Samsung 990 Pro Available Spare at 0%. Don't Flash Firmware First.

Written and maintained by the Desert Data Recovery lab team. · Updated October 10, 2026

We already wrote about SSD failure signs, about NAND wearing out, and about an SSD that vanishes from BIOS. A WD Black that disappears after sleep is a different page: don't flash that firmware first either. This one is the Samsung screen. The number is bad, a forum reply or Magician itself points at a firmware update, and the stick is the only place the files exist.

Read the row before you read the advice

The NVMe SMART health log has more than one percentage, and the 2023 Samsung note named one of them. Available Spare is a normalized percentage, 0 to 100, of remaining spare capacity. Available Spare Threshold is the line below which the drive can raise an event. Critical Warning bit 0 means spare has fallen below that threshold. A different bit means the controller has placed the media in read-only mode. Percentage Used is something else again: a vendor estimate of life used. It is allowed to pass 100, and passing 100 is not, by itself, the spare pool hitting zero.

Samsung Magician's health percentage is not that spare field. In January 2023, before the 990 Pro fix shipped, Tom's Hardware quoted Samsung community moderator DavidB on the inquiries Samsung was looking at: SMART Percentage Used in Samsung Magician, and Health Status in CrystalDiskInfo, on the 990 Pro. If the window in front of you says Available Spare 0%, you are looking at the spare field, which that note did not name. If it says Percentage Used, or a health percentage with no spare row, you are looking at what the note did name. Photograph whichever rows are already on screen, including the firmware string. Do not install a benchmark, and do not run Magician's diagnostic, to get a cleaner screenshot.

A 0 in either place is not proof the NAND was wiped. Spare is the pool the controller uses when blocks fail. Percentage Used is an endurance estimate. The files are addressed through the mapping table until the controller stops presenting them. An empty health bar is a status. It is not a directory listing.

Two firmware notes from 2023, and they are not the same

The 980 Pro came first. Puget Systems' support guide, published January 31, 2023, says they had a surprising number of failures on the 2TB 980 Pro, that Samsung confirmed an issue in older firmware, and that the failure Puget was seeing was a drive suddenly locked read-only. A read-only boot drive does not boot. Puget writes that Samsung confirmed the issue in firmware 3B2QGXA7 and recommended updating 980 Pro drives to 5B2QGXA7 to prevent it. The same guide says Samsung stated the update should prevent the failures and will not reverse the problem on an already failed drive. Puget's own records were not showing the symptoms on 4B2QGXA7 or 5B2QGXA7. Their February 2, 2023 reliability post says the smaller 500GB and 1TB 980 Pro seemed unaffected. That is Puget's field report, not a Samsung capacity matrix we can quote. The guide also says a limited number of users reported problems after the update, which is why Puget emphasized a backup first. Puget's own page tells you the article is from January 2023 and may be out of date. The part that still governs a drive that has already locked is the sentence Samsung is reported as saying: the update does not reverse an already failed drive.

Tom's Hardware, writing up those 980 Pro reports, said owner screens were showing media and data integrity errors and Available Spare below 100 percent. That Available Spare detail is Tom's characterization of user reports. It is not a line in Puget's guide, and it is not a Samsung bulletin that says spare hits 0. The symptom Puget itself documented is the read-only lock.

The 990 Pro note is the other one. The FPS Review, on February 13, 2023, printed a Samsung forum post and said the wording had been adapted at Samsung's request. The post says Samsung had been investigating SMART anomalies on the 990 PRO, that the anomalies were caused by the firmware, that an update released that day fixes them, and that SMART values are not restored to factory settings after the update. Actual SMART values vary by environment and use. Tom's Hardware quotes DavidB the same day: the anomalies were caused by problems with the 990 Pro firmware, firmware was released February 13, 2023 to address them, and SMART values are not reset to factory defaults. The version string on that update is 1B2QJXD7. Puget Systems, on March 14, 2023, names the original firmware they were seeing as 0B2QJXD7 and the fix as 1B2QJXD7. Their translation of a German Samsung community moderator matches the same limit: SMART values will not be restored. Puget says you can check an affected drive before and after the update and the health level stays the same.

What Samsung did not publish, and what we will not fill in, is a teardown that says every lost percentage was a retired block rather than a counter. Ars Technica, February 13, 2023, states that gap directly. The 1B2QJXD7 update does not reset the wear percentage. The bug may have degraded drives early, or the firmware may have been unable to tell that counter from real wear. Puget's February 2, 2023 post, written before this fix, said there was a chance the 990 Pro issue was improper reporting of endurance loss. After the fix, Puget's plan for their own customers was to update, then replace drives whose remaining health was already unacceptable. The flash was not the repair of a drive that had already fallen off.

