Two Samsung 840 Pro's & 1 840 have failed, 4 at [H]ardOCP (Early review sample SSD's)

[MrBighouse]

Damn it! Just when I pull the trigger and buy an 840 and convince my dad to do the same.... Won't be selling my old spinner HD anytime soon.

[Mr. Pedantic]

Originally Posted by serpretetsky

I'm not sure i'm understanding your question. Are you asking why it's important to find out why a design is failing in the real world?

Yes, if several hundred/thousand are failing at the same time. Here, we're talking about 2. Certainly, it's not a good sign that a reviewer gets 2 and they both fail, but it's not outside the realm of possibility that it could happen completely by chance and indeed, it has to happen some time.

[garage_logician]

Great. just got my 840 Pro last night to replace a Crucial Nuetron. :-/

[Hellhammer]

Originally Posted by Mr. Pedantic

Yes, if several hundred/thousand are failing at the same time. Here, we're talking about 2. Certainly, it's not a good sign that a reviewer gets 2 and they both fail, but it's not outside the realm of possibility that it could happen completely by chance and indeed, it has to happen some time.

It's still important to find the cause because the issue could be widespread. For example, if the issue lies in the firmware, all drives are affected. That doesn't mean that all will die because the bug might require a specific workload to be triggered, but it means that all units have the potential to fail. If the issue is in hardware, then Samsung needs to figure out what is failing and what is the reason. Is it simply a production error or is there something wrong in the actual design?

When the first sample died, I strongly believed that it was just a lemon because that happens, the 840 Pro wasn't the first SSD to die on our hands. Now I'm starting to think that there is really problem in the 840 Pro and I agree with Anand that we can't recommend the drive until we hear more from Samsung.

[ensign_lee]

...but but...newegg is selling the 840 pro with assassin's creed 3 for free.

[jwilliams4200]

Originally Posted by Mr. Pedantic

Yes, if several hundred/thousand are failing at the same time. Here, we're talking about 2. Certainly, it's not a good sign that a reviewer gets 2 and they both fail, but it's not outside the realm of possibility that it could happen completely by chance and indeed, it has to happen some time.

That argument, while strictly true, is completely useless.

In the real world, we know that products often do have defects, and both consumers and manufacturers would like to have the defects fixed.

So a more useful question is how do defects get fixed after a product has already undergone its initial validation testing, and some units have passed production testing and been released into the field?

A common answer is that units that fail in the field can be carefully analyzed to determine the cause of failure, and if a widespread or serious failure mode is recognized, then the search for a remedy begins.

Also, it is interesting to try to frame your specious argument quantitatively, to try to make it actually worth something. So, note that a failure rate of about 0.5% per year is commonly considered very good for SSDs (Intel and Plextor have claimed to have failure rates of about 0.5% per year).

If the failures of the two 840 Pro SSDs at anandtech were random and independent (uncorrelated) and the chance of failure of each is less than 0.5% (technically, significantly less since they failed in less than a week each, but for argument sake I will skip that complication), then the probability that two out of two failed is longer than 1 in 40,000. That already makes any reasonable person question our initial assumption of random and independent failures. Add to that the observation that the two SSDs failed with similar symptoms and after running similar workloads, and it looks even more likely that the two failures were correlated in some manner.

So, if the failures were correlated, what can we conclude? Still nothing definite, since it is possible that the two SSDs were both exposed to some stress outside of their specification that caused them to fail. But I tend to discount that possibility since Anand is experienced at testing SSDs and I would expect that if he changed his test equipment recently (or had his test equipment fail in a way that causes undue stress on the SSDs) that it is likely he would have corrected it before now. Therefore, the most likely cause of the correlated failures is a defect in the 840 Pro SSDs.

Now it is incumbent upon Samsung to do a failure analysis on the units, and hopefully get back to anandtech with a failure report.

