Tool AT-W22REV. 2026-A
Hard drive check: what do the SMART values say about your drive?
A few values matter most: for hard drives and SATA SSDs, the reallocated, pending and uncorrectable sectors; for NVMe SSDs, above all the critical warning and the media errors. If any of them is above zero, back up your data. Paste the text from CrystalDiskInfo or smartctl; the check translates each value and tells you: good, monitor or back up now. A good result is, however, no guarantee.
- free
- no sign-up
- reads in your browser
01 / CHECK
Paste the values, read the plain-language explanation and verdict
Check · SMART values of hard drive and SSD
The check reads the values your drive reports about itself and rates them according to studies and manufacturer data. It does not predict a failure and is no substitute for a backup.
Rules and studies as of 14 November 2026. If a drive has already failed, continue with data recovery.
Read in your browser. Nothing is sent or stored.
02 / READING THE VALUES
How to get the SMART values
Every hard drive and every SSD keeps a record of itself: power-on hours, temperature, read errors, spare sectors. This is called SMART (Self-Monitoring, Analysis and Reporting Technology). Windows shows almost none of it. On Windows, the easiest way to read the values is the free CrystalDiskInfo; on Linux and macOS, the command-line program smartctl from the smartmontools package, which is also available for Windows.
| System | How to do it | Command |
|---|---|---|
| Windows | Start CrystalDiskInfo (crystalmark.info), select the drive, press Ctrl+C | Bearbeiten → Kopieren |
| Windows | Install smartmontools, open the command prompt as administrator | smartctl -x /dev/sda |
| Linux | smartmontools package, in the terminal | sudo smartctl -x /dev/sda |
| Linux, NVMe | as above, the SSD is called nvme0 | sudo smartctl -x /dev/nvme0 |
| macOS | smartmontools via Homebrew, list drives with diskutil list | sudo smartctl -x disk0 |
| USB enclosure | many adapters need the SAT passthrough | smartctl -x -d sat /dev/sdb |
- CrystalDiskInfo copies all drives at once. Ctrl+C in the program window is enough; the check separates the drives itself. German attribute names are no problem, as values are read by ID (05, C5, C6 …).
- smartctl -x instead of -a. The detailed output also contains the standardised device statistics, the temperature log and the drive’s temperature limits. The check evaluates all of this, including several runs one after another.
- USB enclosures, NAS, RAID controllers. Many USB adapters only pass SMART through with
-d sat, some not at all. Behind a RAID controller, smartctl requires specific options such as-d megaraid,0. A NAS usually shows the values in its storage manager; from there they can be entered by hand.
03 / HARD DRIVE AND SATA SSD
The values that matter for hard drives and SATA SSDs
Of the roughly thirty attributes on a hard drive, a few carry almost the whole message. The first large field study comes from Google: after the first reallocation, drives failed more than 14 times as often over the following 60 days as drives without one, and after the first sector “on probation” 16 times Pinheiro et al. 2007. In each case the critical threshold was one. A second study with around a million drives shows what happens next: without reallocated sectors, 1.7 per cent failed within 60 days, with more than 40 over half Ma et al. 2015.
| Value | Monitor | Back up now |
|---|---|---|
| Reallocated sectors (5) | from 1 | above 40 for hard drives |
| Pending sectors (197) | – | from 1 |
| Offline uncorrectable (198) | – | from 1 |
| Reported uncorrectable errors (187) | – | from 1 |
| Command timeouts (188) | from 1, usually power or cable | – |
| UDMA CRC errors (199) | from 1, usually the cable | – |
| Temperature (194), hard drive | from 50 °C | from 60 °C |
| Temperature (194), SATA SSD | from 60 °C | from 70 °C |
| Normalised value at or below threshold (any attribute) | once in the past (In_the_past) | now (FAILING_NOW) |
| Last self-test | older test with error | most recent test with error |
| Life used (SSD) | 90 % or more | – |
| Written against TBW (SSD) | 90 % or more | – |
As of 14 November 2026. If the drive supplies its own temperature limits (device statistics or SCT with smartctl -x), these apply.
