News Samsung 990 Pro Firmware Update Addresses Failing SSD Health

[InvalidError]

A decade old paper.

Unless your SSD has enough on-board backup power to ensure critical writes are completed cleanly on input power loss, corruption on random power loss is still a possibility today. Tempt the devil enough times and you'll eventually lose data despite improvements in journaling and write management strategies. HDDs aren't immune to this either, damage is just mitigated by their relatively low speed.

 

[Lafong]

Pro version is for demanding consumers. Buying one I'd expect reliability and performance.

That's what the marketing department would like you to believe.

Pity that engineers don't have the same representation at the meeting when model names such as "Pro" are decided upon. But I suppose engineers may be equally beholden to upper management.

 

[setx]

Unless your SSD has enough on-board backup power to ensure critical writes are completed cleanly on input power loss, corruption on random power loss is still a possibility today. Tempt the devil enough times and you'll eventually lose data despite improvements in journaling and write management strategies.

I believe that's exactly what happened with my Plextor SSD.

I've read that Samsung has some kind of journaling to prevent corruption in such cases. Is there any such information for other consumer drives?

 

[InvalidError]

I believe that's exactly what happened with my Plextor SSD.

I've read that Samsung has some kind of journaling to prevent corruption in such cases. Is there any such information for other consumer drives?

I doubt there are any public details beyond generic reliability improvement claims.

 

[bit_user]

What you say sounds very plausible. On the other hand: how can an SSD wear by 50-80% within just a couple of days or weeks??

I'm not saying this is definitely the case, but imagine the firmware had a bug that caused it to continually rewrite or shuffle data around. That would cause wear by simply leaving the drive powered on. Given the extremely low amount of host writes on some of these drives where people have experienced excessive wear, it seems plausible to me that's what happened.

 

[bigpinkdragon286]

Awful lot of rumor and speculation going on here about just what can be done by Samsung, especially knowing only the symptoms and not precisely what the underlying issue is.

Can the wear level even be reset?

Can the page health be determined accurately by the hardware or do some cheaper driver just count write / erase cycles, which in this case may be difficult or impossible to track back at this point?

Should Samsung really put wear level resetting code into a firmware update for shady resellers or scammers to reverse engineer for the used market?

Why would you reset the wear counter to zero, even if you could, when these drives have been in use? One would think it best to reset the wear level to an accurate level, assuming that's even possible, but certainly not back to zero, which in most cases is likely incorrect even for a relatively new drive.

Does anyone outside of Samsung even know if the wear indication is correct?

Finally, Samsung may be choosing to address the issue in multiple steps without being forthcoming. The initial firmware update may only target a portion of the whole issue to mask symptoms. Redressing customer concerns about lost value may come in the future, after the company has determined the best means to bridge this with accounting and shareholders.

 

[bit_user]

I myself buy GOODRAM SSD's or WesterDigital. The first is a local brand and I wholeheartedly support local business. Second is not the top performer, but it's predictable.

Okay, so they're Polish, in case anyone else was wondering.

As for Western Digital, I recall they got their SSD business by acquiring someone. Can anyone remember who? Maybe SanDisk?

I think Toshiba (now Kioxia) acquired its SSD business by buying OCZ.

I think neither of these brands (SanDisk and OCZ) had stellar reliability before their respective acquisition, but enough time has passed that perhaps that's changed.

 

[bit_user]

I've read that Samsung has some kind of journaling to prevent corruption in such cases. Is there any such information for other consumer drives?

About 10 years ago, Intel provided a wealth of information in their datasheets. Unfortunately, that practice long ago stopped and I haven't seen any other manufacturers provide comparable information.

 

[Amdlova]

Okay, so they're Polish, in case anyone else was wondering.

As for Western Digital, I recall they got their SSD business by acquiring someone. Can anyone remember who? Maybe SanDisk?

I think Toshiba (now Kioxia) acquired its SSD business by buying OCZ.

I think neither of these brands (SanDisk and OCZ) had stellar reliability before their respective acquisition, but enough time has passed that perhaps that's changed.