[jwilliams4200]

There is a guy who works in the IT / storage department of a large company who frequently posts on hardforum, and he is reporting three out of four 840 Pros have failed their standard burn-in / acceptance tests:

http://hardforum.com/showpost.php?p=...4&postcount=34

I'm thinking that the issue is not caused by firmware, since one out of four of his 840 Pros is chugging along fine under the same stress.

Also, my 840 Pro is still alive and kicking. I've been hammering it with sequential and random writes, but it is still going.

Right now, Anand's explanation of the power circuit failing is sounding likely. Maybe the power circuit is marginal, and fails on some units but not all units.

[hhhd1]

Even if it is a circuit issue, it might not be impossible to fix using a firmware update, if the firmware is able to do some voltage control.

Waiting for samsung actions about this ..

[Remobz]

Originally Posted by hhhd1

Even if it is a circuit issue, it might not be impossible to fix using a firmware update, if the firmware is able to do some voltage control.

Waiting for samsung actions about this ..

Hoping they do this within a few weeks at the latest.

[Zap]

Originally Posted by jwilliams4200

There is a guy who works in the IT / storage department of a large company who frequently posts on hardforum, and he is reporting three out of four 840 Pros have failed their standard burn-in / acceptance tests:

http://hardforum.com/showpost.php?p=...4&postcount=34

Wow, that's even worse, because an argument can be made about the reviewer's units dying due to being pre-production or whatever (whether or not it was) but this is end user units dying.

[corkyg]

Isolating such faults is always difficult. Consider that the computers hosting the various failed drives may themselves have some condition that triggers the failure, whereas others seem to work as advertised. To really solve the problem, Samsung may have to replicate the machines that hosted failure - and see if they have a common thread.

With the holiday season coming up, they need some rapid response.

[GPz1100]

^^Indeed** I was/am prepared to pay the premium, but if reliability is questionable, i'll either end up going with the 830 or the plextor**

[Mr. Pedantic]

Originally Posted by Hellhammer

It's still important to find the cause because the issue could be widespread. For example, if the issue lies in the firmware, all drives are affected. That doesn't mean that all will die because the bug might require a specific workload to be triggered, but it means that all units have the potential to fail. If the issue is in hardware, then Samsung needs to figure out what is failing and what is the reason. Is it simply a production error or is there something wrong in the actual design?

When the first sample died, I strongly believed that it was just a lemon because that happens, the 840 Pro wasn't the first SSD to die on our hands. Now I'm starting to think that there is really problem in the 840 Pro and I agree with Anand that we can't recommend the drive until we hear more from Samsung.

Fair enough.

If you put your personal SSD through the same workload as your review drives, would that fail as well?

[philipma1957]

Does this affect the standard samsung 840 ssd's. version



SAM MZ7TD500BW | UPC # 887276006659 | Mfr. Part # MZ7TD500HAFV-0BW00


I can get the 500 gb size for under 340 usd.

I would hate to buy it and have it go bad.

[jwilliams4200]

I have seen no reports of any failures for Samsung 840 SSDs. However, since the 840 SSDs are intended for lighter workloads than the 840 Pro SSDs, people probably have not been stressing the 840 SSDs as hard as the 840 Pro SSDs.

Because the 840 SSDs are still quite new, and because using TLC flash in an SSD is new, I suggest waiting a few months before using an 840 in any application where a failure would be a big hassle.

If I wanted to buy a high quality 512GB SSD right now at a low price, I'd go with a 512GB Plextor M3, which is $360 at newegg.

[philipma1957]

thanks I have a few samsung 830's in the 512gb size.


I think I will just past on the deal.

[videoclone]

Samsung 830 512GB would be much better then the Samsung 840 and 840 pro as a long term drive. i wonder just how long these 830's will last.. SSD's are ment to go for 10+ years and with lower writes to the drive it could last 20+ years

[davewolfs]

Here is my take.

I've had the 840 PRO for almost two weeks now. System has been on 24/7. Not a gaming system. A workstation.

