chelf, B. et al. (2001): An empirical study of operating systems errors. Proceedings of the 18th ACM Symposium on Operating Systems Principles (SOSP), pp. 73–88
Finding An automatic code check of the Linux and OpenBSD kernels found three to seven times as many errors in device drivers as in the rest of the kernel. The newest files had up to twice as many errors as the oldest; code becomes more stable over time.
In the tool A freshly installed or updated driver moves forward in the tool, and the step “Roll back or renew the driver” comes early.
doi.org/10.1145/502034.502042 · Conference paper
Palix, N.; Thomas, G.; Saha, S. et al. (2011): Faults in Linux: ten years later. Proceedings of the 16th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), pp. 305–318
Finding Ten years after Chou et al., drivers still make up a large part of the Linux kernel and contain the most errors. Their error density is now lower than that of other parts such as file systems, however.
In the tool The page explains that drivers are so often to blame mainly because there are so many of them, and not every driver is bad.
doi.org/10.1145/1950365.1950401 · Conference paper
Glerum, K.; Kinshumann, K.; Greenberg, S. et al. (2009): Debugging in the (very) large: ten years of implementation and experience. Proceedings of the 22nd ACM Symposium on Operating Systems Principles (SOSP), pp. 103–116
Finding Windows error reports sort billions of crashes into groups and steer fixes back to the devices. Under Windows XP, crashes caused by drivers fell from 2004 to 2006 in all classes except antivirus and storage; antivirus drivers are constantly being reworked. A processor fix via Windows Update cut crash reports to a fifth within two days, and a driver manufacturer halved its share of all kernel crashes in five months.
In the tool Installing updates comes early in the sequence of steps, programs with their own driver such as antivirus are a cause of their own, and the stop code counts only as a group, not as a diagnosis.
doi.org/10.1145/1629575.1629586 · Conference paper
Dang, Y.; Wu, R.; Zhang, H. et al. (2012): ReBucket: A method for clustering duplicate crash reports based on call stack similarity. Proceedings of the 34th International Conference on Software Engineering (ICSE), pp. 1084–1093
Finding To group crash reports of the same error, the method compares the call stacks from the reports; for five Microsoft products it assigns them more accurately than the previous rules.
In the tool Justifies the limit of the tool: only the call stack in the memory dump shows the culprit, not the stop code alone.
doi.org/10.1109/ICSE.2012.6227111 · Conference paper
Nightingale, E. B.; Douceur, J. R.; Orgovan, V. (2011): Cycles, cells and platters: an empirical analysis of hardware failures on a million consumer PCs. Proceedings of the 6th European Conference on Computer Systems (EuroSys), pp. 343–356
Finding Crash data from around 950,000 consumer PCs: hardware faults come back. After a crash caused by the processor, the chance of the next was 1 in 3.3 instead of 1 in 190; with errors in the disk system, 86 per cent of repeats occurred within ten days of running time, and the authors advise replacing the drive straight away. Overclocked processors crashed 4 to 20 times as often, laptops less often than desktops.
In the tool Repeated crashes and “for weeks” move hardware causes higher up; for drive codes the tool warns at the very first time, and undoing overclocking is a step of its own.
doi.org/10.1145/1966445.1966477 · Conference paper
Schroeder, B.; Pinheiro, E.; Weber, W.-D. (2009): DRAM errors in the wild: a large-scale field study. Proceedings of the 11th International Joint Conference on Measurement and Modeling of Computer Systems (SIGMETRICS), pp. 193–204
Finding Two and a half years of measurements in Google’s server fleet: memory errors are orders of magnitude more common than assumed, with over 8 per cent of memory modules affected each year. Permanent hardware faults predominate; temperature plays a surprisingly small role in the field.
In the tool For recurring crashes with changing codes, memory moves forward, and the memory test is a fixed step.
doi.org/10.1145/1555349.1555372 · Conference paper
Hwang, A. A.; Stefanovici, I. A.; Schroeder, B. (2012): Cosmic rays don't strike twice: understanding the nature of DRAM errors and the implications for system design. Proceedings of the 17th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), pp. 111–122
Finding Memory is one of the leading hardware causes of crashes in data centres. A large share of the memory errors in the field are hard errors, meaning recurring ones, and not one-off glitches.
In the tool Because memory errors come back, the test is worthwhile; the page thereby explains why a memory fault often produces quite different codes.
doi.org/10.1145/2150976.2150989 · Conference paper
Sridharan, V.; Liberty, D. (2012): A study of DRAM failures in the field. International Conference for High Performance Computing, Networking, Storage and Analysis (SC12), pp. 1–11
Finding Eleven months of memory errors in a large compute cluster: permanent errors outnumber transient ones, and there are failures of whole rows or columns as well as errors in the circuitry on the board.
