Samsung 870 Evo -- is it a dud model?

Metanis said:
As an "influencer" it just reinforces how important backup and disaster recovery needs to be.
Click to expand...

As far as I'm concerned, it's even more critical in the age of the SSD than it was when the HDD alone was king. When these things fail, data recovery is not only far more expensive, it's far less likely to be successful in many cases.

With HDDs, you had a good chance of getting quite a bit back if the platters were not physically damaged. That's not true with SSDs, and depending on the failure mode it may not be possible to get anything.
 
SSDs are great for speed and portability (don't crash from being moved), but are super unstable thanks to low quality of NAND technology.

Basically, use SSDs for OS and quick access data, with a very solid backup routine.

Hard drive recovery success rates = usually around 90%
SSD recovery success rates = < 20%
 
lcoughey said:
Hard drive recovery success rates = usually around 90%
SSD recovery success rates = < 20%
Click to expand...

Thanks for this. I keep sharing this sort of thing as far and widely as I can.

There are many who seem to believe that SSDs are "indestructible." And while they are when it comes to shock damage (at least to an exponentially greater extent than a HDD is) they have sudden, catastrophic failure without warning. I've had that happen with several Mushkin drives and one Adata drive. I know of plenty of other instances with other brands, too.

As @nlinecomputers has said, when they go, it's like a light bulb burning out. And once that has occurred, the probability of any recovery is quite low indeed.
 
fincoder said:
What about adding failure rates to the mix though?
Click to expand...

Which would be interesting, but that wasn't the question posed.

Personally, in my short history with SSDs they've been far more failure prone than HDDs had been for several decades. But I don't have a statistically valid sample for either, either.

But I will never trust SSDs as they exist now as much as I trust HDDs as they exist now for long term backup storage.

But, in the end, what I said elsewhere today applies: It's way more important to have a backup on the media of your choosing than worrying about which specific media type that may be. The probability of the original media and the backup media failing at precisely the same time, barring physical destruction of both because they're in the same physical location, is so small as to be zero.
 
fincoder said:
What about adding failure rates to the mix though? Chances of data loss overall may be less with SSDs, or about the same as HDDs.
Click to expand...
The following shows that HDDs are more reliable, overall (assuming 3.5% failure rate/yr for HDDs and 1.5% for SSDs):

1000 HDDs - 3.5% failure rate (i.e., 35 drives) leaves 965 drives that didn't fail. 90% recovery on 35 HDDs yields 31.5 recovered, so 965+31.5 = 996.5 HDDs with data.

1000 SSDs - 1.5% failure rate (i.e., 15 SSDs) leaves 985 SSDs that didn't fail. 20% recovery on 15 SSDs yields 3 recovered SSD, so 985+3 = 988 SSDs with data, i.e., 8.5 fewer than had HDDs been used instead of SSDs.

The best case is an SSD for the OS drive with user data folders relocated to a HDD, and regular back-ups of both.
 
Good post Larry, explaining the maths. My take-away is that data retention is somewhat higher for HDDs but not by a great deal.

The underlying failure rate figures are the key though, and they might be somewhat rubbery. The recovery rate figures are likely to be quite accurate with the only caveat being that without expensive specialist recovery services (often not affordable to home and small office users) that rate would be lower (60% ?).

Considering the SSDs are getting better all the time, and if sticking to certain major brands, and the different usage patterns in client PCs compared to Backblaze's server usage, the failure rates might actually be lower. My own experience suggests that the SSD failure rates are lower than 1.5% and entry-level HDDs made in the last 10 years have a higher failure rate than 3.5%.

Assuming Backblaze's failure rates though, with the high costs of specialist recovery I think it tips in favour of SSDs (for home and small office users) in regards to data retention.

Regardless, I agree with "regular backups". This has always been and always will be the advice for any storage medium, another reason that the relatively minor differences in data retention between HDDs and SSDs is not significant.
 
Yes, the cost to recover the low percentage of SSDs will likely rise to offset the low odds and costs to be able to provide the service. For example, I currently charge the same minor ($500 CAD ) and major ($1250) rates for hard drives and SSDs, but am seriously considering increasing the cost of NVMe SSD recoveries to the $2000-$3000 range.

I have a current case that is occupying the only NVMe port on a my $15,000USD system which will take a month or more to recover, assuming that I can eventually get the 15-20% of bad sector reads down to something more reasonable.
 
So has a consensus been reached on when you replace a perfectly good SSD? 50,000 hours? 75,000? ..or at what percent should it be replaced when reading something like CrystalDiskInfo? ...and how much depends on the brand for longevity Inland vs. Samsung? I've got some pretty old SSDs out there (50,000+ hours). I'm not sure what to recommend to my old customers.
 
Given the different hardware parameters (the NAND itself for example) and different firmware's (and quality of those) I think it's hard to come up with a generic advice. Advice/interpretation by something like CrystalDiskInfo is simplistic and is basically simply following thresholds as determined by the manufacturer (write/erase cycle count). IOW if you take the manufacturer's word for it, CrystalDiskInfo is accurate.
 
Last edited:
I don't worry about usage that much. SSDs can die at any time so as long as the user has their data backed up or living in the cloud I don't worry about it much. Just remind the user that SSDs die like light bulbs. One day they work and the next day they don't and plan accordingly.
 
Diggs said:
So has a consensus been reached on when you replace a perfectly good SSD? 50,000 hours? 75,000? ..or at what percent should it be replaced when reading something like CrystalDiskInfo? ...and how much depends on the brand for longevity Inland vs. Samsung? I've got some pretty old SSDs out there (50,000+ hours). I'm not sure what to recommend to my old customers.
Click to expand...
We have probably 300-400 Inland SSDs installed for our customers and until the last two months not had any failures with them. The last 8-10 we tried to install did not work. Hopefully it was just a bad run but we are going to send the remaining 20+ we have in stock back and exchange them. We carry 128, 240 and 512s,
 
nlinecomputers said:
I don't worry about usage that much. SSDs can die at any time so as long as the user has their data backed up or living in the cloud I don't worry about it much. Just remind the user that SSDs die like light bulbs. One day they work and the next day they don't and plan accordingly.
Click to expand...
Absolutely. That vast majority of users who still think that backup is an unnecessary expense are just going to have to learn the hard way.
 
seashore said:
That vast majority of users who still think that backup is an unnecessary expense
Click to expand...

I don't even think it's thought of as "an unnecessary expense," it's either not thought of at all or falls into the, "Oh, that won't happen to me!," category.

Long before I ever got into this business I went for years and years with no backup drive (mostly because they were obscenely expensive for the home user at that time - now they're dirt cheap) and given that I had no catastrophic failure lulled myself into the latter of the previously mentioned categories.

My first time of losing everything taught me the folly of my way and way of thinking. As a tech, I am actually eternally grateful for having had that experience as I can speak directly to it, and can emphasize that with SSDs, unlike with HDDs, the failure generally occurs with zero warning and a very low probability of recovery at great expense trying. Spending $75 to have a mammoth personal backup drive (or two smaller ones, if you want to alternate between them on a per backup basis), is the best money you'll ever spend for peace of mind.
 
You must log in or register to reply here.
Back
Top