SSD reliablity dependance on what main factor?

[hal2kilo]

Bought an X-25M Intel 120 GB drive and have never looked back. Works like a hose. I only have Sata 1 on my machine so it's all good.

 

[Charlie98]

Sample: 1
Days running/wear level: Unknown

Speaking as the lone generic retail user here... :awe:

Anyone remember the Chrysler commercial with Lee Iacocca and Snoop Dogg? ...Snoop talks about getting the hookup with the nephew, for sure...

"You know, I'm not too sure what you just said, now everybody gets a great deal..."

From my point of view, my OCZ SSD was a great buy, works well and has given me zero problems (50 days in service, wear level... probably nil.) As far as all that benchmark testing... I don't really care; it's faster than an old HDD so what's not to like?

 

[taltamir]

Yes I am an SSD n00b and didn't know about firmware, but you have to start somewhere.

A quantum mechanics professor answering a simplistic question of an interested student saying "Your question is dumb" is kind of a dick.

If you know nothing about it then you are not qualified to be setting up polls about it.
You should start a thread asking "hey, what is the casue of X".
the only reason to set up a poll is if there is a disagreement and you are curious to see what different people vote on

Also he said thread/poll is dumb not the question. The question (what causes failure) is legitimate.
The thread/poll has 2 dumb things
1. "OCZ is known for their poor reliability" - sounds like flame bait.
2. Made a multi choice question where every single option is wrong because the author is clueless about the subject (according to your own admition).
Pro Tip: If you know nothing about a subject then don't presume to make your questions into a multiple choice question but let people educate you instead.

 

[Coup27]

Well said.

 

[Helicobacter]

If you know nothing about it then you are not qualified to be setting up polls about it.
You should start a thread asking "hey, what is the casue of X".
the only reason to set up a poll is if there is a disagreement and you are curious to see what different people vote on

Also he said thread/poll is dumb not the question. The question (what causes failure) is legitimate.
The thread/poll has 2 dumb things
1. "OCZ is known for their poor reliability" - sounds like flame bait.
2. Made a multi choice question where every single option is wrong because the author is clueless about the subject (according to your own admition).
Pro Tip: If you know nothing about a subject then don't presume to make your questions into a multiple choice question but let people educate you instead.

You have a point there, but I think your reaction is over-the-top.

Regarding point 1: Yes, it's provocative it is but is it not true, based on the statistics? Just because it's provocative doesn't mean it's dumb or untrue.
Regarding point 2: I have no evidence (I only know the stats from one retailer and a user survey of a large sample of SSD owners), but I'd be willing to lay strong odds that:

Given a large enough return hub, most of the variation (85+%) in

failure/RMA/return proportion of a particular SSD model

can be explained with just the three poll choices as factors (and perhaps some interaction terms between them) - that's assuming you only look at a particular time frame. Because firmware is specific to the controller and vendor you will reach very accurate results. Also, I asked what is the single most important factor, and one of the three terms will have the greatest effect against the response.

 

[Coup27]

So to summarise, you have made a poll on a subject you knew nothing about with the wrong set of choices. Then you have ignored all of the information from informed users and insist on clubbing together controller and firmware, ignoring all of the technical aspects which go along with them, just so your poll of 3 choices makes sense?

 

[Helicobacter]

with the wrong set of choices.

Have you even read the last post?

Then you have ignored all of the information from informed users

Who says I ignored them?

and insist on clubbing together controller and firmware

I never made that claim.

, ignoring all of the technical aspects which go along with them,

How is picking higher level of granulation wrong?

If someone asks for why people sneeze and someone puts up a survey with choices: A cold, allergy, fever, dust. Then someone else complains, "well it's really just mostly about these proteins binding with those tissues, and then it's triggering the sneeze..."

You want to go one level below "vendor" and say: "well, this mobo manufacturer uses so and so feature, and this doesn't work with this SSD because firmware 2.x defaults to this setting"...fine, all I wanted is to know if it's a reliable heuristic to pick an SSD vendor I can trust not to go with unvalidated controllers, doing very strict QA, and being on top of things - and this seems to be the case with Intel, Samsung and perhaps - recently - Crucial. All that matters to me is:

All else being equal, if I buy this particular SSD, what are the odds that I'll have problems within a time frame of 5 years?

Anyway, there is not much value-add in going to engage these bitter and patchy statements going forward.

 

[Puffnstuff]

Well I've owned two different ssd's from different manufacturers both with sf1222 controllers and they've both failed in the same way with the ocz lasting the longest of the two. OCZ only took a week to turn around the drive once they received it. The other guys, adata, took 1 month to complete the rma. The adata is faster however I don't trust them which is why I'll clone this drive once the ocz comes in and put it away for when the adata fails again. My next ssd will either be intel or samsung.

 

[Obsoleet]

'SSD board manufacturer '

= firmware, QA

At least that's the closest to firmware/QA that I saw on your choices. I think the 510/M4 proved Intel and Crucial can turn Marvell chips into industry-best SSDs.

The Intel 520 will do it again, this time with Sandforce (probably SF).

It's about buying from Samsung, Intel or Crucial.. not the flash or controller.

 

[Obsoleet]

Well I've owned two different ssd's from different manufacturers both with sf1222 controllers and they've both failed in the same way with the ocz lasting the longest of the two. OCZ only took a week to turn around the drive once they received it. The other guys, adata, took 1 month to complete the rma. The adata is faster however I don't trust them which is why I'll clone this drive once the ocz comes in and put it away for when the adata fails again. My next ssd with either be intel or samsung.

Wise choice.
I wish all forumgoers would heed this, or would have heeded this truth since the Intel G1 release (the 1st truly reliable SSD).

 

[groberts101]

FUD. All that matters to most is that they paid less than the Intel.. got more performance than that drive.. and it actually still works.

Some obviously weren't so lucky since I got all the good ones running on my machine. lol

And I should be offering an Intel in return for every two of those locked drives so I can d-flash them back to life and sell them to others who know how to work around their compatability shortcomings.

No returns though unless you use the power configs and hardware/drivers of my choice. :biggrin:

kidding(mostly anyways).. and I do actually feel for those who had/will have issues with that particular controller. Not knowing "what's up" will eventaully bite some in the long run with that first gen SF.

 

[Obsoleet]

FUD. All that matters to most is that they paid less than the Intel.. got more performance than that drive.. and it actually still works.

I don't think that's a concern for those who want one of the most reliable drives on the market.

 

[Shmee]

I voted for controller, at least that seems to me to be the most contributing factor of options present.

With that said, Heliobacter has done nothing wrong, and I applaud him for asking questions. Some other have been a bit harsh here on their responses, and while some do make a point, I believe the attitude is uncalled for.

 

[RampantAndroid]

Wise choice.
I wish all forumgoers would heed this, or would have heeded this truth since the Intel G1 release (the 1st truly reliable SSD).

My intel 510 SSD took 6 months to die. Waiting for a replacement, but I'm not spending another DIME on SSDs for the next 5 years. To hell with em. Magnetic drives are just more reliable.

 

[Veliko]

My intel 510 SSD took 6 months to die. Waiting for a replacement, but I'm not spending another DIME on SSDs for the next 5 years. To hell with em. Magnetic drives are just more reliable.

What are you basing this assumption on?

 

[RampantAndroid]

What are you basing this assumption on?

Experience with SSDs at work, personally, and with friends. And obviously, 20+ years of being around magnetic drives. They're slower, but I will NEVER trust critical data to an SSD. EVER.

Oh, also, need we remember the Vertex 2s?

 

[Veliko]

Experience with SSDs at work, personally, and with friends. And obviously, 20+ years of being around magnetic drives. They're slower, but I will NEVER trust critical data to an SSD. EVER.

Oh, also, need we remember the Vertex 2s?

Do we need to remember the IBM Deathstars?

I have had several magnetic drives fail over the years. Didn't Google publish some data that showed HDDs weren't as reliable as people thought?

 

[pandemonium]

Ever since SSD's introduction to the market I've mainly just been waiting due to the cost/available space to drop before jumping on it. Now, it's reliability.

So far I have 3 internal WD magnetic drives that are still functioning perfectly: 9 years old, 6 years old, and 3 years old. For their performances, I can't really complain.

When data centers around the world actually start using SSDs, I'll consider it.

 

[taltamir]

Ever since SSD's introduction to the market I've mainly just been waiting due to the cost/available space to drop before jumping on it. Now, it's reliability.

So far I have 3 internal WD magnetic drives that are still functioning perfectly: 9 years old, 6 years old, and 3 years old. For their performances, I can't really complain.

When data centers around the world actually start using SSDs, I'll consider it.

Datacenters are using SSDs when they need speed. Otherwise they stick with HDD for cost not for reliability.
You got lucky with some spindle disks and unlucky with an SSD. Sample size is super small and it proves nothing. There are plenty of figures that show your experience is extremely unusual.

But meh, if you feel like not getting an SSD due to reliability concerns. well no skin off my back.

You should be keeping backups regardless of your media type not switching companies/storage type when you lose your data due to lack of backups.

 

[Motorheader]

People get caught up in benchmarks that it becomes a detriment to system reliability and performance. SSD drives were never intended to replace having to backup your data.

It's well established that SSD drives and spinners are not without their caveats. The biggest issue as a tech I've seen with an SSD is data recovery of any type if the drive goes bad. Having said that I go the way of the OEM's when it comes to my and anyone elses data with any SSD - primarily Samsung, Intel, and Toshiba based.

The poll doesn't directly address what is probably the biggest factor in drive reliability which is firmware. The reliability of the dram and pcb have decades of r&d so I don't think that is an issue. Look at all of those WD Silicon Blue 256gb SSD drives that Newegg blew out in early July. You rarely if ever hear/read about anyone bringing up that they had an issue with the drive. It wasn't a world beater controller (second/third rev JMicron controller) or the fastest drive as far as benchmarks go but it seems to just work.

