Investigation: Is Your SSD More Reliable Than A Hard Drive?

Page 5 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
[citation][nom]ram1009[/nom]After all, The circuitry replacing the mechanics is at least an order of magnitude less likely to fail all things being equal.[/citation]

Common wisdom would believe this. I really WANT to believe this. But feedback from buyers is saying otherwise. Too many of the drives fail in the first week or month. Too many of them have "freeze" issues, lockup issues, losing capacity, etc. I think the key is "all things being equal", and obviously all things are not equal. Manufacturers are not building SSD electronics and SSD firmware to the quality that they are putting into rotating drives. I've been dealing with rotating drives for over 25 years and I don't remember the amounts of failures then in rotating hard drives as what we have now in SSD's.

I don't see how infant mortality of an SSD doesn't count as a failure. If I buy one and it fails in the first week, I'm not going to like it and I'm going to consider it a failed product. User feedback has proven that I'm less likely to see infant mortality in a rotating hard drive than in an SSD, so it does sound like an SSD issue.
 
[citation][nom]ntrceptr[/nom]Or better yet do what a real user would.- fill the drive to near capacity (only 5 GB free)- write a program that fill's and re-writes data to that last 5GB until writes fail.[/citation]

I don't believe SSD firmware works quite this way. I think the firmware knows when a part of the drive is being overexercised and it will swap positions with a part of the drive that it knows has not been used very frequently. It can even out the wear all over the drive that way. ("wear leveling"?)
http://en.wikipedia.org/wiki/Wear_leveling

 
I have an OCZ Vertex 2 running in my old Inspiron 1520, and I this laptop has taken many hard falls while running and the harddrive never fails.

I have had multiple External HDs fail because they fall on their side.

My $$$ Insvest == Money well spent!!!!

W/O My laptop and the stuff I have on it that I need all the time I would be on my way to super un-employment.
 
so what's the word about the new 'hybrid drives' in comparison to this article.?
also include WD VelociRatpors as well.
a little more detail and specifics/scenarios is what I'm asking for.
 
I work for a (possibly the) SoftLayer customer with the Intel X25-M drives (these are not a standard SL product, we persuaded them to special-order them for us). SL has rather more than 11 of these drives - we have three 4U servers stuffed full of them - but I can confirm that not one of them has failed in the time we've been using them. We have had a couple of X25-E drives fail though, and always without warning.
 
[citation][nom]cadder[/nom]I don't believe SSD firmware works quite this way. I think the firmware knows when a part of the drive is being overexercised and it will swap positions with a part of the drive that it knows has not been used very frequently. It can even out the wear all over the drive that way. ("wear leveling"?)http://en.wikipedia.org/wiki/Wear_leveling[/citation]

I understand the wear levling thats why I say fill the drive to near capicity. Then it does have much room to play with.
 
The failure rates of “enterprise” and “consumer” drives are very much similar.
Click to expand...
But the warranties of "enterprise" and "consumer" drives are not similar. So am I to understand that the price premium paid for enterprise level hardware is for the longer warranty?
 
[citation][nom]g00ey[/nom]But the warranties of "enterprise" and "consumer" drives are not similar. So am I to understand that the price premium paid for enterprise level hardware is for the longer warranty?[/citation]

Performance is different too. Think about SAS. This discussion only deals with failure rates.



I don't want to draw this out into an argument. But I would suggest reading the article in full. I never said that infant mortality doesn't count as a failure. I said that in a sufficiently large population, it's not a statistically relavant phenomenon. I would strongly suggest that you read the original Google and CMU studies.

User feedback is hardly a scientific study. It would never hold up as a claim and you'd be eaten alive if you were under any sort of academic review. Simply put user feedback doesn't work well for scientific studies. Think about the placebo effect. If a majority of placebo takers claimed an affect, would that make the sugar pill effective?

Furthermore, as you stated you have been dealing with spinners for 25 years. Have you dealt with the same volume of SSDs per time period? Have all of your SSDs been subjected to the same workload under the same conditions for testing? You do know that we're talking about media reliablity NOT durability or write endurance right?

Cheers,
Andrew Ku
TomsHardware.com
 
[citation][nom]Anonymous[/nom]You guys do the most comprehensive research I have ever seen. If I ever have a question about anything computer related, this is the first place I go to. Without a doubt the most knowledgeable site out there. Excellent article and keep up the good work.[/citation]

Tomshardware is great and I'm not trying to take a word of what you said away from them, but another source I suggest, particularly for SSD information, is Anandtech.com.