Perhaps I had lemons, but going through two Intel 520's based on this sites recommendation that all bugs had been worked out is probably the worst hardware advice I have ever taken. This drive with my hardware was an EPIC FAILURE, yet it is regarded by this site as a decent drive.

Now, for me, I have not had a SINGLE issue with the 840 PRO. I also find it hilarious that this site decided to update their article ON THE RELEASE DATE OF THE DRIVE, that a second drive had failed. Perfect timing.

I am taking the Anandtech comments as a grain of salt until proven otherwise.

[jwilliams4200]

Originally Posted by davewolfs

I also find it hilarious that this site decided to update their article ON THE RELEASE DATE OF THE DRIVE, that a second drive had failed.

Why is it hilarious? It seems like a boring coincidence to me -- Samsung sent a second SSD, anandtech tested it, it failed, anandtech updated the article.

You wouldn't happen to be the same person who was posting a few days ago as "katkah", would you?

[videoclone]

Originally Posted by jwilliams4200

Why is it hilarious? It seems like a boring coincidence to me -- Samsung sent a second SSD, anandtech tested it, it failed, anandtech updated the article.

You wouldn't happen to be the same person who was posting a few days ago as "katkah", would you?

Anandtech does nothing but praise Samsung 830, 840, and 840 Pro drives.. They do state that 2 840 Pro drives have died while testing...

Do you think they should "NOT post information on drive failures? Do you think they are being paid bribe money by one of the other 10 SSD players in the market to bag out Samsung's 840 Pro ?... do you think Samsung never got a failed drive back and never sent another one for them to finish testing the 840 Pro with? ...

The Intel 520 is a Sandforce drive.. IT does come with the risk of being a dude due to the controller that has had many issues over its lifetime... my Sandisk 256 SSD with the same Sandforce controller has been rock solid for 6 months... Just because Intel failed to provide you with a rock solid part doesnt mean you got bad advice from here!... Most people on this site would always recommend a Samsung 830 256GB drive over any other out on the market.. it really is the best of the best

[jwilliams4200]

Originally Posted by videoclone

Anandtech does nothing but praise Samsung 830, 840, and 840 Pro drives.. They do state that 2 840 Pro drives have died while testing...

Does anyone else have trouble making sense out of this post? It begins with the two contradictory statements quoted above, and deteriorates from there. The whole post looks like a string of random statements that make no sense when put together. I would typically ignore something like that, but this one happened to quote me so I read it twice and I still cannot figure out any coherent message contained within the post.

[jwilliams4200]

The fourth (of four) Samsung 840 Pro SSDs just failed for the guy at hardforum.com:

http://hardforum.com/showpost.php?p=...8&postcount=65

Note that although he says here "840" he really means "840 Pro" since that is what he mentioned in his other post.

[videoclone]

Originally Posted by jwilliams4200

Does anyone else have trouble making sense out of this post? It begins with the two contradictory statements quoted above, and deteriorates from there. The whole post looks like a string of random statements that make no sense when put together. I would typically ignore something like that, but this one happened to quote me so I read it twice and I still cannot figure out any coherent message contained within the post.

sorry i quoted the wrong person. I clicked below the wrong box.
It was a reply to what "davewolfs" said about Anandtech being an unreliable source for information and to take the two failed drives with a grain of salt

[davewolfs]

Originally Posted by videoclone

sorry i quoted the wrong person... a click below the wrong box.
It was a reply to what davewolfs said about Anandtech being a dodgy about the 840 Pro's failing

I really don't care about what Anandtech said. Seriously. The drive works for me so far and I will report otherwise if it dies.

What I do find dodgy is the timing in which they reported the drive failing.

I mean really, the tester woke up Tuesday morning and the drive just failed? Get real.

[videoclone]

Originally Posted by davewolfs

I mean really, the tester woke up Tuesday morning and the drive just failed? Get real.

That is strange yes. It could also be a coincidence.. either way i would still make sure you have your files backed up just to be safe ( As for the drive itself, it comes with warranty so no worries )