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

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
[citation][nom]cadder[/nom]I've been interested in the 120-128GB range[...] 6822136322[/citation]
You are not taking into account the likely hood of someone posting a review. I would be more likely to post a review if my drive failed in order to warn others than I would to post a review if nothing went wrong.
 

Jax69

Distinguished
Feb 7, 2007
22
0
18,510
one of my HDDs is a 60GB Seagate Baracuda from 2002 and still no problem whatsoever, beat that SSD...:))
 

razor512

Distinguished
Jun 16, 2007
2,130
68
19,890
How about saving up for a few SSD's then kill them

Run spinrite on the SSD's in level 4

(A friend of mine was able to kill a intel x25 v in about a day using spinrite)

Not all SSD's can do the write cycles advertised. No way to tell how much data spinrite is writing to the drive but it can kill a SSD pretty quickly.

Same goes for using one and having your page file on it.

I really want one but will have to wait until write reliability gets to a point where I can have a few of my video editors and photo editors use a SSD for cache
 

cadder

Distinguished
Nov 17, 2008
1,711
1
19,865
[citation][nom]Enzo Matrix[/nom]You are not taking into account the likely hood of someone posting a review. I would be more likely to post a review if my drive failed in order to warn others than I would to post a review if nothing went wrong.[/citation]

That probably does mean something. For instance the average SSD seemed to have 25% bad reviews. I would not want to buy a product where I had a 1 in 4 chance of being extremely dissatisfied, but if only 1/5 of the satisfied buyers posted reviews, then instead of 1 in 4 dissatisfied it would be 1 in 16 dissatisfied which looks a lot better.

OTOH that is why I posted the numbers for the WD mechanical drives. On the average they got a lot better reviews than the SSD's, so you if you applied the same logic of dissatisfied buyers more likely to post reviews, then the mechanical drives overall had fewer dissatisfied buyers.
 

ram1009

Distinguished
IMHO, the supposition that lack of moving parts doesn't equal improved reliability is nuts. I understand that lack of empirical evidence precludes making such a statement today but common sense dictates otherwise. After all, The circuitry replacing the mechanics is at least an order of magnitude less likely to fail all things being equal. As has been pointed out component "infant mortality" issues don't count as SSD failures.
 
G

Guest

Guest
I have a computer store where we do a lot of laptop repairs. We see a massively higher amount of hard drive failure on laptops when compared with desktops. We have been recommending SSDs as replacement drives on laptop drive failures and have yet to see a failure when the customer chooses to go this way. Our sample size is relatively small, as most people still choose the cheaper HDD option, so I won't claim that what we have seen is conclusive, but it certainly seems promising.

On another note, you metioneded that drives that were in a RAID together with a failed drive were more likely to fail. I believe this is because these drives are more likely to come from the same batch (in this case probably bad batches). Bad batches crop up from time to time. I know we have had long stretches of drives with no issues, then we get a bunch of bad ones in the same box.
 

razor512

Distinguished
Jun 16, 2007
2,130
68
19,890
To clarify my point about saving up for 1 or 2 SSD's is to test how much data can actually be written to them,

Just as you at tomshardware have said, the MTBF is in no way accurate and if thats the case then why do you think their claims of write cycles will be anymore honest?

Get a ssd and and get some program to do a write benchmark to the drive. the new SSD's have good write speeds so it shouldn't take too long to write a few TB of data to them in the form of a write benchmark.

If their claims are true then the SSD should survive more than 2 days of writing.

