A hard fact: SSD recovery rates are far below hard drives
The numbers from our own benches are blunt. A mechanical hard drive with a physical failure comes back about 70% of the time, because cleanroom head swaps are a mature, well-established process. An SSD that is completely unrecognized comes back far less often, and in many cases the honest answer is that it cannot be saved. That is not a matter of technician skill. It follows from how the drive is built.
Start by separating the two kinds of SSD failure, because they are handled in completely different ways:
- Logical failure. The drive is still detected, but files were deleted, the volume was formatted or a partition was lost. These are recoverable, with one large caveat that we come to below: TRIM.
- Physical failure. The drive is not detected, drops off the bus, disappears from the BIOS, or reports a capacity of zero or some absurd value. This is the hard category.
TRIM: how "delete" became permanent
On a mechanical disk, deleting a file only writes a "deleted" mark in the directory. The data itself is still sitting on the platter, which is why undelete software can find it. An SSD works differently. When you delete a file, the operating system sends a TRIM command telling the drive those blocks are no longer needed, and the controller physically clears them when it is idle, so that the next write does not have to erase first.
The practical consequence: a file deleted from an SSD can be gone after a night powered on. The recovery window is measured in minutes, not days.
So the first rule of an accidental SSD deletion is cut the power immediately. Not "use the computer less" — cut the power. As long as the drive has power, the controller's garbage collection is quietly emptying your deleted data in the background. Cut the power at once and pull the battery if it is removable; do not shut down normally. On a desktop, power off, pull the drive, and connect it to a read-only device for scanning.
Controller failure: the data is in the NAND, but the key is gone
About 80% of SSD physical failures are in the controller or the firmware. The NAND flash itself is often perfectly healthy. So why not pull the chips and read them directly? Two obstacles:
- The data is sliced, interleaved and often encrypted. The controller splits data into slices and interleaves them across several NAND packages. Most modern SSDs also apply hardware AES, with the key held in the controller. What you read from the chips is a pile of encrypted fragments, and reassembling it needs the algorithm and key of the same controller model. That is why SSD work depends on controller-level repair, transplanting firmware parameters onto a matching controller or reviving the original one by fixing its power supply or oscillator, rather than the cleanroom head swap used on mechanical drives.
- Firmware damage is worse. The FTL, the flash translation layer that maps logical addresses to physical blocks, is the library's index. If the firmware area is damaged, the books are still there but nothing can be found, and the mapping has to be rebuilt by analyzing the NAND package by package. The workload and the uncertainty are both large.
Typical controller-failure symptoms: the drive appears and then vanishes after a while, the BIOS reports a capacity of 0 or 8MB, or the drive casing heats up abnormally. Samsung, Kingston, Western Digital and ZhiTai have all had batches with elevated failure rates, including a firmware issue a few years ago that made one model drop off en masse.
Case: a Samsung 970 EVO Plus, a thesis and two years of data
In June 2026 a graduate student from University Town brought in a Lenovo Legion Y7000P that would not boot into Windows. Sometimes the BIOS saw the 512GB 970 EVO Plus, sometimes it did not, the classic sign of a drive dropping off. On it were the first draft of a graduation thesis and two years of experimental data, with no backup.
We did not keep retrying the machine; every attempt is a risk. We pulled the drive and connected it to a PC-3000 Portable. It powered up and was detected, but reading any sector timed out, which pointed to a firmware-area fault. This drive was lucky: the database for its Phoenix controller is mature. Using vendor-level tools we entered safe mode, rebuilt the firmware area's translator, and the drive was detected again. Imaging the full disk showed about 1.2% of blocks reading unstably, which we recovered with slow-mode reads. The thesis and the experimental data came out intact. The charge was RMB 1,800 for firmware repair and a full-disk image export.
The negative case arrived the same month: a no-name 256GB SSD with a controller missing from every database and completely failed completely and unmarked white-label NAND. It could not be recovered. That is the reality of SSD work. Brand drives succeed far more often than cheap no-name drives because the tool chain supports them. When you buy an SSD, do not chase the cheapest unknown brand; the saving is small and the bet is all of your data.
2026 SSD recovery price reference
| Fault type | Reference price | Notes |
|---|---|---|
| Logical failure (deletion, format, drive still detected) | RMB 300–800 | TRIM makes the success rate fall sharply with time |
| Firmware damage (drive drops off or is not detected, controller repairable) | RMB 1,200–2,500 | High success on mainstream brands; no-name is a gamble |
| Controller-level repair (controller swap, firmware parameter transplant) | RMB 2,000–4,000 | Needs a matching donor board; three to seven days |
| NAND-level analysis (dead controller, chips removed and mapping rebuilt) | From RMB 4,000, quoted after assessment | Encrypted NAND may be impossible; we quote after evaluation |
The trade convention applies here too: no recovery fee if the recovery fails. Testing and assessment are usually free or carry a small bench fee that is credited against the recovery charge.
Five habits that protect data in the SSD era
- Keep at least three copies of anything important. In the SSD era, recovery is not a reliable fallback; backup is. Follow 3-2-1: the machine, a portable drive or NAS, and the cloud, with at least two copies in place.
- Buy brand drives and keep the receipt. Samsung, Western Digital, Kioxia, ZhiTai and Kingston all have mature controller databases and noticeably higher recovery rates, and their warranties are real, often five years. A no-name white-label drive saves maybe RMB 200 and puts every byte at risk.
- Do not fill an SSD. Below 10% free space, wear leveling and garbage collection come under real pressure and the failure rate rises. A 512GB drive holding 480GB for months is asking for trouble.
- Treat any abnormality as a reason to power off and send it in. Slowness, the occasional blue screen, files that corrupt for no reason, a BIOS that sees the drive only sometimes: these are the warning signs of a drive about to drop. Do not wait and see.
- If you work on a laptop, use a dock and a portable drive. Dragging the critical folders across once a week takes ten minutes, and it costs far less than a recovery bill in the thousands.
When an SSD fails, the first step is always to cut the power, then call 020-39029800. Tell us the brand, model, capacity and symptoms, and we can usually say which route the job will take, and whether it is worth saving.