Bought a few used SSD

[tcsenter]

Have had pretty good luck so far getting deals on 'used' SSD drives several times, in terms of 1) working 2) modest/acceptable host or NAND writes. This time not so much...164TB!!




Three of them for $30 delivered. Judging by the power on hours count, which is a few hundred less than the SSD I am writing this on that only has 7.4TB host writes and 17.5TB NAND writes, I'm guessing these have seen some 24/7 server use. Probably wise to just toss them? Every other metric there is great.

 

[mikeymikec]

Do you have a use for them as backup drives?

 

[tcsenter]

That's what I was going to use them for. I acquired some spiffy USB external enclosures I wanted to make use of and was going to give them as tech gifts!

 

[fzabkar]

That number doesn't make sense. Show us all the attributes.

If attribute 0xA9 really is Max Erase Count, then the SSD has been completely erased a maximum of 58 (= 0x3A) times. That corresponds to about 15TB of NAND writes.

I would write a 1GB file to the SSD and record the Host Writes before and after. Also, I would examine the average erase count attribute, if it exists.

This is from the smartmontools database:

https://git.centos.org/rpms/smartmontools/blob/c8/f/SOURCES/drivedb.h

  { "SK hynix SATA SSDs",
    "SK ?hynix SC(210|300|308|311|313).*|" // tested with
      // SK hynix SC210 mSATA 256GB/20002L00,
      // SKhynix SC300 HFS256G32MND-3210A/20131P00,
      // SK hynix SC308 SATA 128GB/30001P10,
      // SK hynix SC311 SATA 512GB/70000P10,
      // SK hynix SC313 HFS256G32TNF-N3A0A/70000P10
    "HFS(128|256|512)G3[29]MND-(2200|3[23]10)A|" // HFS128G32MND-2200A/20200L00,
      // HFS512G32MND-3210A/20100P00, HFS512G39MND-3310A/20002P00
    "HFS(120|250|500)G32TND-N1A2A|" // SL308, tested with HFS500G32TND-N1A2A/30000P10
    "HFS(128|256|512)G39TND-N210A", // SC308, tested with HFS128G39TND-N210A/30001P10
    "", "",
  //"-v 1,raw48,Raw_Read_Error_Rate "
    "-v 5,raw48,Retired_Block_Count "
  //"-v 9,raw24(raw8),Power_On_Hours "
  //"-v 12,raw48,Power_Cycle_Count "
    "-v 100,raw48,Total_Erase_Count "
    "-v 168,raw48,Min_Erase_Count "
    "-v 169,raw48,Max_Erase_Count "
    "-v 171,raw48,Program_Fail_Count "
    "-v 172,raw48,Erase_Fail_Count "
    "-v 173,raw48,Wear_Leveling_Count "
    "-v 174,raw48,Unexpect_Power_Loss_Ct "
  //"-v 175,raw48,Program_Fail_Count_Chip "
    "-v 176,raw48,Unused_Rsvd_Blk_Cnt_Tot "
  //"-v 177,raw48,Wear_Leveling_Count "
  //"-v 178,raw48,Used_Rsvd_Blk_Cnt_Chip "
  //"-v 179,raw48,Used_Rsvd_Blk_Cnt_Tot "
    "-v 180,raw48,Erase_Fail_Count "
    "-v 181,raw48,Non4k_Aligned_Access "
    "-v 183,raw48,SATA_Downshift_Count "
  //"-v 184,raw48,End-to-End_Error "
  //"-v 187,raw48,Reported_Uncorrect "
  //"-v 188,raw48,Command_Timeout "
  //"-v 194,tempminmax,Temperature_Celsius "
  //"-v 195,raw48,Hardware_ECC_Recovered "
  //"-v 196,raw16(raw16),Reallocated_Event_Count "
  //"-v 198,raw48,Offline_Uncorrectable "
  //"-v 199,raw48,UDMA_CRC_Error_Count "
    "-v 201,raw48,Percent_Lifetime_Remain "
  //"-v 204,raw48,Soft_ECC_Correction "
    "-v 212,raw48,Phy_Error_Count "
    "-v 231,raw48,SSD_Life_Left "
    "-v 234,raw48,Unknown_SK_hynix_Attrib "
    "-v 241,raw48,Total_Writes_GiB "
    "-v 242,raw48,Total_Reads_GiB "
    "-v 243,raw48,Total_Media_Writes "
    "-v 250,raw48,Read_Retry_Count "
  },

Attribute 173 (= 0xAD) is the Wear_Leveling_Count or Average Erase Count. In your case it is 10 (= 0xA), so the total NAND Writes is 2560 GiB.