Fig. 1 · Figures from Pinheiro et al. 2007. The surface scan is not a standard attribute; reallocation and pending sectors correspond to attributes 5 and 197.
- Reallocated sectors (5). The drive has replaced defective sectors with spares. A few can stay stable for years, but the risk is raised from the first one, and the more there are, the shorter the remaining time Ma et al. 2015. Hence: monitor from 1, back up now for hard drives above 40. On SSDs, the same attribute is often called “reallocated blocks”; there, a few defective blocks are followed by hundreds in half of cases Schroeder et al. 2016.
- Pending and offline uncorrectable sectors (197, 198). These sectors cannot be read at the moment. Such errors cluster in space and time, and a drive with one is more likely to get more Bairavasundaram et al. 2007. Without a second copy, a single one can already cost data Schroeder et al. 2010. From 1, the advice is therefore: back up now.
- Reported uncorrectable errors (187). Data that error correction could no longer save. Here too, the next error is much more likely after the first Schroeder et al. 2016.
- CRC errors and timeouts (199, 188). Usually it is not the drive but the connection. In the Google study, CRC errors pointed more towards cables and connectors Pinheiro et al. 2007, and connections caused a considerable share of all failures in large storage systems Jiang et al. 2008. Replace the cable first, then monitor.
An example from the check, the Seagate drive under “Load example”: it reports “PASSED” itself. At the same time, the list shows 184 reallocated and 16 pending sectors, and the last self-test aborted with a read error. Verdict: back up now. The drive’s overall verdict alone would have revealed none of this.
04 / NVME SSD
NVMe SSDs: reading the health log
NVMe SSDs do not have an attribute list but a standardised health log (SMART/Health Information, Log 02h). That is more convenient: the fields mean the same for every manufacturer. smartctl prints them with English names, CrystalDiskInfo as lines 01 to 0F.
| Field | Monitor | Back up now |
|---|---|---|
| Critical Warning | bit 1 only (temperature) | any other bit |
| Available Spare | below 100 % | below the Available Spare Threshold |
| Percentage Used | 90 % or more | – |
| Media and Data Integrity Errors | – | from 1 |
| Temperature | from the Warning Comp. Temp. Threshold, otherwise from 70 °C | from the Critical Comp. Temp. Threshold, otherwise from 80 °C |
| Unsafe Shutdowns, Error Log Entries | note only | – |
| Written against TBW | 90 % or more | – |
| Bit | Value | Meaning |
|---|---|---|
| 0 | 0x01 | Spare below the threshold |
| 1 | 0x02 | Temperature above or below the limit |
| 2 | 0x04 | Reliability degraded by media or internal errors |
| 3 | 0x08 | read-only now (write protection) |
| 4 | 0x10 | Backup of the volatile memory failed |
| 5 | 0x20 | Persistent Memory Region read-only or unreliable |
- Media and Data Integrity Errors. Counts data the SSD could not recover. Between 20 and 63 per cent of the SSDs at Google had at least one such error within four years; after one, the probability of another in the following month was almost 30 instead of 2 per cent Schroeder et al. 2016. So from 1, back up now, even if the SSD may well keep running for a long time.
- Available Spare. The reserve of spare blocks in per cent. If it falls below the threshold (Available Spare Threshold), the SSD itself sets the critical warning. If it drops below 100 %, the SSD has already replaced blocks; the check then says “monitor”.
- Percentage Used. The manufacturer’s wear estimate. According to the specification it may rise above 100, and 100 does not necessarily mean the SSD fails. In the field, errors did not suddenly rise beyond the cycle limit Schroeder et al. 2016, and failures did not follow wear evenly Meza et al. 2015.
- Unsafe Shutdowns. How often power was lost without the system having unmounted the SSD. In the lab, 14 of 17 SSDs lost data in such power failures, up to complete failure Zheng et al. 2016. A few are normal, for example after a crash. If they pile up, shut down cleanly and, for a computer that runs around the clock, provide a UPS; how large it needs to be is worked out by the UPS calculator.