Later Magician builds have offered newer 990 Pro strings after 1B2QJXD7. A Samsung community thread documents an offer of 3B2QJXD7 to a drive that was already on 1B2QJXD7. A newer string is not a statement that Available Spare comes back. A 990 Pro with Heatsink is still a 990 Pro, but the string on the screen is the one you photograph. A 990 EVO is a different product. A plain 980, without the Pro, is not the drive in Puget's 3B2QGXA7 note. Do not point one model's firmware file at another because the box says Samsung.

A firmware update does not put spare back

Spare the controller has already taken out of service stays out of service. That is what a spare pool is. Samsung's 990 Pro note says the SMART values are not restored. Puget's report of Samsung's 980 Pro note says an already failed drive is not reversed. If the 990 Pro number was only a bad counter, Samsung still said the counter is not reset. If the blocks were retired, the file you are about to write does not put them back. The update is the wrong tool for the number on the screen.

It is also a write, on a drive you are worried cannot spare the writes. Samsung's Magician 9 installation guide, revision 1.9 (December 2025), says there is always a risk of data loss when updating SSD firmware, and that backing up important data first is imperative. The same guide says not to hibernate, sleep, or power off the computer during the update, because that can cause a system error or a firmware update failure, and that the PC shuts down on its own after the update. Magician counts down 20 seconds and powers off. Puget's 990 Pro instructions start at the same place: back up, because a firmware update can brick the drive. They call that chance tiny. Tiny is not zero, and their sentence assumes the backup already exists.

On a drive that is already at a critical spare reading, already read-only, or already dropping out of BIOS, that shutdown is another power cycle of a controller in a bad state. An update that hangs is the moment people are told to move the stick to another M.2 slot and try again. That is a second flash attempt. Do not do it with the only copy. A bootable firmware ISO from Samsung's tools page is the same class of write. It is not a gentler one because Magician would not start.

The case the 2023 updates were actually for is the other one. The drive still behaves, the firmware string is still 0B2QJXD7 on a 990 Pro or 3B2QGXA7 on a 980 Pro, and a second copy has already been made and opened. Puget told people in that position to update so the decline would stop, and to keep watching the health number afterward. This page is the drive that does not have that second copy, or that has already reached 0% spare, a critical warning, or read-only. Those are not the preventative case.

Read-only is the drive protecting the data

When a Samsung NVMe goes read-only, Windows may say the disk is write-protected, files open and will not save, or the volume mounts and every write fails. Puget's 980 Pro guide names that lock as the common failure they saw, and names the practical response: copy the data to another drive before you replace the SSD. The firmware update, in that guide, does not bring a locked drive back.

The NVMe critical-warning bit for "media is read-only" is not the same bit as "spare is below the threshold." A 990 Pro can show a bad health percentage and still accept writes. A drive that has actually gone read-only has stopped accepting them. Samsung's 990 Pro bulletin described SMART anomalies. It did not say every 990 Pro goes read-only. If yours has, treat the lock as the controller refusing writes. Do not decide it is the 980 Pro bug, and do not decide a 990 Pro firmware file will clear it.

Fighting the lock is how the remaining readable window gets spent. diskpart attributes disk clear readonly is a host flag. It is not a spare-block repair. Initialize Disk writes a partition table. Format writes a filesystem. CHKDSK /f and Startup Repair write metadata. A full-surface scanner reads the drive hard and, if you let it "fix," writes. None of those put spare back. If the files still open, copy them off once. If Windows is installed on the stick, every boot is more activity on it. Do not reboot it so Magician can try an update that needs a shutdown at the end.

The other Magician buttons are not a fix

Firmware is the button people are sent to. The rest of that window can do worse, faster.

• Secure Erase. Samsung's Magician 6 leaflet describes it as a permanent delete and a complete wipe. It is not a way to clear a critical warning or restore performance on a drive you still need.
• PSID Revert. The Magician 9 guide says this uses the PSID printed on the label, and that afterward all data on the drive is deleted. The guide's model list for that function names the 980 Pro. A 990 Pro label can carry a PSID too. Either way the code is not a key that unlocks your files, and the leaflet's line about using it when other recovery methods are exhausted is still a wipe. Do not type it in because spare looks bad.
• Over Provisioning. The Magician 9 guide describes this as a change to the last partition. It is not a donation of fresh spare blocks to a drive that has already reported the pool empty.
• Diagnostic scan. The same guide says a short scan needs 5GB of free space, so it is not a read-only look at the drive. A full scan is a long read of a drive that is already reporting it is out of spare. Skip both on the patient.
• Performance Optimization. That is a TRIM pass, and only on NTFS. TRIM tells the drive it may erase blocks the operating system calls free. The guide says disconnecting the SSD during it can corrupt data. It does not raise Available Spare.
• Data Migration. The guide says it only clones a source drive with an operating system on it, and that a damaged source may not clone properly. Puget already found it would not do the job on a read-only 980 Pro. A straight file copy of what you need is the move, if the drive is still reading. It is not a license to keep retrying.