😉
 
As someone who thinks a bit like a physicist, I think it's a bit odd that people talk in terms of the platter-based hard drives as being "mechanical and have parts to wear out" and the solid-state hard drives as "having fewer parts to wear out".

I suggest looking at it from a microscopic viewpoint. Okay, sub-microscopic, at the atomic level. Inside a memory chip, that glass-smooth semi-conductive substrate the conductor material gets deposited on to make the circuit is all peaks and valleys, and the material that gets deposited isn't uniform, it's thicker here, thinner there, and after it's etched, wider here and there, and these things vary from one chip to the next, from one production run to the next. At this level, the differences would be easy enough to see.

So what, they all just sit there, right? Wrong. the molecules vibrate and move around. Generally not much because of the kind of material they are - but they aren't stationary and fixed, and it's incorrect to think of them as "not moving parts"; they in fact move all the time. More important, these things get heated up and cooled down all the time, expanding and contracting, moving at faster rates, then slower, and different materials - atoms and molecules - are effected differently by heating and cooling, so if they are "bound" to each other, that bond is stressed and can and does break. weak points on the hills and valleys give way, and those micro-circuits stop behaving the way we want them to, in the bigger world we live in failures occur.

Many of you have done the thing where you take a piece of copper wire and bend it rapidly back and forth until it breaks. Aluminum will do it even faster. A lot of metals will do this because the molecules change when you work the metal, and when you heat it up. When I think of the billions of circuits we have in some of our technology in this light, it's simply amazing to me our manufacturing technology is to the point where these things are as reliable as they are.

The old story about the first computer ever built was that some set of calculations showed there were so many parts it would fail before it would ever work properly - but they built it anyway. Look how far we've come in such a short time!

My thinking, when it comes to comparing the reliability of types of hard drives (just comparing reliability here, not anything else), is that the manufacturing process is what is in question, not the type of product. Can the SSDs be built to the same level of reliability because of the level of technologically developed skill of the manufacturers? The "mechanical" argument is a non-issue, it's beside the point, and really incorrect in the end.

The universe is a mechanical place - with the possible exception of what goes on inside neutron stars and black holes - from atoms to galaxies and beyond, really, I think.

My next rig will be built using an SSD (maybe 2) as it's primary work drive, I'm not worried much about it in any case, despite the fact that, to my way of thinking, it has more parts to wear out, not less. Besides, that will be at least a year from now, a relative long time for more bugs to get worked out and prices to drop.

😉
 
[citation][nom]Marcus52[/nom]Tomshardware is great and I'm not trying to take a word of what you said away from them, but another source I suggest, particularly for SSD information, is Anandtech.com.[/citation]

Is that a plug for AnandTech? 😛 Having worked for AT back in the early days, Anand is still an old friend and mentor. But, I'd like to think this article stands on its own.

Cheers,
Andrew Ku
TomsHardware.com
 
One main thing was forgotten:

with a standard hard drive, the chance of a recovery in case of a failure is much better. any lab can replace the PCB or even place the actual platters in a new identical drive.

with ssd's this remains a mystery whether or not the data can be restored.
 
Server farms with huge numbers of HDs or SSDs may give the biggest sample, but non-volatile storage is used elsewhere in far less benign environments. I have seen HDs fail earlier used offshore, where temperature cycling, humidity changes and vibration appear to reduce useful life. (We have changed out to SSDs for these reasons.)
There must be some papers relating HD MTBFs with vibration in particular. I would be interested in other folks' experiences using SDDs in higher vibration environments.
 
First of all, sorry for my english.

I find this article very informative.
But, i Think something is missing, why not suggest some of the best reliable SSD?

Right now, I'm looking for a Sata 3 reliable drive...and it's hard to find, when I see all the reviews on Newegg or amazon from buyers.

Thanks

 
[citation][nom]Caleb D[/nom]First of all, sorry for my english.I find this article very informative.But, i Think something is missing, why not suggest some of the best reliable SSD?Right now, I'm looking for a Sata 3 reliable drive...and it's hard to find, when I see all the reviews on Newegg or amazon from buyers.Thanks[/citation]

I don't think that there are any confirmed reliability studies. It is not even known if SSDs are more reliable than HDDs right now, though most people would say that they are (no moving parts)
 
How about a refresh of this article in January 2012, given that the SSD space is moving so quickly.
 


Sorry late to the game but just read through this article. Reliability goes a lot further than just SSD NAND. I've had two SSD's fail on me because of the controller. I think that's a much more likely cause of failure for SSD's and quite prominent. I'd be interested in getting that information over NAND write cycles.