- Error Information Log Entries. Also counts harmless cases, such as commands the operating system tries and the SSD does not know. High numbers without media errors are not a warning sign.
- Data Units Written. In thousands of 512 bytes: a value of 1 is 512,000 bytes. smartctl converts to TB itself in square brackets.
05 / RAW VALUE AND THRESHOLD
Normalised value, threshold, raw value: why “PASSED”
Every ATA attribute has two numbers. The normalised value (VALUE, “Cur” in CrystalDiskInfo) usually runs downwards from 100 or 200. The manufacturer sets a threshold (THRESH, “Thr”). If the normalised value is at or below the threshold, the attribute counts as failed; smartctl then writes FAILING_NOW, or In_the_past if it was once the case. The raw value (RAW_VALUE) is the actual count, such as the number of reallocated sectors.
The catch: how the normalised value is derived from the raw value is decided by each manufacturer. Since ATA/ATAPI-4, the meaning of the attributes has been entirely vendor-specific, and the command for reading the attributes has been marked obsolete since ACS-4, according to the smartctl documentation. On Seagate, raw values in the millions are normal for attributes 1, 7 and 195, and attribute 188 packs three counters into one number. The check recognises this and rates such attributes by the threshold only.
The thresholds themselves are set generously so that healthy drives do not trigger false alarms. In an evaluation of 3744 drives, the usual threshold rule detected only a small share of later failures; statistical methods on the same values found three to four times as many Hughes et al. 2002. That is why the check looks at the raw values of the important attributes and not just the overall verdict “PASSED”.
| decimal | hex | Value |
|---|---|---|
| 5 | 05 | Reallocated sectors |
| 9 | 09 | Power-on hours |
| 187 | BB | Reported uncorrectable errors |
| 188 | BC | Command timeouts |
| 194 | C2 | Temperature |
| 197 | C5 | Pending sectors |
| 198 | C6 | Offline uncorrectable |
| 199 | C7 | UDMA CRC errors |
06 / TEMPERATURE AND AGE
Putting temperature, power-on hours and write volume in context
Temperature. Heat does less harm than often thought. Below 50 °C, failures and sector errors rose only slowly with temperature; large fluctuations had more effect El-Sayed et al. 2012. At Google too, an effect showed only at the upper end and for older drives Pinheiro et al. 2007. At Microsoft, however, temperature was more strongly linked to failures than utilisation, and installation location and fans mattered more than load Sankar et al. 2013. For hard drives, the check therefore says “monitor” from 50 °C and “too hot” from 60 °C, the upper limit on many data sheets. For SATA SSDs, 60 and 70 °C apply. NVMe SSDs usually report their own limits; there too, higher temperatures went with more failures Meza et al. 2015.
Power-on hours. Data sheets quoting 1 to 1.5 million hours mean time between failures lead you to expect at most 0.88 per cent failures a year; in the field, mostly 2 to 4 per cent of drives were replaced, and the rates rose steadily with age Schroeder and Gibson 2007. From 43,800 hours, five years of continuous operation, the check recommends planning the replacement. Conversely, youth is no protection: with SSDs, a conspicuous number failed early Alter et al. 2019.
Write volume and TBW. SSD manufacturers state in the data sheet how many terabytes may be written during the warranty (TBW). The check sets the amount written against this and says “plan replacement” from 90 per cent. This remains a rule of thumb, because what the operating system writes does not accurately reflect the wear of the cells Meza et al. 2015. CrystalDiskInfo gives the amount under “Host Writes” in binary gigabytes; the check converts it to decimal terabytes, as they appear in the data sheet.
07 / METHOD
What to do depending on the result
Verdict
Good
- Check your backup: is there a current copy on a second device?
- Read the values again in six months, more often for older drives.
Verdict
Monitor
- Back up today, not at the weekend.
- Read the values again in one to two weeks and compare. If a counter keeps rising, replace the drive.
- For CRC errors, replace the cable and port first, then check again.
Verdict
Back up now
- Copy the most important things first (documents, photos, mailbox), not the whole drive in one go.
- Do not run checks or repairs such as chkdsk /r, and do not defragment: they read the whole drive and can cost you the rest.