Western digital > SanDisk
Toshiba > ocz
Seagate > lsi "sandforce"



Ocz has some firmware problems, when got in transaction with the controller Sandforce to the barefoot own design "sandforce has some awful design and problems too" but in these times it's the "better product".

 

[LJHCPA]

I have two 990 Pro's in RAID 0 and two 980 Pro's also in RAID 0. Samsung Magician won't report any information on these drives because they're in RAID, but Crystaldisk Info shows the firmware versions, serial numbers, read and writes in GB and disk health. All four of my drives show 100% health. Did I win the silicon lottery? 990 firmware 0B2QJXD7 and 980 firmware 5B2QGXA7. are these firmware ID's free of the bug?

I bought the 990's from Newegg on their pre-order so I assume the drives I got were among the earliest they made. I bought one of the 980's new from Best Buy and one from them that was refurbished. The new 980 shows 574 GB in reads and 547 GB in writes. The refurbished drive shows 770 GB in reads and 2,410 GB in writes, yet both show 100% health, The 990's report reads of 1,174 GB and writes of 1,605GB. I haven't had any problems with any of the drives and they're pretty darn fast. 990 read score of 14,226 and write score of 13,021 (CrystalMark) the 980's report read score of 13,850 and write score of 10,760. Should I book a trip to Vegas, or buckle up for a bumpy ride down the road?

System specs: MB ASUS Maximus Z790 Hero, i9 13900KF, 64GB GSkill Z5 6000 RAM, MSI Gaming Trio 4090. (990's are the boot drive in the ASUS Hyper M.2 card in PCIe slot 4 and the 980's in MB slots M.2_2 and M.2_3

 

[kyzarvs]

Western digital > SanDisk
Toshiba > ocz
Seagate > lsi "sandforce"



Ocz has some firmware problems, when got in transaction with the controller Sandforce to the barefoot own design "sandforce has some awful design and problems too" but in these times it's the "better product".

SanDisk early SSDs were top of the charts - Toms recommeded them all the time. My fist SSDs were all SanDisk - I think my home 24/7 linux server is still running one as a boot drive (no essential data on it these days, all on the NAS) - that's been going for... 6-7 years at a guess. It's powered by an FX8350 to show its vintage!

 

[Gillerer]

If it were really the case that the wear was fake, it seems like the firmware update probably would've reset the counters.

- -
Either the wear is real or it could just be that the scratchpad used to track writes cannot be rolled back.

I don't think it would be a good idea to implement a wear level counter in any way that it could be reset or manipulated.

Imagine the profiteering possibilities for refurbished drives - even by (on the surface) legitimate retailers in case of returned products that would normally have to be sold as "open box".

 

[bit_user]

I don't think it would be a good idea to implement a wear level counter in any way that it could be reset or manipulated.

Imagine the profiteering possibilities for refurbished drives - even by (on the surface) legitimate retailers in case of returned products that would normally have to be sold as "open box".

I expect that it's ultimately just data that resides on the very flash it's managing. There are ways to protect it from tampering that don't involve making it unresettable, such as by limiting the controller to executing only signed firmware images.

 

[KyaraM]

Okay, so they're Polish, in case anyone else was wondering.

As for Western Digital, I recall they got their SSD business by acquiring someone. Can anyone remember who? Maybe SanDisk?

I think Toshiba (now Kioxia) acquired its SSD business by buying OCZ.

I think neither of these brands (SanDisk and OCZ) had stellar reliability before their respective acquisition, but enough time has passed that perhaps that's changed.

WD is SanDisk, yes. For a regular end user, though, their reliability is not an issue. Any manufacturer should be fine, maybe with an exception or two. My system had 0 fails in a year so far, and while 1 year is no age for an SSD, I don't expect anything to happen to my WD Black SN850 any time soon, nor to my Kingston game drive for that matter, which is a simple NV1 drive.

 

[Deleted member 431422]

That's what the marketing department would like you to believe.

Pity that engineers don't have the same representation at the meeting when model names such as "Pro" are decided upon. But I suppose engineers may be equally beholden to upper management.