 

[tcsenter]

That number doesn't make sense. Show us all the attributes.

If attribute 0xA9 really is Max Erase Count, then the SSD has been completely erased a maximum of 58 (= 0x3A) times. That corresponds to about 15TB of NAND writes.

Crystal Disk Info is not showing NAND writes value in particular, only Host Writes. Here is the rest from Crystal Disk and also from SK Hynix Drive Manager SMART reporter. One of those Secure Erase cycle counts is attributable to myself:

 

[fzabkar]

I don't trust either tool. They both refer to attribute 0xF1 (= 241) as Total LBAs Written.

If that were true, then ...

0x289fc x 512 bytes = 85.2 MB​

​

https://www.google.com/search?q=0x289fc+x+512+bytes+in+MB​

​

I suspect that the actual unit is 16MiB per count. I'm betting that, if you write a 1 GB file to the SSD, the count will increment by 64 rather than 1.

 

[tcsenter]

Some weird things going on. I tried copying 1GB file noticed it was unexpectedly slow. Took ~40 seconds total. Disabled MS Security Defender realtime monitoring, in case it was scanning the file on copy or something. Still took ~40 seconds. I noticed it was quite slow to (quick) format too, about 30 seconds. Usually takes these things about six or eight seconds to quick format.

I changed over to another PC, I had been testing older hardware from 2009 era, Intel G41+ICH7 with non-AHCI SATA controller in ATA/IDE legacy mode, running Windows 7-64. Unable to manually invoke TRIM. Now running on 'newer' hardware, Intel 9 Series with Haswell/Devil's Canyon from 2014, in AHCI mode on Windows 10-64, with TRIM support. After running TRIM on the drive, cleaning partition, creating new partition, both quick format and copying 1GB file were now appropriately speedy. But...

Here is before and after copying 1GB file:



Aaaannd after:



CrystalMark results are in-line with the SATA 3Gbps port/interface it is on.

 

[fzabkar]

The difference is 66.

0x28C05 - 0x28BC3 = 66

https://www.google.com/search?q=0x28C05+-+0x28BC3+in+decimal

1 GiB / 66 = 15.5 MiB

Therefore, you should divide the reported Host Writes count by 64 to get the real TBW of your SSD. I'm suggesting 64 rather than 66 because it is more likely that a binary value is being used.

 

[tcsenter]

Thanks for catching that. I was trusting the SMART reporting and probably would have tossed them. So the two I have tested so far are fairly low usage less than 4 TB! Not a bad dealio.

 

[tcsenter]

Not related to the host/nand writes topic, but still related to the drives I acquired. I am testing these using this mSATA & M2 SATA NGFF (not simultaneous) adapter board:



I had been using this just fine with mSATA SSD, but there is problem with M.2 NGFF SATA SSD. BIOS doesn't see or detect it as a bootable or boot drive. But if you boot to Windows (another drive) with it attached as 2nd storage, it is exposed to Windows and works fine. Tried two different motherboards, more than one of the M2 SATA SSD; SC311 and SC401. Same in each.

I acquired one of these alternative designs, that unlike the one above has a tiny manual switch to select between mSATA and NGFF SATA:




Everything works fine recognized in BIOS, can select as boot device and works in Windows. Anyone encounter something similar or hazard a guess at what is happening? Gave me fits trying to figure out is it the drives? The mobo? The adapter?

 

[tcsenter]

Edit: Just noted the drive using problematic first (green) adapter disappears from Windows after resuming from standby. Tried using 'scan for hardware changes' in Dev Man but nothing. Had to restart and it appears again but only to Windows (and the Intel IRST driver).