- If the drive clicks or grinds, switch it off immediately. Then only data recovery can help.
- Afterwards, replace the drive and move the system to the new one.
Rising values say more than a single reading. Predictions from SMART values work best with the trend over days and weeks Botezatu et al. 2016. For “monitor”, therefore, read the values again after one to two weeks and compare. What a backup should look like is described by the 3-2-1 rule. If several drives sit in a NAS, the NAS and RAID calculator shows how long a rebuild takes when one is replaced.
Where the check stops
- Many failures come without warning. Over 56 per cent of the drives that failed at Google showed nothing in the four strong values, 36 per cent in none at all Pinheiro et al. 2007. “Good” therefore does not mean “safe”.
- Slow is also a fault. Drives and SSDs can become noticeably slow without any value triggering Gunawi et al. 2018. If the computer hangs when opening files, the drive should be checked despite good values.
- A snapshot. The check sees a reading, not a trend. Only comparing two readings shows whether a counter is rising.
- Vendor-specific data. Attributes the check does not know are rated only by the manufacturer’s threshold. Older SSD controllers with their own raw-value format (8 bytes in CrystalDiskInfo) it cannot read.
If the drive clicks, is no longer detected or files are already missing, reading values no longer helps. Then follow the advice in the Hard drive failed guide: switch it off and do not keep trying. We bring the data back through data recovery. You can get the new drive through hardware ordering, and how quickly a business is working again after such a failure is checked by the IT emergency check.
08 / RESEARCH
What the research says
The thresholds in the check come from field studies of hundreds of thousands of hard drives and SSDs in data centres. Five findings underpin the most important decisions; the full list, with a review note, is further down.
- The first error counts. For reallocated and pending sectors, the critical threshold was one Pinheiro et al. 2007; with more than 40 reallocated sectors, more than half of the drives failed within two months Ma et al. 2015.
- Errors rarely come alone. Sector errors cluster in space and time Bairavasundaram et al. 2007; with SSDs, an uncorrectable error is often followed by the next Schroeder et al. 2016.
- The built-in warning is too sluggish. Manufacturer thresholds detect only a small share of failures Hughes et al. 2002.
- Not every warning concerns the drive. A large share of storage failures traces back to connections Jiang et al. 2008.
- Temperature matters, but moderately. Below 50 °C the effect is small El-Sayed et al. 2012; installation and ventilation matter more than load Sankar et al. 2013.
09 / 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_festplatten_check- Transport
- Streamable HTTP, read-only
| Name | Required | Meaning |
|---|---|---|
text | no | Full text output of smartctl -a or smartctl -x (ATA/SATA and NVMe, also several runs one after another) or the text copied from CrystalDiskInfo (Ctrl+C, all drives). Pass it on unchanged; do not summarise. |
hand | no | Alternative to text: individual values, if only a screenshot or a list is available. Give known values only. |
tbw_tb | no | Optional: the manufacturer’s TBW rating in TB from the data sheet. Applies to every SSD in the text; for several SSDs, run the check separately. |
Questions where an assistant should call it
- “CrystalDiskInfo shows ‘Caution’ for my hard drive. Here is the copied text: how bad is it?”
- “What do these smartctl values of my NVMe SSD mean, in particular Percentage Used and Unsafe Shutdowns?”
- “My drive has 8 Current Pending Sectors and 0 Reallocated Sectors. Do I need to replace it?”
- “The UDMA CRC errors on my SSD are rising. Is the SSD faulty?”
- “My Samsung SSD has written 180 TB and the data sheet says 600 TBW. How much life is left?”
How the connection works and what MCP can do in your own business is explained on the page about MCP integrations.
10 / COMMON QUESTIONS
Common questions about SMART values
What does “Reallocated Sector Count” mean?
The drive has replaced sectors it could no longer read or write reliably with spare sectors. The raw value counts them. Zero is normal. In a large field study, from the very first one the risk of the drive failing within the next 60 days rose to more than 14 times. With more than 40 reallocated sectors, over half of the drives in a second study failed within 60 days. So the check says “monitor” from 1 and “back up now” for hard drives above 40.
