Overclock.net banner

1 - 11 of 11 Posts

·
Registered
Joined
·
154 Posts
Discussion Starter #1 (Edited)
Hi everyone. It's been a few years since I've posted. Wow I can't believe I joined way back in 2006! I wanted to give you a little info on SSD endurance.

A lot of people seem to be concerned about MLC SSDs because of the problem of lower endurance of these drives. I've seen this myself even with the newer 3D SSDs sometimes after just 6 to 8 months of use. This issue is greatly influenced by the increased electrical wear that will occur within the MLC design. While certain 3D MLC drives can mitigate this problem there are a few other SSD options that will give much increased endurance per GB while using hybrid type MLC that comes close to the endurance of the SLC design. This can be achieved by programming or other proprietary means and some are available with greater operating temperatures. They cost more but nowhere near what SLC drives cost. Anyways, here are some SSD options that are available that exceed typical endurance specs of a common consumer SSD. These are usually industrial and I've done a few tests of similar military speced drives from other companies but the results are about the same minus some of the physical endurance and material quality characteristics that the mil spec drives have. Not really much of an advantage that will matter for most people. If you want more examples just do a search for iSLC and pSLC.

Transcend SSD510K (128GB), SuperMLC
Endurance, 1420 TBW

APACER SU210-25 (128GB), SLC-Lite
Endurance, 1315 TBW

Innodisk 3IE4 (128GB), iSLC
Endurance, 1388 TBW

ATP Velocity XE (512GB) aMLC
Endurance, 17066 TBW

Amtron UA Series (128GB) pSLC
Endurance, 2568 TBW
 

·
Senioritis Member
Joined
·
8,360 Posts
TBW is not the only factor to consider when rating the reliability of an SSD. There are other ways for SSDs to die, such as the controller chips dying. Also, many SSDs continue to last far longer than the manufacturer's guaranteed write life.

If you are looking for high endurance, I suggest looking into the new Samsung 860 Pros. Samsung 840 and 850 Pros have an excellent track record and there is no reason to think the 860 Pros won't be just as reliable as the 840s and 850s, if not more so. The warranted write life of the 860 pro 256GB (the smallest one made) is 300 TBW over five years, far, far more that most consumers will ever exhaust over five years. That 300 TBW rating is just for warranty purposes; most likely, it will last much longer. Samsung rates the 256GB 860 Pro reliability at 1.5 million hours MTBF (Mean Time Between Failures). That's around 170 years of continuous operation, outrageously more than anyone will ever use one.

However, any SSD (or HDD) is subject to irrecoverable failure at anytime without warning, no matter how good they may be. The only way to ensure the safety of one's data is to maintain at least one onsite and one offsite backup. If continuous operation is important, then a RAID or some other kind of redundancy should also be employed (btw, RAID or similar, in itself, is NOT a backup!).
 

·
Registered
Joined
·
154 Posts
Discussion Starter #3
TBW is not the only factor to consider when rating the reliability of an SSD. There are other ways for SSDs to die, such as the controller chips dying. Also, many SSDs continue to last far longer than the manufacturer's guaranteed write life.

If you are looking for high endurance, I suggest looking into the new Samsung 860 Pros. Samsung 840 and 850 Pros have an excellent track record and there is no reason to think the 860 Pros won't be just as reliable as the 840s and 850s, if not more so. The warranted write life of the 860 pro 256GB (the smallest one made) is 300 TBW over five years, far, far more that most consumers will ever exhaust over five years. That 300 TBW rating is just for warranty purposes; most likely, it will last much longer. Samsung rates the 256GB 860 Pro reliability at 1.5 million hours MTBF (Mean Time Between Failures). That's around 170 years of continuous operation, outrageously more than anyone will ever use one.

However, any SSD (or HDD) is subject to irrecoverable failure at anytime without warning, no matter how good they may be. The only way to ensure the safety of one's data is to maintain at least one onsite and one offsite backup. If continuous operation is important, then a RAID or some other kind of redundancy should also be employed (btw, RAID or similar, in itself, is NOT a backup!).
It's true that TBW or even something like DWPD aren't the only factors when it comes to endurance but I think they're something people should look for anyways. I didn't bring up the quality of controllers because I think that some of the problems I've seen are probably because of the increased incidence of electrical wear and degradation by chronic temperature changes that's going to happen with an MLC design. I suspect that the controllers are probably doing their job right, especially with more recent SSDs. This hasn't happened with SLC drives from my own experience.
 

