SK hynix Gold S31

[otter]

I'd deployed 13 laptops with new SK hynix Gold S31 500GB HDDs but had never used HDS with them before. Today I having imaged a laptop with one of these new SSDs and trialing HDS with one of them for the first time.

HDS increments the number of bad sectors on disk by the number 40 each time HDS is launched. CrystalDiskInfo and HWiNFO don't report any bad sectors. If HDS isn't running I can power off and on the disks as many times as I like with no increment. Only when HDS is launched does HDS increment the bad sectors by 40 on each launch. Interesting is if I refresh repeatedly during perhaps the first minute of launch HDS increments the bad sectors by lesser numbers but stops at 40 each time.

I tested with another identical new SSD and observe the same behavior. Both have the latest firmware. Both have been tested with an internal Intel controller in AHCI mode and also a USB SATA controller. Same behavior with HDS 5.70.5 beta. Having connected both SSDs to my desktop PC with HDS license I was able to run full read tests on both SSDs without it reporting any bad sectors upon test completion.

HDS still keeps the incremented bad sector count despite deleting the DISKDATA data dat, hdd, hdi files.

My guess is HDS needs a slight tweak to adjust how the SMART info is interpreted from this SSD firmware?

[hdsentinel]

Thanks for your message and attention and I agree that it seems so interesting. But I can confirm that there is absolutely no bug/limitation/problem in Hard Disk Sentinel, exactly the opposite.
I can confirm that no, of course HDS does NOT increments the number of bad sectors at all.

The SSD itself increments the number of bad sectors: the SSD automatically reallocates the sectors, as this reflected in the S.M.A.R.T. attribute 198 Uncorrectable Sector Count in Hard Disk Sentinel. If required, this can happen any time, for example any disk activity (reads/writes) and it is a coincidence that you may see the increase when you launch a disk monitoring/checking software (of course, which reports the change).

As you can see, the other tool also reports that value (see C6 Uncorrectable Sector Count, showing 0200 hex), so wonder why you wrote that it "don't report any bad sectors.".
However, that tool seems not really compatible with this model, as it does not really report the proper name/value of most of the attributes. If I'm correct, Hard Disk Sentinel shows the attributes properly (note: I do not really understand why you did not send image about HDS S.M.A.R.T. page in this forum, why you post about a different tool

Generally the SSD performs the reallocations itself. This is completely automatic and usually unnoticed, as (compared to hard disks) SSDs do not make "clicking" sounds, there may be no noticeable delays or so. But I'm afraid SSDs can also have bad sectors, exactly as described at

https://www.hdsentinel.com/ssd_case_bad_sectors.php

I'm terrible sorry that other tools may be not compatible with this model and/or do not report problems with them correctly, so THEY may require changes/improvements.
But if Hard Disk Sentinel detects/reports them, it is not really a bug, but exactly the opposite: Hard Disk Sentinel designed exactly for this purpose, to reveal/show possible issues.

Personally examined/checked numerous SK Hynix Gold S31 SSDs and I just re-checked all the reports received about them. I see no similar issue with any them, the above mentioned counter were 0 in all of these SK Hynix Gold S31 SSDs.
Not sure, but it is possible that your SSD has a slightly different firmware and counts/reports issues differently than expected, so I'd be more than happy to examine this particular model and its status and (if required) adjust Hard Disk Sentinel but according what you sent, I'm afraid the other tools may require update (too)...

I suggest to please use Report menu -> Send test report to developer option. Then I can check the status of the SSD.

Maybe you can send multiple reports, for example once as a base and then after some time of use and/or if you may perform some tests, exactly to reveal all possible issues (as there can be more) long before they can cause failure or "just" data corruption. Good to hear that the surface test showed no issues - as it confirms that the SSD could perform the reallocation and probably no data yet stored.


> HDS still keeps the incremented bad sector count despite deleting the DISKDATA data dat, hdd, hdi files.

This is completely normal and expected situation.