 

[Obsoleet]

Experience with SSDs at work, personally, and with friends. And obviously, 20+ years of being around magnetic drives. They're slower, but I will NEVER trust critical data to an SSD. EVER.

Oh, also, need we remember the Vertex 2s?

I wouldn't trust critical data to a single SSD or HDD. Nothing in the world can replace a backup or a mirror.

I work in one of the largest hosting environments in the world, a 15 billion dollar business.. we use everything and there isn't a single drive by itself that is entrusted by itself with business-critical data.
Sorry to hear about your trouble with the failed drive, I experienced that a number of times with a number of OCZ drives and haven't had trouble since avoiding those but I hope you have better experiences going forward.

 

[RampantAndroid]

Do we need to remember the IBM Deathstars?

I have had several magnetic drives fail over the years. Didn't Google publish some data that showed HDDs weren't as reliable as people thought?

In the last 12 years, let's see if I can get this list even halfway right.

2000 - 40GB drive. WD. Lasted 7 years, retired.
2001 - 30GB drive. WD. Lasted 3 years, retired.
2003 - 2x 36GB Raptors. Lasted 5 years, retired.
2005 - 2x 74GB Raptors. Lasted 5 years, retired. Still on my desk. A few bad sectors, IIRC.
2003, 160GB WD Drive. Still in use, I think (parents)
2005 - Seagate drive DOA for my father.
2005(?) 640GB external MyBook drive, made internal, still in use.
2007 - 2x 750GB Drive, seagate. For my dad's system doing video work. He wanted space...still in use.

More recently, 2008 750GB Seagate. Still in use
1TB drive in 2009(?) still in use
2TB Drive 2010, still in use

Number of drives that have failed on me while in use: 0. Bad sectors? Sure. Outright "drive won't even spin up"? None. Going back further, there's a few 700MB drives in there that never died. There's a 20MB hard drive that is 5.25 in size. Last I knew it still worked (probably last booted that system in 2000 or so, for s**** and giggles.)

I have a history of spinning drives that WORK. I back up some data to 2 places (one is my own HD, one is my web server on a user account that is accessible only through SSH, and the host backs up the drives.

 

[frostedflakes]

My intel 510 SSD took 6 months to die. Waiting for a replacement, but I'm not spending another DIME on SSDs for the next 5 years. To hell with em. Magnetic drives are just more reliable.

It depends on the SSD. I can't speak for the Intel 510, but the X25-M and X25-E have a proven track record of being extremely reliable drives. Intel claims ~0.4% annual failure rate for the X25-M based on a sample size of >1 million drives in deployment. That's only 4 out of every 1000 SSDs they sell. You need to look at the aggregate data, personal anecdotes are pretty meaningless for drawing conclusions about reliability, although I can understand that if a person has a brand or product that fails on them it does negatively affect their perception of that brand or product and make them less likely to buy it in the future. That's just human nature I guess.

Tom's also did an article examining SSD reliability in a couple data centers that have a large number of SSDs and HDDs deployed. The results were basically the same, the Intel drives tended to be very solid with low failure rates.

http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923.html

The article also brings up an interesting point, which is that generally a single SSD is replacing multiple HDDs, which further compounds the reliability advantage of SSDs, at least in I/O heavy servers.

We simply get significantly higher I/O [with SSDs] at a lower cost than we'd be able to get with standard drives. We've had many customers needing more I/O than what 4x 15k RPM SAS drives in RAID 10 provide, and an upgrade involves moving to a larger server chassis to support more than four drives, a larger RAID card, etc. Other configurations have needed 16+ 15k RPM drives to get the necessary I/O. Going with a single SSD (or a couple SSDs in RAID) greatly simplifies the configurations and makes them much cheaper overall.

That is then compounded by the fact that you generally use one SSD to replace 4+ standard drives on average. You're then looking at a 20%+ AFR with hard drives and 1.6% with an SSD.

 

[Helicobacter]

I wish there was a public Pareto chart of SSD failures.

When the consensus is firmware is the leading root cause, is it because of compatibility issues? Why would so many SSD's suddenly die after, say, 150-1000 days if that were the case? Only the minority (i.e., enthusiasts) flash their devices every now and then to cause such an issue after months of seamless operation. The mainstream user keeps the hardware and its firmware constant - why would things suddenly go wrong after months because of firmware code?

Is it not the quality assurance level of NAND- and controller-chips in the manufacturing process that's at play?

 

[LokutusofBorg]

Software (firmware is software) is inherently complex, and it's a mathematical bylaw that you cannot test all failure scenarios.

Again, if you have to ask why things would suddenly go wrong after months of use because of (not-updated) firmware code, then you're just making it all the more obvious that you're not very up to speed on SSDs at all. I would suggest you go over to the Anandtech main site and read up on all the SSD articles. Trying to grow your knowledge by asking questions in the forum is admirable, but you're not doing that. You're here trying to posit your theories, trying to answer the questions before you even know what questions to ask.