Fix it now
0x80070017 is Win32 23, ERROR_CRC, a data error on a cyclic redundancy check. 0x8007045D is Win32 1117, ERROR_IO_DEVICE – the request could not be performed because of an I/O device error. Both say Setup cannot reliably read its own source. Verify the download first, because everything downstream of a corrupt ISO is wasted effort.
Get-FileHash -Path .\Windows.iso -Algorithm SHA256
- If the hash does not match, stop. Download it again over a connection you trust and re-check before touching any hardware.
- Rebuild the media on a different USB stick from the verified ISO, and use a port directly on the machine rather than a hub, a dock or a front panel.
- If you are upgrading rather than installing fresh, copy the ISO to the internal disk and mount it, then run setup from the mounted drive. That removes the removable media path entirely.
- Before touching the target disk, read its health:
Get-PhysicalDisk | Get-StorageReliabilityCounterin PowerShell, and check the System log in Event Viewer for disk warnings around the failure. - If the counters or the log show real errors, back up or image the drive before running any repair. A write-repairing pass on a drive that is already throwing I/O errors can make recovery harder.
- Only then check the file system, and start with the read-only scan:
chkdsk C: /scan, which Microsoft documents as an online scan on NTFS. - Test the memory. Bad RAM produces exactly these symptoms during large copies; run the built-in Windows Memory Diagnostic and let a full pass complete.
chkdsk /f requires the volume to be locked, so on the system drive it schedules itself for the next restart and then repairs. Do not run it on a suspect drive before the data is safe.
If Setup copies past the point it previously failed, you are done. If not, the next section works down the chain from the download to the target disk.
Why it happens
Every code here is a Win32 error about reading or writing bytes, and Windows is repeating it verbatim rather than interpreting it. That is worth knowing because it means the codes are honest: they describe what the operating system was told by the layer beneath it, and none of them is a statement about Windows, your licence or the update itself.
0x80070017 is ERROR_CRC, ‘Data error (cyclic redundancy check)’ – the bytes that arrived did not match the checksum they came with. 0x8007045D is ERROR_IO_DEVICE, ‘The request could not be performed because of an I/O device error’ – the read did not complete at all. 0x8007001E is ERROR_READ_FAULT, ‘The system cannot read from the specified device’. 0x80070015 is ERROR_NOT_READY, ‘The device is not ready’, which is what a removable device looks like when it drops off the bus mid-operation.
0x800703EE is subtler: ERROR_FILE_INVALID, ‘The volume for a file has been externally altered so that the opened file is no longer valid’. Setup had a handle open and the volume underneath it changed. On removable media that is a stick reconnecting; on a network path it is the share going away and coming back.
0x8007025D – 0x2000C is the one this article previously got wrong, and the correction matters. The result half is Win32 605, ERROR_BAD_COMPRESSION_BUFFER: ‘The specified buffer contains ill-formed data.’ It is not a read failure. The extend half decodes as SAFE_OS / APPLY_IMAGE, so it is raised while the image is being applied. Microsoft’s published row for that exact pairing names two causes – corrupt ISO metadata, or a problem with the storage medium such as a RAM module containing bad blocks – and the published mitigation is to redownload the ISO or rebuild the media with the Media Creation Tool.
The downloaded image is corrupt
You have this one if The hash does not match, or the ISO was fetched over an unreliable connection or copied from a shared drive.
- Download the ISO again, from Microsoft, over a connection you trust.
- Run
Get-FileHash -Algorithm SHA256on the result and compare it with the published value. - Build media only once the hash matches. Everything downstream of a bad ISO is wasted effort.
Microsoft’s own mitigation for 0x8007025D – 0x2000C starts here: redownload the ISO, or re-create the installation media.
The stick or the port is unreliable
You have this one if The same media fails on more than one machine, or the failure moves around between attempts on one machine.
- Rebuild on a different stick, preferably a known brand rather than a promotional giveaway.
- Use a port directly on the machine, at the back on a desktop, avoiding hubs, docks and extension cables.
- On a machine that is fussy about USB 3, try a USB 2 port; the older controller path is often more forgiving during Setup.
Memory is corrupting data in transit
You have this one if Failures land on a different file each attempt, and the machine has a history of unexplained crashes or has been overclocked.
- Run the built-in Windows Memory Diagnostic and let a full pass complete; a single error explains everything on this page.
- In firmware, turn off any memory overclocking profile and return the memory to its default speed.
- If the machine has several modules, test with one at a time to identify the faulty one.
Microsoft names a RAM module with bad blocks directly as a cause of 0x8007025D – 0x2000C, so this is not a long shot.
The target disk is failing
You have this one if The media reads fine on other machines, and this machine also throws errors during normal use or stalls on file operations.
- Read the health counters first:
Get-PhysicalDisk | Get-StorageReliabilityCounter, and look at ReadErrorsUncorrected, WriteErrorsUncorrected and Wear. - Check the System log in Event Viewer for disk warnings recorded around the failure.
- If either shows real errors, image or back up the data before you run any repair.
- Use the read-only
chkdsk C: /scanon a suspect drive, and treat a repairing pass as something you do after the data is safe. - If the drive is failing, replace it rather than installing on it.
Installing Windows on a disk that is already producing read errors buys a working machine for days at most. Replace it first.