In the tool Together with Schroeder 2009 and Hwang 2012, supports the ranking: if crashes accumulate, a memory fault becomes more likely.
doi.org/10.1109/SC.2012.13 · Conference paper
Meza, J.; Wu, Q.; Kumar, S.; Mutlu, O. (2015): Revisiting Memory Errors in Large-Scale Production Data Centers: Analysis and Modeling of New Trends from the Field. 45th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN), pp. 415–426
Finding Fourteen months of memory errors across Facebook’s entire server fleet: most errors came not from the memory chips but from the memory controller and memory channel.
In the tool The memory test step advises testing modules individually and in another slot, rather than suspecting only the module.
doi.org/10.1109/DSN.2015.57 · Conference paper
Schroeder, B.; Damouras, S.; Gill, P. (2010): Understanding latent sector errors and how to protect against them. ACM Transactions on Storage 6 (3), pp. 1–23
Finding Unreadable sectors occur in the field in significant numbers and become more common with larger drives. Even a single one can cost data in systems without redundancy.
In the tool For codes that point to read errors, the tool puts the backup before any repair.
doi.org/10.1145/1837915.1837917 · Journal article
Ma, A.; Traylor, R.; Douglis, F. et al. (2015): RAIDShield: characterizing, monitoring, and proactively protecting against disk failures. ACM Transactions on Storage 11 (4), pp. 1–28
Finding Failure data from around one million SATA drives over up to five years: the number of reallocated sectors is closely linked to an impending failure.
In the tool The step “Check the hard drive or SSD” begins with the SMART values and refers to the hard drive check.
doi.org/10.1145/2820615 · Journal article
Kalyanakrishnam, M.; Kalbarczyk, Z.; Iyer, R. (1999): Failure data analysis of a LAN of Windows NT based computers. Proceedings of the 18th IEEE Symposium on Reliable Distributed Systems (SRDS), pp. 178–187
Finding Event logs of Windows NT mail servers over six months: most restarts had software causes, but the restart did not always solve the problem, and failures occurred in clusters.
In the tool The restart comes first, but if you choose “Several times”, you get the full sequence of steps instead of advice to wait and see.
doi.org/10.1109/RELDIS.1999.805094 · Conference paper
Crameri, O.; Knezevic, N.; Kostic, D. et al. (2007): Staged deployment in Mirage, an integrated software upgrade testing and distribution system. Proceedings of the 21st ACM Symposium on Operating Systems Principles (SOSP), pp. 221–236
Finding In a survey of 50 administrators, two thirds estimated that 5 to 10 per cent of the updates they installed cause problems; 18 per cent had already experienced severe outages caused by updates. Common reasons are broken dependencies and faulty updates.
In the tool If the blue screen began after an update, “Uninstall the update” moves to the top of the steps.
doi.org/10.1145/1294261.1294283 · Conference paper
Standards, authorities and manufacturer data
- Bug Check Code Reference (all stop codes) · Microsoft Learn
- Advanced troubleshooting for Stop error or blue screen error issue (distribution of causes 70/10/5/15) · Microsoft Learn
- Troubleshooting unexpected restarts and stop code errors in Windows · Microsoft Support
- System File Checker: DISM first, then sfc · Microsoft Support
- Windows startup settings and Safe Mode · Microsoft Support
- Recovery options in Windows · Microsoft Support
- Uninstall a Windows update · Microsoft Support
- System Restore · Microsoft Support
- Find your BitLocker recovery key · Microsoft Support
- Perform a clean boot in Windows · Microsoft Support
- Windows RE: technical reference (start after two failed attempts) · Microsoft Learn
- chkdsk · Microsoft Learn
- Driver Verifier · Microsoft Learn
11 / MORE TOOLS
Still have a question?
- AT-W2 · Home & business
Windows 10 check
Does this PC still get security updates, and what needs doing by October 2027? - AT-W3 · Home & business
Wi-Fi planner
How many access points or mesh units does your home, office or warehouse need? - AT-W4 · Businesses
IT emergency check
How quickly could your business get back to work after an outage? - AT-W5 · Businesses
E-mail check
Are your e-mails landing in spam? Your domain’s SPF, DKIM and DMARC in plain language. - AT-W6 · Businesses
UPS calculator
Which UPS keeps your server, router and till running long enough during a power cut? - AT-W7 · Home
Digital legacy planning sheet
Which accounts, devices and contracts should your family know about?
Does the blue screen come back despite everything? Send us the copied result. We read the crash logs, remotely or on site in Rodewisch and throughout the Vogtland, and fix the cause.
+49 3744 4399760[email protected]Mon–Fri 8:00–17:00 · Straße der Jugend 5, 08228 Rodewisch