Magician often cannot see a drive sitting behind Intel RST or VMD. The replies under those posts tell people to pull the stick, turn VMD off, or update the members one at a time. If these 990 Pros are a motherboard RAID, do not delete the volume and do not disable VMD so a firmware button will light up. Delete Volume wipes RAID metadata. A stripe has no second copy of each block, which is the RAID 0 problem. Breaking the array to flash a member is two writes, not a diagnostic.

What this is not

A drive that vanished after sleep, with an empty M.2 slot and no SMART screen at all, is a controller that did not come back. That is the SN850X page and the SSD not detected page. You do not have an Available Spare reading if the drive never identified. Do not run the 990 Pro steps against a stick BIOS cannot see.

A click, grind, or beep is a hard drive. A 990 Pro does not have heads. Power a clicking drive off. That failure is the clicking-drive page.

A Phison stick that identifies as SATAFIRM S11 is a different controller in ROM mode. Pulling the NAND packages off the board is the wrong first move on a modern NVMe, which is the chip-off page. Do not heat the packages because Magician showed a bad health percentage.

BitLocker is a different lock. A 48-digit key opens ciphertext. It does not refill spare. Where Windows may have turned encryption on without a clear ask is the Device Encryption page. If the drive is already unhealthy, the order is image first, then unlock the copy.

A warranty replacement is a different drive. Samsung's February 13, 2023 post points at Samsung's warranty terms. An exchange does not return the files that were on this one. Do not ship the only copy to a warranty depot as if that were the recovery. Get the data off, or have it evaluated, before the stick leaves for a swap.

Ordinary wear still exists. A 990 Pro that has been on 1B2QJXD7 or a later string for years, with a large amount of data written, can report Available Spare at 0 because the spare was used up. Puget's records were not showing the 980 Pro failure on 4B2QGXA7 or 5B2QGXA7. A stick already on the post-fix firmware is not explained by "you never installed the February 2023 update." The copy-first rule is the same either way. We will not tell you from a model number which of those you have.

What to do right now

1. Stop the fixes. Close Samsung Magician. Do not start a firmware update, a bootable firmware ISO, Secure Erase, PSID Revert, Over Provisioning, Performance Optimization, or a diagnostic scan. Do not clear read-only, initialize, format, or run CHKDSK.
2. If a SMART window is already open, photograph it. Firmware string, Available Spare, Available Spare Threshold, Percentage Used, critical warning, model, and capacity. Then leave the numbers alone.
3. If the drive still reads and the files are nowhere else, copy the ones that matter, once, to different media. Open a few of them on that other media before you do anything else to the original. If the copy errors, the drive stalls, or it disappears, stop. Do not retry with a cloner.
4. If it is read-only, copy what opens and do not try to make it writable. Read-only is the remaining way to read it. A firmware update will not reverse a 980 Pro Puget already called failed, and it will not restore 990 Pro SMART values Samsung said are not restored.
5. If it does not enumerate, or the copy cannot finish, power it down. One careful removal from the laptop is enough if you can do that without guessing. Do not seat it in a string of USB enclosures. Pack it the way you would any failed drive — how to ship a failed drive safely. If Windows asked for a BitLocker key, bring the key. Do not unlock the drive to "make the copy easier" before it leaves.

If you already have a second copy, you have opened it, and the drive is still on 0B2QJXD7 or 3B2QGXA7 and still reads and writes normally, the 2023 updates were the preventative step Puget and Samsung described. Magician's guide still says to back up first, because the flash can lose data. That is a different moment from a drive at 0% spare with one copy of the files.

How a lab handles it

We do not flash the patient because Magician is offering a newer string. We do not Secure Erase it, type the PSID into a revert box, over-provision it, TRIM it, initialize it, format it, or run CHKDSK against the original. We do not clear a read-only lock so the bench can write.

If the controller will identify, the drive is imaged write-blocked first, as far as it will read, and the file work happens on the copy. If it will not identify, or it drops as soon as it is read, the evaluation is why the controller is refusing. That is the job on the SSD not detected page. It is not a Magician session, and it is not a reason to lift the NAND. A BitLocker key, if you have one, is used on the image.

What that cannot do: put retired spare blocks back, reset SMART values Samsung said the firmware update does not restore, or rebuild a volume a finished Secure Erase or PSID Revert already wiped. A firmware update that was interrupted is its own damage. We will tell you if the copy shows that. We will not promise that a flash the manual warned could lose data left the volume untouched.

The public case log has no Samsung 990 Pro or 980 Pro with this failure. We will not invent one. We don't advertise a success rate, and we can't promise a turnaround from the model string. The evaluation is at Desert Data Recovery, in the Phoenix area. The number on this page is (602) 686-2622. For the work itself, see SSD data recovery.