What is a pending sector (Current Pending Sector)?
A sector that the drive cannot read at the moment and has therefore flagged. If it is later overwritten successfully, it disappears from the counter; otherwise it is reallocated. Until then, whatever was stored there cannot be read. On a PC with a single drive, that can mean a damaged file. That is why the check says “back up now” from 1.
My UDMA CRC errors keep rising. Is the drive faulty?
Usually not. The counter (attribute 199) records transmission errors on the way between drive and motherboard, that is, the cable, connector or port. In the Google study, such errors pointed more towards the cable than the drive. The counter never goes down. Replace the SATA cable, plug it into a different port and check after a few days whether the number is still rising.
CrystalDiskInfo shows “Caution”. What now?
CrystalDiskInfo warns from the first reallocated, pending or uncorrectable sector, and for SSDs at 10 per cent remaining life. That is deliberately cautious. Back up your data first. Then paste the copied text (Ctrl+C in the program) into the check: it tells you which value triggers the warning and how serious it is.
After how many operating hours should you replace a hard drive?
There is no fixed limit. In the field, replacement rates rise steadily with age, not only after five years, and mostly 2 to 4 per cent of drives were replaced each year. After around 43,800 hours, that is five years of continuous operation, it is worth planning the replacement rather than waiting for a failure. For drives that only run during the day, that is considerably more calendar years.
What does “Percentage Used” mean on an NVMe SSD?
The manufacturer’s estimate of how much of the drive’s life has been used up. According to the NVMe specification, 100 per cent does not necessarily mean the SSD fails; the value may rise above 100. The check advises planning a replacement from 90 per cent. With SSDs, media errors and the spare reserve matter more than wear.
Why is the raw value on my Seagate drive so huge?
For some attributes, Seagate packs several counters into one number. For the read error rate (1), the seek error rate (7) and the corrected ECC errors (195), values in the millions are normal. The meaning of raw values is not standardised; only the normalised value against the manufacturer’s threshold counts there. The check therefore does not judge these attributes by their raw value.
Where does the hard drive check stop?
It reads what the drive reports about itself. In a large study, more than a third of the drives that failed had shown nothing in any single value beforehand. A drive can also become noticeably slow without any value triggering. If a drive clicks, drops out or data is already missing, the check cannot help. In that case we back up whatever is still readable, in Rodewisch and across the Vogtland, or it goes to data recovery.
11 / STUDIES AND SOURCES
What the tool is based on
The guide values and notes above are based on these 15 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.
Pinheiro, E.; Weber, W.-D.; Barroso, L. A. (2007): Failure Trends in a Large Disk Drive Population. 5th USENIX Conference on File and Storage Technologies (FAST ’07), pp. 17–29
Finding Drives in Google’s data centres: after the first reallocation, drives failed over 14 times as often in the following 60 days, after the first sector “on probation” 16 times, after the first offline reallocation 21 times. CRC errors pointed more towards cables and connectors. Over 56 per cent of the drives that failed showed nothing in the four strong values, 36 per cent in none at all.
In the tool Even the first reallocated, pending or offline unreadable sector counts; CRC errors lead to the cable; and an unremarkable result explicitly does not mean “safe”.
www.usenix.org/conference/fast-07/failure-trends-large-disk-drive-population · Conference paper
Hughes, G. F.; Murray, J. F.; Kreutz-Delgado, K.; Elkan, C. (2002): Improved disk-drive failure warnings. IEEE Transactions on Reliability 51 (3), pp. 350–357
Finding The built-in SMART warning with fixed thresholds detects, with few false alarms, only a small share of later failures. Statistical tests on the same values found three to four times as many in a sample of 3744 drives, but still at most 40 to 60 per cent.
In the tool That is why the check rates the raw values itself and explains why “PASSED” means little.
doi.org/10.1109/TR.2002.802886 · Journal article
Ma, A.; Douglis, F.; Lu, G. et al. (2015): RAIDShield: Characterizing, Monitoring, and Proactively Protecting Against Disk Failures. 13th USENIX Conference on File and Storage Technologies (FAST ’15), pp. 241–256
Finding Around a million SATA drives over up to five years: without reallocated sectors, 1.7 per cent failed within 60 days; with more than 40 reallocated sectors, over half failed. The more sectors, the shorter the remaining time.