 

razor512

Distinguished
Jun 16, 2007
2,130
68
19,890
[citation][nom]bobwya[/nom]Great article.Shame that THG keeps wiping my carefully crafted and thought-provoking comments every time I login... :)[/citation]


the site currently has a problem where you will randomly get logged out, and if you type up a comment and try to post it, it will say you are not logged in, and if you enter your login info before continuing, it will log you in but not post your comment and you will not be posted and you cant go back to copy and paste it for a new comment
 

NuclearShadow

Distinguished
Sep 20, 2007
1,535
0
19,810
While SSD drives are no doubt better performance and like the article says brand does have a lot to do with quality and failure rate. They are no where near going mainstream at this time. The average PC user does not have a strong enough reason to go for them at their current price to storage ratio. Even most gamers don't have a reason for the switch as its biggest advantage is loading times in games and the modern game takes up a huge amount of space.

Unless you are truly going to use the SSD for its real capabilities there is little reason to make the switch before it hits mainstream.
 

mayne92

Distinguished
Nov 6, 2009
743
0
18,980
Excellent article Andy! I started reading this during my lunch break at work and couldn't stop reading; now I'm 32 minutes over lunch :( - Excellent article and a great read!
 

acku

Distinguished
Sep 6, 2010
559
0
18,980
[citation][nom]mayne92[/nom]Excellent article Andy! I started reading this during my lunch break at work and couldn't stop reading; now I'm 32 minutes over lunch - Excellent article and a great read![/citation]

That's what I like to hear... Though, I'm not crazy about killing your lunch break. :) If I could, I would give you 32 minutes back but we have yet to perfect our time machine here in the lab.


Cheers,
Andrew Ku
TomsHardware.com
 

ntrceptr

Distinguished
Jan 26, 2006
97
0
18,630
I still believe this all depends on how full the drive is, which it should. The formulas used are assuming the drive is empty and you have all that space to work with otherwise there should be very small difference in the failer time for 80gb vs 160gb drives.
I understand they allocate slack space to be used when cells start failing and the bigger the drive the more of this space there is.

The failure times 80GB = 18.7 yrs and 160GB = 37.5 yrs. That's a load of crap unless they are calculating with all free space and who has that, Isn't it on the same 25nm cells???

Do i need to write a program for you to test this.
- Analyise drive layout/size/sectors etc.
- find a beloved patriot of free space on the drive (according to what the drive returns, since in reality it won't be the same physical space everytime you write to xxxx sector)
- re-write data to that chunk of space until writes fail.

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.

Note: be sure the program used writes different data each time to avoid the firmware not actually re-writing to that cell.

I bet it's much much lower than the manufacturers are claiming.
Whos fills up there SSD like this, probably not many home users.
Enterprise systems using these for databases may have this situation more often.

I think tom's should do a live write test of a nealry full SSD so we can watch it fail within days, if that long.
 

acku

Distinguished
Sep 6, 2010
559
0
18,980
[citation][nom]ntrceptr[/nom]I still believe this all depends on how full the drive is, which it should. The formulas used are assuming the drive is empty and you have all that space to work with otherwise there should be very small difference in the failer time for 80gb vs 160gb drives. I understand they allocate slack space to be used when cells start failing and the bigger the drive the more of this space there is.The failure times 80GB = 18.7 yrs and 160GB = 37.5 yrs. That's a load of crap unless they are calculating with all free space and who has that, Isn't it on the same 25nm cells???Do i need to write a program for you to test this.- Analyise drive layout/size/sectors etc.- find a beloved patriot of free space on the drive (according to what the drive returns, since in reality it won't be the same physical space everytime you write to xxxx sector)- re-write data to that chunk of space until writes fail.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.Note: be sure the program used writes different data each time to avoid the firmware not actually re-writing to that cell.I bet it's much much lower than the manufacturers are claiming.Whos fills up there SSD like this, probably not many home users.Enterprise systems using these for databases may have this situation more often.I think tom's should do a live write test of a nealry full SSD so we can watch it fail within days, if that long.[/citation]

Again, that is write endurance. It is a spec'ed failure. We are looking at random failures.
 

bobwya

Guest
May 21, 2005
692
0
18,980
[citation][nom]acku[/nom]Again, that is write endurance. It is a spec'ed failure. We are looking at random failures.[/citation]

Yeh. People don't seem to be reading the article! It's a difficult subject to cover since the failures are by their nature hard to pin down and dependent on a wide variety of factors (firmware bugs/glitches, component tolerances/manufacturing defects, power fluctuations - for the majority of home users anyway, etc.)