The "bad sector count" is reported by the SSD itself. Hard Disk Sentinel just reads and displays the counter and stores the value in its statistics files to show statistics/changes.
But deleting the files do not delete the counter in the SSD itself of course.
The counter remains in the SSD as-is, as it reflects the total amount of bad sectors in the lifetime of the device.

Hard Disk Sentinel allows you to manually clear them by using the Offset column for the attribute, exactly to reset them and restore the health and be notified about possible new issues.
In general, this is described at:
https://www.hdsentinel.com/faq_repair_h ... _drive.php
and I can send step-by-step details for this of course.


> My guess is HDS needs a slight tweak to adjust how the SMART info is interpreted from this SSD firmware?

Interesting is that from the image I see Firmware version as 90000Q00
The other models I tested (and received reports) also have the same firmware version and they all reports 0000 in the attribute 198 Uncorrectable Sector Count (so no bad sectors).

So there is still chance that your firmware works somehow differently: may use this counter for slightly different purposes.
This is why checking the complete status would be essential to verify if there is any difference (in the operation of the firmware) compared to the previous tested models - and then (if required) make adjustments.

In such/similar situations, using the Report menu -> Send test report to developer option is always the best as then it is possible to check and confirm what happened.

But from what you sent, I'm afraid that
- Hard Disk Sentinel (only that) reports the errors correctly, does everything exactly as should
- other tools do not interpret/report the errors as they may be not compatible with this SSD model

[otter]

Always so much for me to learn. Thanks for such a thorough point by point response and even glancing at the SMART data I and others submitted online for this model. I always forget to translate hex numbers...Thanks for clueing me in.

I just now submitted separate test reports for both SSDs.

Another observation, likely entirely benign, but both of these SSDs increment their SMART value 195 On the Fly ECC Uncorrectable Error rate by 1 every ten seconds. I recorded videos of both SSDs and both incremented by 60 over the course of ten minutes. Looking at HDS chart for one of the SSDs shows SMART value 195 going from 27000 one day to 15000 the subsequent time I powered it on so I'm guessing that SMART value zeros out each time the unit is powered off or something like that.

Days past I only ran full read tests on the SSDs. I have no problem with running a full write test on one or more of these SSDs to see how many bad sectors get recorded and if the uncorrectable sector count ceases increasing thereafter. I also realized I have a third one of these SSDs sitting unopened in a box as well. I'll test it any way advise to best learn the peculiarities about these SSDs.

I greatly appreciate your brilliant expertise about this stuff.

[hdsentinel]

Thanks for the reports, yes, I see them and saw the increased bad sector count. I also noticed the increase of the attribute 195 too.
Interesting, as other SK Hynix Gold S31 models tested did not show this/similar for attribute 195 until now, they all show constant there too.

From the reports, I see these connected with USB adapter/enclosure/dock.
Did you use the SSDs only this way? Not sure, but maybe it's possible that it is related, as (while there should be no problems) some USB adapters/docks do not "like" all hard disks/SSDs and there can be some compatibility issues.

Personally I'd surely check the status for slightly longer time and verify if the health will change/degrade or not further. If not, I'd restore the health by clearing the current error counters, to be notified about possible new issues (as described at https://www.hdsentinel.com/faq_repair_h ... _drive.php and I can assist personally by e-mail).

Personally I'd surely try the third SSD in a completely different environment: by connecting directly to a SATA port of a completely different computer (without USB adapter just to be safe) and examine things there too. This may reveal if the current operating environment is somehow affected and could cause troubles.
Maybe you can ask also the manufacturer too about the situation (especially if you have results of the 3rd SSD too) as maybe they can explain and/or give some thoughts for this case.

[otter]

I first observed the behavior with a laptop with an intel Integrated controller in AHCI mode. The USB adapter only came into play when I wanted to run a full read test on it and that feature was only available on my desktop PC. I then also opened a second new Hynix SSD and tested it with the USB adapter as well.

I can open the third new Hynix SSD and attach it to my LSI SAS HBA since HDS seems to work fine with the SAS HBA. Unless you suggest differently, I'll go ahead and do that a couple days from now.