·
Senioritis Member
Joined
·
8,360 Posts
It's true that TBW or even something like DWPD aren't the only factors when it comes to endurance but I think they're something people should look for anyways. I didn't bring up the quality of controllers because I think that some of the problems I've seen are probably because of the increased incidence of electrical wear and degradation by chronic temperature changes that's going to happen with an MLC design. I suspect that the controllers are probably doing their job right, especially with more recent SSDs. This hasn't happened with SLC drives from my own experience.
You are still placing too much emphasis on drive reliability. Again, it does NOT matter how good a drive is; it is still susceptible to sudden, irrecoverable failure at any time, even when brand new. Granted, the better the drive quality, the lower the frequency of failures in a batch of them but that still doesn't ensure against failure. Out of the 34 SSDs I've owned (and I still have 32 of them), only two have failed me. One was a 128GB Samsung 840 Pro that died after almost 5 years of 24/7 operation (it was still under warranty--barely--but I didn't bother to file a warranty claim because I no longer have any use for a drive under 500GB) and the other was a 4TB Samsung 850 EVO that was DOA (I had it quickly exchanged for a new one under the vendor's 30 day warranty).

In the case of the 128GB SSD that died, I would have had to reinstall the OS and around a hundred programs and manually restore all my settings, a job that would have taken hours, if not a day or two. Instead, I was able to remove the dead drive, install a replacement drive (a retired 500GB SSD I had), then restore my latest backup image of the old drive to the replacement drive. The entire process took only a little over an hour (most of that time being getting at the dead drive due to its poorly planned location in the case; my new case will not have that problem). Having a solid backup scheme in place is the only reliable way to protect your data (keep in mind that, while a RAID can be part of a backup scheme, in itself, it is not a backup).

Again, if protection against downtime due to drive failure is essential, or desirable enough to warrant the expense, redundancy, such as a RAID, is called for.
 

·
Premium Member
Joined
·
3,806 Posts
You are still placing too much emphasis on drive reliability. Again, it does NOT matter how good a drive is; it is still susceptible to sudden, irrecoverable failure at any time, even when brand new.

Having a solid backup scheme in place is the only reliable way to protect your data (keep in mind that, while a RAID can be part of a backup scheme, in itself, it is not a backup).

Again, if protection against downtime due to drive failure is essential, or desirable enough to warrant the expense, redundancy, such as a RAID, is called for.
Yep. For the price of a pseudo-SLC or high endurance MLC SSD, might as well grab multiple SSDs (and/or HDDs) and maybe a cloud subscription for backup and redundancy.

I can understand needing high endurance for database and OLTP-type stuff. However, for common user workloads I reckon capacity would long be an issue before drive health, which is typically tied to remaining P/E cycles, drops down to 90% and that's even with TLC/3-bit MLC NAND. I know I have a collection of retired 128GB and 256GB class SSDs that I just use as oversized flash drives. Even there sometimes capacity is still too low. Drive health? Iirc, lowest is 95%.
 

·
Registered
Joined
·
154 Posts
Discussion Starter #6
You are still placing too much emphasis on drive reliability. Again, it does NOT matter how good a drive is; it is still susceptible to sudden, irrecoverable failure at any time, even when brand new.
Of course it matters. The companies that offer it are aware of problems with MLC even with moderate use and know that for some it's not economically realistic to use SLC, at least not for everything, especially when factoring costs of other components. It's simple. More TBW the SSD can take means less wear and chance of failures and also increased data retention. I didn't start the thread to debate every factor that can play a role in overall SSD quality and I mentioned in my first post that 3D MLC can mitigate the problem. But tech like iSLC and pSLC make SSDs better quality for the reasons that I mentioned just like things like better PCB design, VRM quality and materials make motherboards better.

"Out of the 34 SSDs I've owned (and I still have 32 of them), only two have failed me."

You are extremely lucky. None of my first drives work after a year of power off and the last one eventually started getting slower followed by occasional shut downs. All MLC drives.
 

