Question CrystalDiskInfo: how to decipher Total Disk Writes value?

[stason99]

How can I find out the total TB written for this Chinese SSD ? What does F1 value of 22E229 mean?

https://imgur.com/a/WBRBmK4

 

[USAFRet]

TBW is irrelevant.
Do you currently own this?

Use it until it dies.

Which might be tomorrow, or 5 years from now.

 

[stason99]

Yes, I own this. I do want to know TB written. Any way to find that out?

 

[USAFRet]

Yes, I own this. I do want to know TB written. Any way to find that out?

If the unknown crappy SSD does not expose this info, probably can't be found.

 

[fzabkar]

"22E229" is the hexadecimal raw value. CrystalDiskInfo can be switched between decimal and hexadecimal modes.

Alternatively, use Google's calculator:

https://www.google.com/search?q=0x22E229+in+decimal

0x22E229 = 2286121 (decimal)​

You now need to determine the units of measurement. To this end, copy a 1GB file to the drive and record the before-and-after raw values of attribute F1. Divide 1GB by the difference in the raw values. That will tell you if each increment in F1 corresponds to 512 bytes or 1KB or 1MB or something else.

I suspect that attribute AA is the average erase count, ie 0x18 = 26. If so, then the total writes to NAND would be 3.3TB:

https://www.google.com/search?q=128+GiB+x+0x18+in+TB

This means that the total host writes would be at least that much less than that, allowing for write amplification.

I'm betting that each increment in F1 corresponds to 2MB.

 

[stason99]

"22E229" is the hexadecimal raw value. CrystalDiskInfo can be switched between decimal and hexadecimal modes.

Alternatively, use Google's calculator:

https://www.google.com/search?q=0x22E229+in+decimal

0x22E229 = 2286121 (decimal)
You now need to determine the units of measurement. To this end, copy a 1GB file to the drive and record the before-and-after raw values of attribute F1. Divide 1GB by the difference in the raw values. That will tell you if each increment in F1 corresponds to 512 bytes or 1KB or 1MB or something else.

Thank you! I followed your advice, changed to decimal, copied 1GB and the value changed about 1000 which means that the unit of measurement is Mb. The current TBW is 2.28Tb. Do you have any idea what this 128Gb SSD may be rated for in terms of TBW?

This is the disk in question. Do you know what endurance can it have? I can't find that info in it's specs.

 

[USAFRet]

Thank you! I followed your advice, changed to decimal, copied 1GB and the value changed about 1000 which means that the unit of measurement is Mb. The current TBW is 2.28Tb. Do you have any idea what this 128Gb SSD may be rated for in terms of TBW?

This is the disk in question. Do you know what endurance can it have? I can't find that info in it's specs.

The TBW number is only for the warranty.
Just like the year value.

Say it is 3 years, 100TBW.
All that means if it goes over either of those values, the manufacturer won't give you a new one if it dies.

It does NOT mean it WILL die at that age or TBW.

Yours at a supposed 2.2TB so far...a trivial amount.
At 14k+ hours, almost 2 years of 24/7. But the actual purchase age is much older? It is already out of warranty, no matter the TBW usage.

Use it until it dies. Replace it when it does.

 

[stason99]

It does NOT mean it WILL die at that age or TBW.

This is not what I've understood about TBW but I could be wrong. actually it has been on 24/7 for 8 years.

 

[USAFRet]

This is not what I've understood about TBW but I could be wrong. actually it has been on 24/7 for 8 years.

14436 Power On Hours = 601.5 days.
Under 2 years of 24/7.

 

[MWink64]

I suspect that attribute AA is the average erase count, ie 0x18 = 26. If so, then the total writes to NAND would be 3.3TB:

https://www.google.com/search?q=128+GiB+x+0x18+in+TB

This means that the total host writes would be at least that much, allowing for write amplification.

I'm betting that each increment in F1 corresponds to 2MB.

Judging by the firmware and SMART attributes I'm guessing it may be an old Phison controller. Phison S11 drives are just slightly different. On them, attribute AA is Bad Block Count. F1 (Host Writes) is usually in GB but that's obviously not the case here. I've seen this attribute correspond to 512 byte sectors on SMI based drives and 32MB on Realtek drives. I've never seen it equal 2MB.

Thank you! I followed your advice, changed to decimal, copied 1GB and the value changed about 1000 which means that the unit of measurement is Mb. The current TBW is 2.28Tb. Do you have any idea what this 128Gb SSD may be rated for in terms of TBW?

This is the disk in question. Do you know what endurance can it have? I can't find that info in it's specs.

As others have mentioned, it's out of warranty so it doesn't really matter. I don't know what type of NAND it has (MLC or TLC, planar or 3D) but I'd venture a guess they warranted it for somewhere in the range of 30-50TB. Keep in mind that host writes do not take into account write amplification or other factors that will greatly impact actual NAND wear. For what it's worth, I recently encountered a 128GB SK Hynix (TLC) drive that had 27TB of writes and claims to have 95% health remaining. If your drive has 2.3TB of writes after 8 years, you're not going to have to worry about the NAND wearing out in this lifetime.