The source is a disc or a network path
You have this one if You are installing from a burned disc, or running setup from a share or an external drive that disconnects.
- Copy the entire source to a local disk and run setup from there, so a momentary disconnection cannot break a read.
- If a disc is unavoidable, burn a new one at the slowest speed the drive offers.
- For network installs, check for link flapping and for power management on the network adapter putting the connection to sleep.
Full reference
What each code actually says
| Code | Win32 | Published text |
|---|---|---|
0x80070017 |
23 | Data error (cyclic redundancy check) |
0x8007045D |
1117 | The request could not be performed because of an I/O device error |
0x800703EE |
1006 | The volume for a file has been externally altered so that the opened file is no longer valid |
0x80070015 |
21 | The device is not ready |
0x8007025D - 0x2000C |
605 | The specified buffer contains ill-formed data, raised while applying the image in the Safe OS phase |
0x8007001E |
30 | The system cannot read from the specified device |
Narrowing down which link is bad
| Observation | What it points at |
|---|---|
| The ISO hash does not match the published value | The download, before any hardware is involved |
| The same stick fails on several machines | The media, not the machines |
| Several sticks fail on one machine | That machine’s ports, memory or target disk |
| It fails at a different file each attempt | Memory or a marginal connection |
| It fails at the same file every time | A specific unreadable area on the media |
| Upgrading from a mounted ISO on the internal disk works | The removable path was the fault |
Checking a disk in a safe order
- Read
Get-PhysicalDisk | Get-StorageReliabilityCounterand note the error counters, wear and temperature. - Read the System log for storage warnings around the time of the failure.
- If either shows real errors, get the data off the drive – image it if you can – before anything else.
- Run
chkdsk C: /scan, which is the online read-only scan, and read the result. - Only run a repairing pass once the data is safe and you have decided the drive is worth repairing rather than replacing.
The order matters more than the commands do. A repairing pass rewrites metadata, and on a drive that is already failing to complete reads it can turn a recoverable situation into an unrecoverable one. Nothing on this page is urgent enough to justify skipping the backup.
Working with mounted media instead
For an in-place upgrade, copying the ISO to the internal disk and mounting it removes the USB stick, the port, the hub and the cable from the problem in one move. If the upgrade then works, you have proved the source path was the fault. It also lets you read %windir%\Panther\setuperr.log while the upgrade runs rather than after a rollback.
When Setup keeps failing at the same file
- Compare the file named in the log against the same file on freshly built media; a consistent failure point is usually a bad area on the stick.
- Rebuild the media rather than copying the single file across, because a stick that has one bad area rarely has only one.
- If the same file fails from freshly built media on two machines, go back to the ISO and the hash.
- If it fails at a different file every time, stop blaming the media and test the memory.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
0x80070017 |
Win32 23, ERROR_CRC: data error, cyclic redundancy check – what was read did not match its checksum | Microsoft Learn |
0x8007045D |
Win32 1117, ERROR_IO_DEVICE: the request could not be performed because of an I/O device error | Microsoft Learn |
0x800703EE |
Win32 1006, ERROR_FILE_INVALID: the volume for a file was externally altered, so the opened file is no longer valid | Microsoft Learn |
0x80070015 |
Win32 21, ERROR_NOT_READY: the device is not ready | Microsoft Learn |
0x8007025D - 0x2000C |
Win32 605, ERROR_BAD_COMPRESSION_BUFFER: ill-formed data while applying the image in the Safe OS phase. Microsoft names corrupt ISO metadata or bad storage, such as a RAM module with bad blocks | Microsoft Learn |
0x8007001E |
Win32 30, ERROR_READ_FAULT: the system cannot read from the specified device | Microsoft Learn |
Confirm the fix worked
- The ISO hash matches the published value before you build anything from it.
- Setup copies files past the point where it previously failed.
Get-PhysicalDisk | Get-StorageReliabilityCountershows no uncorrected read or write errors on the target drive.- A full memory diagnostic pass completes without reporting a problem on any machine where memory was suspected.
- The installed machine runs a full day of normal use without storage warnings appearing in the System log.
Questions people ask about this
Could this be a licensing problem?
No. Every code here is a hardware or download problem, your entitlement is unaffected, and no key or upgrade licence changes what a failing USB stick does. If somebody offers you a product as the fix for a CRC error, they have misread the code.
Does one memory error really matter?
Yes. A single flipped bit during a multi-gigabyte copy produces exactly the checksum and buffer errors on this page. Microsoft names a RAM module with bad blocks as a published cause of 0x8007025D – 0x2000C.
Should I run chkdsk /f first?
No. Read the reliability counters and the System log first, and get the data safe if either shows errors. chkdsk C: /scan is the online read-only scan and is the right first command on a drive you already suspect; a repairing pass comes after the backup, not before it.
Why does the mounted-ISO route work when the USB stick does not?
Because it removes the stick, the port, the hub and the cable from the path. If the upgrade succeeds from a mounted ISO on the internal disk, you have proved where the fault was.
Is it worth retrying the same media a few times?
Once, to see whether the failure lands at the same file. Same file every time means a bad area on the media; a different file each time means memory or a marginal connection. After that, retrying is not diagnosis.