[hdsentinel]

I'd worry about using with the LSI SAS HBA.
It seems these SK Hynix Gold S31 SSDs are somehow very "sensitive" and according the experiences, in general some SSDs have compatibility issues with LSI SAS HBAs (but it can be related to other factors, eg. SAS backplanes/enclosures/hot swap bays if used).

Probably the most safe would be yes, to try with onboard Intel AHCI controller with latest possible Intel chipset drivers to be used. Let's see how it will work and what it will report (compared to the others). Maybe if possible, I suggest to please use Report menu -> Send test report to developer option immediately when first detected (so let's have a "base" report) and then a new report after some use/testing, to check the differences.

[otter]

I didn't see your latest reply until just now, so I didn't submit developer reports from the fresh SSD(s). Sorry about that. Yet, I'm very happy how things have turned out for me.

I've learned so much. Thank you! The bad sectors incrementing by 40 on each connect/disconnect was super weird, yet that did finally stop after running full write tests. @hdsentinel please switch this thread to the Questions sub-forum given my ignorance.

Having run full random write/read tests on the suspect SSDs (I'll refer to them as SSDs A&B) the end result was 203 out of 920 spare blocks getting used on SSD-A and 88 out of 824 spare blocks getting used on SSD-B.

I wound up having not one, but two more unopened units of this SSD on hand. I'll refer to them as SSDs C&D. I tested C&D with full random write and read tests as well. Testing C&D reveals no bad sectors on either.

What stands out about suspect SSDs A&B is SMART ID 195 On-the-fly ECC uncorrectable error rate is 13,000,000 on SSD-A and 50,000,000 on SSD-B. Those figures are ever increasing even when doing read only tests and having already completed full write tests. SMART ID 195 On-the-fly ECC uncorrectable error rate on SSDs C&D remains ZERO.

For others with this series SSD, I learned Amazon reviews had many disgruntled people exclaiming that Hynix is dodging honoring the advertised 5 year warranty for these SSDs and instructing persons to contact Amazon for refund with 20% restocking fee if within one year of purchase. Hard to believe, yet that is exactly how it transpired for me as well. Fortunately I opened these two suspect SSDs after ten months from order delivery, so Amazon did accept return of the two. Could the suspect SSDs have been fine going forward? Perhaps @hdsentinel has further insight about that. Yet, since Amazon made it so easy for me to return the suspect ones I just went along with that.

I do have one final question though. I was curious what Hynix' SMART threshold values actually are for these SSDs and noticed HDS reported all SMART threshold values as zero, not only for these SSDs, but actually all my disks. Yet, other SMART utility apps like CrystalDiskInfo do actually indicate threshold values for many fields. Here is a side by side example from a spinning disk of mine.



I'm totally fine to quick check in an alternate app if that's what's necessary to satisfy that curiosity, but thought you might have some insight as to why at least on this PC HDS isn't reporting the threshold values?

Thanks again for your awesome expertise!

[hdsentinel]

The Thresholds may be normal if missing. It is not a bug/limitation of course, exactly the opposite.

With some controllers (for example older LSI HBAs, you mentioned that, so I suspect it is the case now) sometimes it is simply not "safe" to attempt the detection of the Thresholds: according the experiences, very rarely (depending on the controller firmware, driver too) the detection of the Thresholds can cause BSOD/crash, resulting data corruption/data loss.
Other tools happily ignore this and always attempt to detect the thresholds, expecting the best - but when such/similar "suspicious" situation found, Hard Disk Sentinel (to be 100% safe) may not attempt to detect them, as you can see.

As you can see, it is not related to the SSD itself but related to the controller / HBA (so other disks on the same controller/HBA affected too).

I suggest to please use Report menu -> Send test report to developer option now, as then I can check the situation, examine the current controller/driver version and also can help to enable the detection of the Thresholds too - and in case things will work safely, then change the behaviour to automatically enable with your particular controller/driver combination.

A final question if possible: after advertising an other app here many times, please now do the same with Hard Disk Sentinel too: please upload the images from Hard Disk Sentinel too on the page/forum of the other app too similarly to show the differences there too