I guess the solution is not use these ones for M2 SATA. mSATA SSD has no problems. I ordered two more of the black ones. I contacted seller of the green ones and asked if they know anything about it but haven't heard back. This seller has sold like 1000 of them so maybe.

 

[Torn Mind]

My experience is that heavily used SSDs can fail in ways not detected by SMART. These drives are only good for temporary storage of text documents.

 

[fzabkar]

Can you tell us the markings on that rectangular IC that has the SATA Tx/Rx differential inputs and outputs? This would be a 2:1 electronic switch or MUX of some sort. Perhaps there is a way to force the switch into the desired position.

Edit:

This appears to be the IC:

ASM1456, ASMedia, 2:1 MUX/DEMUX switch, 1.8V, TQFN-28L:
https://img.hqew.com/File/Others/110000-119999/113018/Electronic/201485015696529.pdf

Pin #3 selects between mSATA and m.2.

https://i.redd.it/9g8hu9k2s0211.jpg (PCB photo)

Actually, the pinout in the datasheet doesn't match the PCB. Something isn't right. Maybe the "B" version is different, ie AMS1456B (PCB) versus AMS1456 (datasheet)???

Edit #2:

Yes, it is different. The "B" version is a 3.3V part, whereas the other is 1.8V:

https://www.asmedia.com.tw/product/A28yq4brRDTnDqT3/bE5Yq49RG7gS8tGf
https://www.asmedia.com.tw/product/40cYQFaRGfHSdnh0/bE5Yq49RG7gS8tGf

 

[tcsenter]

Can you tell us the markings on that rectangular IC that has the SATA Tx/Rx differential inputs and outputs? This would be a 2:1 electronic switch or MUX of some sort. Perhaps there is a way to force the switch into the desired position.

Yep I have the 1456B version. There appears to be a different maybe newer/revised design of this board (from Ebay photos), that has a different component (FET?) and sheds one LED indicator:


I do not have this one so I don't know if it solves anything. Since the black one labeled "NFHK" with the little manual switch is working in all scenarios, I'll just stick with those ones. I actually ordered two more of them. Curiously (to me), I am unable to find any bridge/mux/dmux IC on the black one at all? So that Asmedia IC is just for auto-selecting/switch function?

Still haven't heard back from seller. These things were so inexpensive I'm not worried about any refund. It still works for mSATA, not a deal breaker. Thanks again for the reply!

 

[fzabkar]

AISI, there is essentially no difference between the black and green adapters. The black one switches the SATA Tx/Rx pairs manually, whereas the green one uses an electronic switch.

The ASM1456 switch has a single SELect pin. This would be a digital input, either 0 or 1, and this would then select the appropriate m.2 or mSATA connector.

The ASM1456B switch has a different pinout, but I can't find its datasheet, so I don't know how it is controlled. I'm hypothesising that the selection process goes awry. I'm wondering if it would be possible to force the switch into the desired state by manually manipulating its SEL pin(s).

As for the other green PCB, U1 is a 3.3V LDO linear regulator. This replaces the 3.3V switchmode regulator (better) in your green adaptor. It also has the ASM1456 version of the switch. You can just make out the R1 resistor on the SEL pin (pin #3). This pin is either a 0 or a 1. In your adaptor, R1 is connected to pin 23 (472 = 4.7K). If you can measure the voltage at this resistor, you will know how the mSATA and m.2 SSDs influence this pin.

This is the switchmode PSU on the black PCB:

https://i.ebayimg.com/images/g/BfQAAOSwArRhHLpa/s-l1600.jpg

This is the step-down regulator IC (SY8009AAAC):

https://www.olimex.com/Products/Breadboarding/BB-PWR-8009/resources/SY8009.pdf

I suspect that the SEL pin is connected to the m.2 slot's CONFIG_1 pin (pin #69). This pin is grounded when an m.2 SATA SSD is present, and open when an m.2 PCIe SSD is present. If the SEL pin is pulled up to Vcc, then the mSATA slot would be selected whenever the m.2 slot is empty or populated with an NVMe SSD.

M.2 format identification:
https://www.hddoracle.com/viewtopic.php?p=18130#p18130