In the tool With more than 40 reallocated sectors, “monitor” becomes “back up now” for hard drives.
www.usenix.org/conference/fast15/technical-sessions/presentation/ma · Conference paper
Bairavasundaram, L. N.; Goodson, G. R.; Pasupathy, S.; Schindler, J. (2007): An analysis of latent sector errors in disk drives. Proceedings of ACM SIGMETRICS 2007, pp. 289–300
Finding 1.53 million drives over 32 months: 3.45 per cent developed undetected sector errors, larger drives more often. A drive with such errors was more likely to get more, and the errors occurred clustered in space and time.
In the tool Pending sectors and a self-test with a read error lead to “back up now”, because further errors often follow.
doi.org/10.1145/1254882.1254917 · 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 When a sector becomes unreadable, a single such error can cost data if there is no redundancy. The paper analyses field data on such errors and compares methods for finding them early.
In the tool In a PC with one drive, every unreadable sector can mean data loss; that is why 197 and 198 say “back up now” from 1.
doi.org/10.1145/1837915.1837917 · Journal article
Schroeder, B.; Gibson, G. A. (2007): Understanding disk failure rates: What does an MTTF of 1,000,000 hours mean to you?. ACM Transactions on Storage 3 (3), Art. 8
Finding Replacement data from over 110,000 drives: mostly 2 to 4 per cent were replaced each year, instead of the at most 0.88 per cent implied by data sheets. The rates rose steadily with age, not only after five years.
In the tool The text on power-on hours says that risk grows with age and that even a healthy drive needs a backup.
doi.org/10.1145/1288783.1288785 · Journal article
Jiang, W.; Hu, C.; Zhou, Y.; Kanevsky, A. (2008): Are disks the dominant contributor for storage failures? A comprehensive study of storage subsystem failure characteristics. ACM Transactions on Storage 4 (3), pp. 1–25
Finding Around 39,000 storage systems with 1.8 million drives: drives caused only 20 to 55 per cent of failures; connections and protocols accounted for a large part. Dual connections cut the failure rate by 30 to 40 per cent.
In the tool For CRC errors and timeouts, the check first advises looking at cable, connector and port.
doi.org/10.1145/1416944.1416946 · Journal article
El-Sayed, N.; Stefanovici, I. A.; Amvrosiadis, G. et al. (2012): Temperature management in data centers: why some (might) like it hot. Proceedings of ACM SIGMETRICS/PERFORMANCE 2012, pp. 163–174
Finding Field data from several data centres: below 50 °C, failures and sector errors rose only slowly and more or less linearly with temperature. Large temperature fluctuations had more effect.
In the tool For hard drives, “warm” starts at 50 °C; below that the check gives no temperature warning.
doi.org/10.1145/2254756.2254778 · Conference paper
Sankar, S.; Shaw, M.; Vaid, K.; Gurumurthi, S. (2013): Datacenter Scale Evaluation of the Impact of Temperature on Hard Disk Drive Failures. ACM Transactions on Storage 9 (2), pp. 1–24
Finding Tens of thousands of drives at Microsoft: temperature was more strongly linked to failures than utilisation. Installation location and fans influenced temperature more than load.
In the tool At high temperature, the check advises ventilation and installation location, not less use.
doi.org/10.1145/2491472.2491475 · Journal article
Schroeder, B.; Lagisetty, R.; Merchant, A. (2016): Flash Reliability in Production: The Expected and the Unexpected. 14th USENIX Conference on File and Storage Technologies (FAST ’16), pp. 67–80
Finding Six years of SSDs at Google: 20 to 63 per cent had at least one uncorrectable error within four years; after one, the probability in the following month was almost 30 instead of 2 per cent. 30 to 80 per cent developed bad blocks, and after 2 to 4, hundreds followed in half of cases. Beyond the cycle limit there was no sudden rise.