Obviously everyone that read through the article would observe that the SSD statistical data is a little short in the temporal domain (2 years of large-scale observations doesn't cut it really).

It would be interesting to see a larger sample set comparing newer generation SSDs with older SSDs (which might indicate more stable firmware revisions and more stable NAND cells - hmmm does the latter equate to greater overall reliability or only better wear-levelling).
 

TNM

Distinguished
Jul 15, 2007
11
0
18,510
Many factors to consider here for sure.

Working for a large internet firm which have hundreds of servers with SSDs, I know for sure the more reliable flash memory storage products are way more reliable than HDDs. There is no competition.

We have services where we have gone from 6 15k rpm drives/server to 2 flash storage unit per server and we have gone from 1 HDD failure per week (hundreds of servers...) to a couple of flash units failing in a year. No comparison at all, although some of the difference may be from reduced number of units.

For consumers, I think the big advantage would be on laptops. No doubt SSDs will be much more robust there than 2.5 inch drives whenever the laptop is moved. With an ssd, I "throw" my laptop around all the time without suspending it. With HDD inside, my stomach is not happy whenever I have to move faster than windows can suspend.

Finally, I do wonder what the general failure rates are on reasonably complex electronics?
I remember a batch of seagate drives some years back which all failed due to some bad capacitor on the controller.

If the expected failure rate for electronics in general is 1.5%, then I expect that to be the case both for HDDs and SSDs and the difference would really be in the remainder part above that 1.5%.
 

boletus

Distinguished
Mar 19, 2010
69
0
18,630
[citation][nom]bobwya[/nom]Great article.Shame that THG keeps wiping my carefully crafted and thought-provoking comments every time I login... :)[/citation]

I (try to) always copy my longer comments to the clipboard before hitting "submit", here or anywhere else. I've had a few too many of those.
 

tomfreak

Distinguished
May 18, 2011
1,334
0
19,280
Try use the same amount of money and buy HDD for the same GBs. a unit 500GB SSD is enough to buy 40 unit 500GB HDD. I will mirror all 40unit together and I only use 1 active in my PC and place the other 39 offline as my backup drive, lets see if ur SSD can be more reliable than my 39+1 HDDs. if u take away the speed advantages, only focus on reliability per price. SSD will not come close at all.
 

bobwya

Guest
May 21, 2005
692
0
18,980
[citation][nom]tomfreak[/nom]Try use the same amount of money and buy HDD for the same GBs. a unit 500GB SSD is enough to buy 40 unit 500GB HDD. I will mirror all 40unit together and I only use 1 active in my PC and place the other 39 offline as my backup drive, lets see if ur SSD can be more reliable than my 39+1 HDDs. if u take away the speed advantages, only focus on reliability per price. SSD will not come close at all.[/citation]

Interesting arguments there!
1) You will have a slow system. Low IOPs (Server) or head seaking (perceived as stuttering on Desktop systems). So is your comment relevant? It takes 6x 15K RPM drives to match a single SSD. I have seen this my desktop (where 2x 15K RPM drives cannot match the performance of a single SSD).

2) HDDs are vulnerable to power-cycling (every week/month you fire up each of those 39 HDDs to do drive backups - right??!!)

3) What happens when you are hit by a tornado, tsunami, earthquake? :)

Seriously the article already stated that random failure rates for SSDs may track those for HDDs!! The jury is out is all - due to lack of a enterprise adoption over a long enough timescale...

Also one of the things that motivated me (originally) to get an Intel X25 G2 SSD was a ZDNET video of Rupert Goodwin's irradiating it with an Alpha paticle source - and the transfer rate being unaffected plus a lack of CRC errors!!
 

ntrceptr

Distinguished
Jan 26, 2006
97
0
18,630
Well until I see tom's or some other site do some endurance testing like metioned above I will not get an SSD for my OS
My opinion.
 
