SSD reporting 80TB of written/read data (mistake?)

I am considering buying a used tablet/pc with a SD9TN8W-256G-1006 SSD. According to CristalDiskInfo though, it has 80TB of written and read data (the disk's 'maximum' is at 100TB I think, but I am not sure), 2325h of work, 83 shutdowns. It's hard to imagine how all that data could have been used in such a short time. SanDisk Dashboard reports the Life Remaining at 100%.

Is the 80TB value false ?

CristalDiskInfo

SanDisk SSD Dashboard

Speccy (also reports a similar number of read/written data, but I am not sure is they are Terabytes)

alfred

Posted 2018-12-04T17:30:03.897

Reputation: 363

Host reads and writes are pretty much meaningless, only writes that get through to the flash matter. 609GB of total NAND writes is a meaningful number. – David Schwartz – 2018-12-04T19:16:29.577

These numbers don’t add up either way, because the unit of the raw value of attributes F1 and F2 is supposed to be “LBAs”, ie. 512-byte sectors. Which means 0x1386a LBAs = 40,948,736 bytes = 39 MiB. – Daniel B – 2018-12-04T19:37:26.993

I'll try writing 100MB on the disk and see what happens to that number. I just noticed I am supposed to have written 200GB between those two screenshots – alfred – 2018-12-04T19:47:19.280

CrystalDiskInfo reads now 82.3TB Read and 82.7TB Write. There is something wrong with the way S.M.A.R.T reports the data, or there is some loop (not credible) – alfred – 2018-12-05T09:30:27.617

Answers

Since it is the drive reporting this information, there isnt much we can do to prove that is it is false. However, we can look at the information given and draw our own conclusions.

2325 hours is just a hair shy of 100 days. At 80TB over that 100 days, that means there was an average of 800GB read and written per day. While that number is not out of the realm of possibility, it is an extremely high number and an unlikely probability. Looking at the drive's datasheet from SanDisk, I see no maximum writes listed, so I am not aware of the 100TB max you mentioned. Additionally, the SanDisk software reports the drive is at 100% health.

In my opinion, something is not adding up. If there has been that much usage, the drive should not be at 100% health. It would appear the SMART data being reported is incorrect. Whether its the 80TB, the 100% health, or both I cannot say.

You could contact SanDisk support and ask them about the information. They might be aware of any firmware bugs on the drive that misrepresent the data, or perhaps they have a better tool you can use to analyze the data.

Keltari

Posted 2018-12-04T17:30:03.897

Reputation: 57 019

I just wish to draw attention to the attribute of Uncorrectable Soft Read Error Rate whose value is 998, and is defined as:

Number of soft read errors that cannot be fixed on-the-ﬂy and requires deep recovery provided by RAISE.

R.A.I.S.E. is a complementary technology to Error Correcting Code (ECC).

This means that 998 bits on the disk are bad, but the data was recovered thanks to Error Correcting Code techniques.

Another suspicious attribute is Program Fail Count whose value is 480.149. I don't know exactly what this means, perhaps it has been normalized, but this is defined as:

The number of times when write to a flash memory failed. The write process is technically called "programming the flash memory", hence the attribute name. When the flash memory is worn out, it cannot be written to any longer and becomes read-only. The Raw value shows the actual number of failures.

There is also Erase Fail Count, value 768, defined as:

S.M.A.R.T. parameter indicates a number of flash erase command failures.

I think therefore that even if the other values don't make sense, this SSD seems to have been hard used and might be on the verge of failure.

I can't recommend buying it, unless for light use. It's in my opinion not trustworthy, as there are too many warning signs.

harrymc

Posted 2018-12-04T17:30:03.897

Reputation: 306 093

0xB5 can refer to "Non-4K Aligned Access Count" (wiki), which makes this value sound reasonable. – Attie – 2018-12-04T19:04:37.180

1@Attie: Possible, but nowadays is less likely, as partitions are usually well-aligned. That is, unless this tablet is really old, which is another reason to avoid it. – harrymc – 2018-12-04T19:10:31.440

@harrymc The program fail block count is zero. Many drives use attribute B5 for other purposes. I believe this drive uses it to report the status of an internal diagnostic on the backup capacitor that allows it to flush writes after a loss of power. 5 recoverable read errors is nothing to worry about. – David Schwartz – 2018-12-04T19:54:29.633

@DavidSchwartz: I had a typing error, the real number is 998. – harrymc – 2018-12-04T20:05:37.330

To put it another way... ~80 TB written over 2,325 hours is a sustained average of ~10 MB/s - over the SSD's entire lifetime. This is high, but not impossible.

To give you a feel for it, my workstation only averages ~160 kB/s over a 24 hour period.

Cherry picking a particularly write-intensive day still only lands me at an average of ~870 kB/s over a 24 hour period (71.5 GB written).

It's not impossible to see the 80 TB figure you've quoted, but I'd be suspicious that either:

Something is misreporting ("powered on hours" or "bytes written" - SMART isn't perfect)
The previous owner used it heavily from a storage point of view

To give further credence to the "something is misreporting" hypothesis, the "Total LBAs Written" (0xF1) value of 79,978 implies an absolute maximum of ~312 MB (if we assume 4096 byte sectors)... The value is stored as 48-bit, so even 80 TB wouldn't have caused it to wrap. But, as mentioned above, SMART isn't perfect, and it certainly isn't implemented the same from manufacturer to manufacturer.

0xF9 ("Total NAND Writes. Raw value reports the number of writes to NAND in 1 GB increments.") however suggests ~609 GB written over the SSD's lifetime, which is quite a reasonable figure.

Even if the 80 TB figure is accurate, modern SSDs are expected to last significantly longer than people originally anticipated... This SSD range was announced in Jan 2018. A rough estimate is "a few thousand times the capacity", or ~750 TB in your case.

The datasheet's claimed "Endurance" of 100 TBW is more likely to be related to conservative reliability and warranty than a prediction of "it will fail as soon as it passes 100 TBW".

Attie

Posted 2018-12-04T17:30:03.897

Reputation: 14 841