... That's some hard world reality there. It's all true tough. You tend to forget the suffix "Pro" has little meaning these days.

 

[Deleted member 431422]

Okay, so they're Polish, in case anyone else was wondering.

As for Western Digital, I recall they got their SSD business by acquiring someone. Can anyone remember who? Maybe SanDisk?

I think Toshiba (now Kioxia) acquired its SSD business by buying OCZ.

I think neither of these brands (SanDisk and OCZ) had stellar reliability before their respective acquisition, but enough time has passed that perhaps that's changed.

You're right. I remember SanDisk being a reliable brand. OCZ wasn't as popular here. I followed WD after SanDisk acquisition. They're not the top performers, but I didn't see too many problems with their SSD's and they're reasonably priced.
WD acquired HGST as well. Not an SSD, but those were the most reliable HDD's.

 

[Mon12]

I had a Samsung PM981 1TB OEM SSD die on me after 3 years and I never had the power cut as I always had a good battery in my laptop; the documentation for the SSD model claims the Mean Time Between Failure is as high as 1,500,000 hours! And that the Annual Failure Ratio is 0.4%... well, I don't believe that's true in practice 🙁 I've not seen a firmware update for this one, only for PM981a.

 

[USAFRet]

the documentation for the SSD model claims the Mean Time Between Failure is as high as 1,500,000 hours! And that the Annual Failure Ratio is 0.4%

And there are always outliers, on both ends of the fail scale.

 

[bit_user]

And there are always outliers, on both ends of the fail scale.

Yeah, but also the reliability stats are merely estimates and there's no penalty (other than warranty claims) if they're wrong.

I'm also a bit jaded about reliability stats. Since all such product specifications come with a disclaimer that they're "subject to change without notice", I put very little faith in them.

 

[USAFRet]

Yeah, but also the reliability stats are merely estimates and there's no penalty (other than warranty claims) if they're wrong.

I'm also a bit jaded about reliability stats. Since all such product specifications come with a disclaimer that they're "subject to change without notice", I put very little faith in them.

Right.
But the fact that I had a 16TB Toshiba Enterprise drive die at 7 months does not mean they all would.
Mine just happened to be on the left side of the bathtub graph.

Another one of those in a different PC might last 20 years.

 

[bit_user]

Right.
But the fact that I had a 16TB Toshiba Enterprise drive die at 7 months does not mean they all would.
Mine just happened to be on the left side of the bathtub graph.

Another one of those in a different PC might last 20 years.

We're not debating the existence of the bathtub curve, here (or, at least I'm not) - just whether to trust the manufacturer's claims about its shape & size.

 

[dimar]

I stay away from everything Samsung ever since microSD slot was removed from recent phones. WD and Seagate SSDs are great these days, no reason to buy a Samsung SSD at all.

 

[bit_user]

Samsung phones with batteries that become bloated and dangerous.

BTW, I remembered a key detail about this.

It's not that other phones don't have problems with defective batteries swelling. Indeed, I have a LG phone and its original battery did that after about 2 years! The problem Samsung stumbled into is they designed a shell which wouldn't expand when that happened. Somehow, by keeping the battery compressed, the result was a violent explosion.

 

[InvalidError]

The problem Samsung stumbled into is they designed a case which wouldn't expand when that happened. Somehow, by keeping the battery compressed, the result was a violent explosion.

If old Samsung phones are anything like my first Samsung Tab A 7, Samsung used to overcharge the crap out of their LiPo batteries at over 4.3V. If you want LiPo to last, you shouldn't push them beyond 4.2V.

 

[xyzzyxyzzy]

Samsung also provide an ISO that you should be able to boot from in order to update the firmware. That said, I've thus far been able to create a bootable USB that actually works (Rufus, ISO mode and DD mode). Even if I get it to boot I doubt it'll be of use to me - I have three 990s in a RAID array, and Samsung Magician can't see the drives (though HWInfo64 can, go figure).

Addendum: the trick to get the ISO to boot is to disable secure boot in the BIOS. It still can't see the drives underlying my RAID array though, so I'm going to have to update my backup of all 3TB then break the array before trying again.