D

Deleted member 270419

Guest
Articles like this make me wish I was born today so by the time I build my first computer 20 years from now, I won't have to wonder about these things. Of course, by then we'd probably all be on cloud and just have a screen that we talk to but still . .
 

OWC Grant

Distinguished
May 3, 2011
2
0
18,520
[citation][nom]cangelini[/nom]To the contrary! We noticed that readers were looking to see OWC's drives in our round-ups. I made sure they were invited to our most recent 120 GB SF-2200-based story, and they chose not to participate (this after their rep jumped on the public forums to ask why OWC wasn't being covered; go figure).They will continue to receive invites for our stories, and hopefully we can do more with OWC in the future!Best,Chris Angelini[/citation]

Sorry we couldn't be in this article...perhaps more explanation required to explain Chris's statement.

As I've discussed with Chris offline, we were in final stages of decision on product evolution and knowing that reviews take some time...and in this case, this review took 2 months to actually publish...we didn't want dated testing information on our drive to be indicative of what the product would soon (and now currently does) truly offer in terms of performance.

I'm sure OWC drives will be featured again on Tom's Hardware. Heck, going back to near same time last year, we got an award-winning one from the site: http://www.tomshardware.com/reviews/windows-7-ssd-trim,2705-11.html

In meantime, you can visit our site, our blog, and other sources for the latest info and test results on our SSD line...the first to offer a Money Back Guarantee and five year warranty (for SandForce based SSDs).

Thanks!

OWC Grant
 
G

Guest

Guest
This was a great article to read since I'm still looking and waiting to buy my first SSD. I have been thinking RAID 1 is the way to go, but its quite expensive. This article sheds some light on the matter and gives me another technical point to considers.

I just wanted to point out this english error on the last page of the article:

"For example, we was extremely impressed by Intel's reliability presentation at IDF 2011."

Probably should correct to:

"For example, we WERE extremely impressed by Intel's reliability presentation at IDF 2011."

On another note, I work for a top Data Warehousing database company that is selling a hardware configurations that uses SSD combined with 15k RPM SAS for Hot and Cold data storage. I'll have to poke around and see if our hardware guys have any numbers on ARR. We partner with a number of drive vendors and look at reliability when we make hardware choices in the final products.
 

tomfreak

Distinguished
May 18, 2011
1,334
0
19,280
bobwya said:
Interesting arguments there!
1) You will have a slow system. Low IOPs (Server) or head seaking (perceived as stuttering on Desktop systems). So is your comment relevant? It takes 6x 15K RPM drives to match a single SSD. I have seen this my desktop (where 2x 15K RPM drives cannot match the performance of a single SSD).

2) HDDs are vulnerable to power-cycling (every week/month you fire up each of those 39 HDDs to do drive backups - right??!!)

3) What happens when you are hit by a tornado, tsunami, earthquake? :)

Seriously the article already stated that random failure rates for SSDs may track those for HDDs!! The jury is out is all - due to lack of a enterprise adoption over a long enough timescale...

Also one of the things that motivated me (originally) to get an Intel X25 G2 SSD was a ZDNET video of Rupert Goodwin's irradiating it with an Alpha paticle source - and the transfer rate being unaffected plus a lack of CRC errors!!
1) Here the thing, I already said thats only valid if u do not take performance into account. We are talking about reliability here dont get move into performance.

2)You dont have to take 39 HDD and do backup all at once. Say a month for each 36 HDD and put 3 other Raid with first one? Thats puts u cycle only once per 3 year.

3) All natural disaster will wipe everything, dont start telling me ur SSD is waterpoof can survive the Tsunami.

if u solely focus on only reliability on 1 SSD vs 40 HDDs. I personally take the latter.
 
Status
Not open for further replies.