·
Senioritis Member
Joined
·
8,360 Posts
Of course it matters. The companies that offer it are aware of problems with MLC even with moderate use and know that for some it's not economically realistic to use SLC, at least not for everything, especially when factoring costs of other components. It's simple. More TBW the SSD can take means less wear and chance of failures and also increased data retention. I didn't start the thread to debate every factor that can play a role in overall SSD quality and I mentioned in my first post that 3D MLC can mitigate the problem. But tech like iSLC and pSLC make SSDs better quality for the reasons that I mentioned just like things like better PCB design, VRM quality and materials make motherboards better.

"Out of the 34 SSDs I've owned (and I still have 32 of them), only two have failed me."

You are extremely lucky. None of my first drives work after a year of power off and the last one eventually started getting slower followed by occasional shut downs. All MLC drives.
You are still missing the point. I never said that quality, reliability, and/or endurance didn't matter. What I'm saying is you are placing far too much importance and reliance on it. For the third time, no matter how good a drive is, it is still subject to sudden irrecoverable failure at any time without warning. No drive has ever been made (and probably never will be made) that is 100% reliable. Given enough time, every drive will fail; it's just a matter of when. Obviously, the better the drive, the fewer problems you are going to have but you will never eliminate them. That's why it's essential to have backups and, if continuous operation is needed or desired, redundancy.

Today's MLC, and even TLC, drives are far more durable and reliable than the early SLC drives. They are far more cost effective than SLC drives.

Btw, I've owned only one MLC drive so far (ironically, it was one that failed in use; the other "failed" drive was a DOA TLC). All my other drives are TLC, most of them 4TB. My success with them has nothing do with luck. It comes from being willing to pony up for good drives (Samsungs), then treating them properly, such as formattting them correctly, not writing unnecessarily to them, and not overfilling them.
 

·
Registered
Joined
·
154 Posts
Discussion Starter #9
Curious, what drives were those? There were plenty of controller missteps with early MLC SSDs.
Mushkin Callisto, Mushkin Chronos (2), Corsair Force LS, Crucial MX100.
 

·
Senioritis Member
Joined
·
8,360 Posts
Mushkin Callisto, Mushkin Chronos (2), Corsair Force LS, Crucial MX100.
Curious. the Crucial MX100 was a pretty fair SSD even though it used a Marvell controller. The others? Not so much (to put it politely). The Mushkins used the problematic Sandforce controllers (even the Marvells were better than the Sandforces). Corsair used a Phison controller; not exactly a great one but better than Sandforce (pretty much anything was better than Sandforce, especially the early ones.

You buy low end products, you are going to get low end results. The Samsungs aren't cheap but, except for the early 840 EVOs, they have all been solid (even the problematic 840 EVOs were reliable; they were just really slow for an SSD).
 

·
Registered
Joined
·
154 Posts
Discussion Starter #11
Curious. the Crucial MX100 was a pretty fair SSD even though it used a Marvell controller. The others? Not so much (to put it politely). The Mushkins used the problematic Sandforce controllers (even the Marvells were better than the Sandforces). Corsair used a Phison controller; not exactly a great one but better than Sandforce (pretty much anything was better than Sandforce, especially the early ones.

You buy low end products, you are going to get low end results. The Samsungs aren't cheap but, except for the early 840 EVOs, they have all been solid (even the problematic 840 EVOs were reliable; they were just really slow for an SSD).
I've been looking a little into possible issues with the controllers mentioned the past couple of days and there is quite a bit of information from users about problems with the Sandforce controller. So yeah perhaps this influenced the failures for those more than others. But the others were used moderately (maybe slightly more than what a "average user" might use) with many times being powered off/on and exposed to temperature variations that were probably a little wider in my case, especially in florida. So my conclusion is something like [moderate use w/occasional higher writes + wider temp variations = approx 1/3 - 2/3 less lifespan per 2 years].

I am going to give one last try with Mushkin since I've had a few good RAM and USB stuff from them. A Source 3D TLC SSD. It has an updated Silicon Motion controller. But I will have a higher quality backup drive for some things.
 
1 - 11 of 11 Posts
Top