 

[fzabkar]

This is the disk in question. Do you know what endurance can it have? I can't find that info in it's specs.

I can't find any TBW spec, either:

https://silicon-power.com/web/product-Slim_S60

The product brochure mentions MLC NAND and SLC cache, so it should have good endurance.

The web site has a firmware update dated 2018/03, so you might want to consider that. The included INI file mentions Phison.

BTW, I goofed when I said that the host writes would be greater than NAND writes. Obviously, the reverse is true.

 

[fzabkar]

Judging by the firmware and SMART attributes I'm guessing it may be an old Phison controller. Phison S11 drives are just slightly different. On them, attribute AA is Bad Block Count. F1 (Host Writes) is usually in GB but that's obviously not the case here. I've seen this attribute correspond to 512 byte sectors on SMI based drives and 32MB on Realtek drives. I've never seen it equal 2MB.

I goofed when I said that the host writes would be greater than NAND writes. Obviously, the reverse is true. Therefore, I should have rounded downwards, in which case I would have arrived at 1MB rather than 2MB. Thanks for the heads up.

 

[Pextaxmx]

SiliconPower is a legit Taiwanese brand. It's not like no-name brand Chinese drives.

 

[fzabkar]

Could attribute AD be min / max / average erase count?

0x16 / 0x90 / 0x58 --> 22 / 144 / 88​

If so, then the write amplification would be quite high.

NAND writes / host writes = (128 GiB x 0x58) / 2.28 TB = 5.3​

 

[MWink64]

Could attribute AD be min / max / average erase count?

0x16 / 0x90 / 0x58 --> 22 / 144 / 88​

If so, then the write amplification would be quite high.

NAND writes / host writes = (128 GiB x 0x58) / 2.28 TB = 5.3​

Make of this what you will. Here's some select SMART info from a 240GB Phison S11 based drive I've encountered:

From Gsmartctl:

170 Bad_Blk_Ct_Erl/Lat PO---- 085 085 010 - 0/278
173 MaxAvgErase_Ct -O--C- 100 100 000 - 75 (Average 39)
231 SSD_Life_Left PO--C- 100 100 000 - 96
241 Lifetime_Writes_GiB -O--C- 100 100 000 - 6379

CrystalDiskInfo gives attribute AD (Erase Count) a decimal value of 2555979 (which appears to convert to 27004B in hex).

 

[fzabkar]

0x27004B --> 0x27 / 0x4B --> 39 / 75

It doesn't look quite the same, but SMART is hardly ever consistent.

I haven't seen "Lat PO" before, though.

This document seems to explain it (page 2):

https://media.kingston.com/support/downloads/MKP_521_Phison_SMART_attribute.pdf

170 = “Bad Block Count (Early / Later)”

​

Raw Value Byte [1~0]: Early bad block count​

Raw Value Byte [5~4]: Later bad block count​

173 = Erase count (Average, Max Erase Count)

Raw Value Byte [1~0]: Max erase count​

Raw Value Byte [3~2]: Average erase count​

​

I suspect that "early" corresponds to the factory defects that the SSD shipped with, while "later" corresponds to grown defects that have developed during use.

241 (F1h) = Host Writes (Sector Unit)

Raw Value Byte [5~0]: Host write increments in 1GB values​

 

[MWink64]

0x27004B --> 0x27 / 0x4B --> 39 / 75

It doesn't look quite the same, but SMART is hardly ever consistent.

I haven't seen "Lat PO" before, though.

You're telling me! The inconsistency of SMART attributes annoys me to no end.

It's hard to read the Gsmartctl data when it's presented this way. "Bad_Blk_Ct_Erl/Lat" and "PO----" are different sections.

P = prefailure warning
O = updated online

I think "Bad_Blk_Ct_Erl/Lat" is supposed to be Bad Block Count Early/Later. However, the Early/Later part is reversed. This drive had 278 bad blocks early (when manufactured) and supposedly has not accumulated any more while in service (assuming you trust what SMART reports).

 

[stason99]

14436 Power On Hours = 601.5 days.
Under 2 years of 24/7.

Then the SMART data reported is wrong, because the drive is powered on for 8 years.

The web site has a firmware update dated 2018/03, so you might want to consider that. The included INI file mentions Phison.

Thanks, I'll try updating it.

 

[fzabkar]

Then the SMART data reported is wrong, because the drive is powered on for 8 years.

Maybe the POH count is not incremented when your machine goes to sleep?

 

[stason99]

Maybe the POH count is not incremented when your machine goes to sleep?

It's set to never going to sleep because it's a surveillance cameras system.