In the tool Media errors lead to “back up now”, a spare reserve that has been touched to “monitor”, and 100 per cent life used on its own does not lead to an alarm.
www.usenix.org/conference/fast16/technical-sessions/presentation/schroeder · Conference paper
Meza, J.; Wu, Q.; Kumar, S.; Mutlu, O. (2015): A Large-Scale Study of Flash Memory Failures in the Field. Proceedings of ACM SIGMETRICS 2015, pp. 177–190
Finding Almost four years of SSDs at Facebook: failure rates did not rise evenly with wear. Higher temperatures went with more failures; throttling mitigated this. What the operating system writes does not accurately reflect the wear of the cells.
In the tool The TBW calculation is a rule of thumb, not a measure of wear; for SSDs, heat is also rated.
doi.org/10.1145/2745844.2745848 · Conference paper
Zheng, M.; Tucek, J.; Qin, F. et al. (2016): Reliability Analysis of SSDs Under Power Fault. ACM Transactions on Computer Systems 34 (4), pp. 1–28
Finding 17 SSDs from six manufacturers, over 3000 deliberate power failures: 14 showed errors, from corrupted bits and half-written data to complete failure.
In the tool The “unsafe shutdowns” value is explained, with the advice to shut down cleanly and to use a UPS if power cuts are frequent.
doi.org/10.1145/2992782 · Journal article
Alter, J.; Xue, J.; Dimnaku, A.; Smirni, E. (2019): SSD failures in the field: symptoms, causes, and prediction models. Proceedings of SC ’19 (High Performance Computing, Networking, Storage and Analysis), pp. 1–14
Finding 30,000 SSDs in a Google data centre over six years: a conspicuous number of early failures. The causes were often different from those expected and hard to identify.
In the tool For young SSDs, the check explicitly says that few power-on hours are no protection.
doi.org/10.1145/3295500.3356172 · Conference paper
Gunawi, H. S.; Suminto, R. O.; Sears, R. et al. (2018): Fail-Slow at Scale: Evidence of Hardware Performance Faults in Large Production Systems. ACM Transactions on Storage 14 (3), pp. 1–26
Finding 114 reports from 14 institutions: drives, SSDs and other hardware can become slow without failing, with varying symptoms and long chains of causes.
In the tool Under the limits it says: a drive can become noticeably slow without any SMART value triggering.
doi.org/10.1145/3242086 · Journal article
Botezatu, M. M.; Giurgiu, I.; Bogojeska, J.; Wiesmann, D. (2016): Predicting Disk Replacement towards Reliable Data Centers. Proceedings of ACM SIGKDD 2016, pp. 39–48
Finding Over 30,000 drives, evaluated continuously for 17 months: from the SMART values, a necessary replacement could be detected 10 to 15 days ahead, with up to 98 per cent accuracy.
In the tool That is why, for warning signs, the check advises reading the values again after one to two weeks and comparing: the trend says more than a snapshot.
doi.org/10.1145/2939672.2939699 · Conference paper
Standards, authorities and manufacturer data
- NVM Express Base Specification 2.1: SMART / Health Information (Log 02h, Fig. 206) · NVM Express
- NVME_HEALTH_INFO_LOG: fields of the SMART/Health log · Microsoft Learn
- smartctl(8): attributes, normalised value, threshold, WHEN_FAILED · smartmontools
- CrystalDiskInfo: Health Status (rules for Good, Caution, Bad) · Crystal Dew World
- CrystalDiskInfo, source code of the copy function (DiskInfoDlgCopy.cpp) · hiyohiyo
- Data sheet BarraCuda 3.5: operating temperature up to 60 °C · Seagate
- Brochure 870 EVO/QVO: operating temperature 0 to 70 °C · Samsung
12 / 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?
Worried about the result? Send us the copied text. We back up your data while the drive is still running, fit a new one and move the system across, in Rodewisch and throughout the Vogtland.
+49 3744 4399760[email protected]Mon–Fri 8:00–17:00 · Straße der Jugend 5, 08228 Rodewisch