Your CNC control PC died: keeping old machines earning

First, stop and protect the evidence

A dead PC invites fast, destructive fixes: reformat the drive, swap in a new disk, let a well-meaning technician "clean it up." Every one of those can destroy the only copy of something the shop needs. Before anyone touches the drive, decide one thing: is there a verified image of it? If not, imaging the drive is step one, and it comes before diagnosis, before repair, and before replacement. A failing drive often still reads once. It rarely reads twice.

Imaging means a sector-level copy of the whole drive, made read-only where possible, stored on separate media, and verified by reading it back. It is not the same as copying the visible files. Deleted installers, configuration files in odd folders, license files, and the exact operating system setup live outside what a file copy grabs. The image preserves all of it, including the state the programs were actually running in. If the drive is already clicking, grinding, or failing to spin up, stop powering it on. Repeated attempts on a dying drive make recovery harder, and a professional drive recovery is cheaper than a new machine but only while the drive still has something left to read.

What actually breaks when a control PC dies

"The PC died" covers several different failures, and the fix depends on which one it is:

• The hard drive failed. The most common case on machines that have run for twenty years. The PC hardware may be fine. The data is the casualty, and the image is the rescue.
• The power supply or motherboard failed. The drive may be perfectly healthy. Move it to another machine or image it before assuming anything is lost.
• The operating system stopped booting. Corruption, a failed update on a machine that should never have updated, or a dying CMOS battery resetting the BIOS. The programs and files are often intact underneath.
• The CMOS battery died and the settings were lost. On old PCs, a dead coin cell can reset boot order, drive geometry, and serial port configuration. Replacing the battery is cheap. Losing the settings notes is not.
• The monitor, keyboard, or serial hardware failed. The PC may be running fine behind a dead display. Check the obvious before declaring the system dead.

Notice what is not on this list: the machine tool itself. The control, the spindle, the ways, and the servos do not fail because a PC beside them did. Treating a PC failure as a machine failure is the expensive mistake this guide exists to prevent.

What only that PC holds

Before you decide anything, inventory what lived on that PC and nowhere else. In most shops it is some mix of:

• Part programs. The G-code and conversational programs the shop actually runs, often edited on that PC over years and never backed up anywhere else.
• Transfer and DNC settings. The baud rate, data bits, parity, stop bits, flow control, and cable wiring for each machine. These look trivial and are maddening to rediscover by trial and error.
• Installers and license files. The setup programs for the transfer utility, editor, or post-processor, plus dongle drivers and activation files. The vendor may be gone, so there is no download page to go back to.
• Machine parameter backups. If anyone ever backed up the control parameters, offsets, or tool tables to the PC, that backup may be the only copy. Guard it like the machine depends on it, because it does.
• Notes and tribal knowledge. A text file with the serial settings, a folder named after the operator who set it up, a README nobody else has read. Copy it all. Sort it later.

If any of these exist only on the dead drive, the drive image is now the most valuable object in the shop. Label it, store a copy off the machine, and do not work from the original image. Work from a copy of the copy.

The image-first checklist

Do these in order, and do not skip ahead to replacement shopping until they are done:

1. Stop powering a failing drive. Clicking, repeated read errors, or a drive that appears and disappears means every power-on risks more loss.
2. Image the drive, sector by sector. Use a write-blocking method where possible. Verify the image by reading it back and comparing hashes, the same standard applied in our published Millplus control-disk audit.
3. Record the PC setup from memory while it is fresh. Who set it up, what cable goes where, which serial port, which program talks to which machine. Write it down today, not next month.
4. Inventory the installed programs from the image. List every program on the drive, and mark the ones production actually uses. That list is the recovery scope.
5. Back up the part program library separately. The library should live in at least two places that are not the machine PC.
6. Only then price the options. Repair the PC, replace the PC, recover the programs, or replace the machine. In that order of cost, not the reverse.

What recovery can and cannot do here

At Orphanware Recovery, Josiah Wilson directs every engagement and signs off on every result, and the decompilation, reconstruction, and testing pipeline is carried out by AI agents under his direction.

Recovery covers the PC-side software: the transfer programs, editors, translators, and utilities the shop depends on to feed its machines. If the binary for a program survives on the drive image, it can be decompiled and reconstructed into readable source, then rebuilt and tested against the original, using the shop's own part programs as the test cases. That is the same pipeline behind the measured results on this site, and it starts with the $45 Recon Report: a verdict on whether a specific program can be recovered, before anyone commits to a full recovery.

The boundary matters. Proprietary controller firmware and ladder logic inside the control are the control vendor's domain, and this work does not claim to recover them. But in most dead-PC cases, the control is fine and the PC layer is the failure. Recovering that layer is what puts a sound machine back to work without replacing it.

The economics, in the right order

A replacement machine prices new iron, installation, tooling, retraining, and lost production. A replacement PC prices a box. A recovered program prices the software question itself. The recon costs $45 and returns a verdict in about 3 days. If a program cannot be recovered, you learned it for $45 and can plan around fact instead of hope. If it can, you have a scope and a fixed price for putting the machine back to work.

The order is the whole point. Shops get into trouble when the machine replacement conversation starts before anyone has imaged the drive or established whether the software can be saved. Image first, inventory second, price the options third. While the original drive still reads, every option stays open. Once the last readable copy is gone, every option gets harder, slower, and more expensive.