Also I believe clarification on reliability and durability needs to be defined. SSD's are much more durable due to lack of moving parts, completely solid state. But reliability is still yet to be determined. I have three Intel X25-M's though that have been flawless. Had an OCZ Vertex 2 fail on me, as well as a Kingston card, both were controller failures. Thankfully replaced without question for brand spankin' new ones. So far so good. But it doesn't change the fact that they CAN fail and now more likely that you'll receive a refurb SSD if one fails than a new one.
 
it doesnt matter what you say about ssd reliablity, it doesnt change the following fact : 100% of ssd's ive installed and used have failed, whereas 99% of mechanical drives ive installed and used are still working reliably long after the ssds have died.
 
i have yet to own a ssd that lasts longer than 11 months. ssd are a total disaster in my experience. compare that to physical harddrives and ive not had a single failure - just one corrupted drive that needed reformatting, and these are drives that are over 10 years old.
 
Very informative article and comments. I'm just learning, researching, starting out today... I just want to know right now... what does "nm" stand for as in

"This topic is even more relevant now, in the wake of a swift shift from 3x nm NAND to flash memory manufactured at 25 nm."?

If possible, can you point me toward and article?

thanks, john b.
 
HDD tech goes back ~60y. Yes; have you seen one? They're the size of sewing cabinets or washing machines, and they're un-freaking-believably heavy. Despite their gigantic physical size, they hold remarkably little data. And they're slow. And each such drive uses about as much electricity as a modern rural neighborhood. Nevertheless, these ancient drives are important to help illustrate a point:

Drives of all types are generally very much improved over the first examples of rapid-i/o semi-permanent storage systems.

If you want maximum permanence of data, use something like glass tablets through which 2mm-diameter holes have been melted: it's extremely low-density, and it's fragile, but as long as you have a way of deciphering what the holes mean and the glass plates don't break, it's got a nearly infinite data-integrity lifetime.

There aren't any "mobile bits" and, provided the plates are clean (no holes occluded by waste or vandalism, etc.), the data isn't likely to be corrupted or lost, even over periods of hundreds of thousands of years. Nevertheless, such a storage system is generally impractical and, on a meta- or omni-inclusive- scale, impracticable.

I've seen studies claiming magnetic tape has far greater (both) longevity and reliability than optical discs; I don't see a rush to DT backup, let alone high-i/o storage. Ask anyone who has had a child, or who personally has experienced one or more moments of clumsiness or poor judgment -- you know: the person whose tapes were stored in the automobile's center console when the Mr. Pibb got dumped in there, or the person who had to clean a PB&J sandwich from the VCR's innards -- and you'll likely discover someone who believes optical discs are far more rugged and reliable than tapes.

In short, the specifics of the operating environment are hugely important in determining the relative performance of different technologies, but that's the meta concern.

Studies such as these wrongly attempt to ascribe to a host of devices superficially similar but in important ways internally different, classification characteristics identifying the relative merit of the technology of the superclass.

In short, truly meaningful comparisons may be made only between identical architectures, technological executions and applications: each such superset then being compared and contrasted against a meaningfully competitive representative superset. Each such superset may be restricted to a particular make, model and revision, type, capacity, installation, end use, etc.

In the end, some trends may emerge; however, the present research seems hardly any more credible than forcing data to support the biases of the observer; the alternative is even worse: an analyst incompetent to recognize the importance of the foregoing predicate.

The tomshardware site timely provides ordinarily remarkably useful information, and while I agree with the author's conclusion that further observation and study is warranted, I disagree with the summary conclusion that all HDDs may for the purposes of competent understanding and analysis be lumped together with the cited exemplars, and that all SSDs may likewise be lumped.
 
I read the article word for word, and really appreciate it. I'm a gamer, and a person who values the huge amount of stuff I've built up over the years in "My Documents" folder. I want fast, but what sticks in my mind, as I contemplate whether to go full ssd or stay with my ssd-hdd combination, is that hdd's really do tend to fail "more gracefully." You can usually feel the failure coming in time to move your valuable information to a new drive. I've even stuck a failing hdd in the freezer, so i could spin them one last to get my precious data out. heck, that's almost fun! So, for now i'll continue with the gamer way of programs on ssd and files I really want to save on hdd.

But, because of this article, I'll probably also add a new behavior: backing up as a regular habit rather than as a rarity. That's one advice of the article that is very useful.

Thanks again. Your site does offer some very nice consumer-friendly serious research.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom