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

Page 2 - 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.

acku

Distinguished
Sep 6, 2010
559
0
18,980
[citation][nom]dereksmall[/nom]I've been considering switching to an SSD for a very specific reason that I would have like to have seen covered. G-shock tolerance, and long-term vibration resistence. I ride a motorcycle to work on a regular basis for the past three years. Since doing so I have found that I'm only getting about 8-12 months out of the conventional HDD in my notebook, before the vibrations cause the drive to start failing. (Yes the heads are parked when it's power off). I have been considering switching to an SSD to reduce failures due to the constant vibrations when my notebook is in a saddle bag.Also I have lost several notebook hard drives over the years when it is dropped even from a couple inches while it's powered up and running (causing a head crash). Not the entire HDD isn't usually trashed, but it usually creates enough bad sectors that I have to reformat the drive.For these two reasons it seems that SSDs would offer more security in mobile devices like notebooks. For my desktop I'll stick with conventional drives (for now at least).[/citation]

As far as shock resistance, we don't even need a long term study for that. Flat out, everyone is going to admit that's an inherent benefit of SSDs. If you look at aircraft (military and civilian), data storage goes to solid state now because of its ability to take on environmental elements easily.

We're strictly looking at the micro-level and trying to answer the question "is the solid state inherently more reliable than hard drives when it comes to retaining data."

And on that note, there's an interesting poll on our Facebook page that suggests that durability/shock resistance isn't even a big reason people use SSDs in their notebook. Completely unexpected.

Cheers,
Andrew Ku
TomsHardware.com
 

spookyman

Distinguished
Jun 20, 2011
670
0
19,010
Actually SSD's are still in there infancy. I remember when FM and MFM drives came out in the 80's. Talk about high failure rate. You actually had to use a mapping program to mark bad sectors on a hard drive. But it time hard drives improved in quality and reliability. Average user can get 3-5 years out of a HDD. In time SSD's will be just as reliable if not more. You do not have the mechanical factor to worry about. That is a big plus.
 
G

Guest

Guest
We believe SSD to not be better. We installed them, and are affected by numerous problems, firmware issues, and failures. A quick note about Intel drives, we purchased SSDSA2MH160G1GN 2.5 inch 160GB SSD models, and with a two year period every single one has failed. We outfitted a US office and some UK staff and not a single drive is left in a working state.
The failure is signified by a slowing of the drives, then more and more stuttering, and then dead drive. These were installed in Dell laptops not doing significant workstation loads, but rather lighter weight normal use.
The worst part is that Intel were exceptionally poor about warranty support on these drives, so poor that we've dictated never ever to buy them again. We can tolerate hardware that fails, we don't have unreal expectations, but not supporting your own products is unacceptable.

I can't support intel drives on either performance, reliability, nor even warranty. But on the whole, we also use crucial, OCZ and other makes, and have consistent issues and failures. Our feel from our side is that you are correct. SSD falls short of its promise on reliability, but not so much on performance.

We still have questions about hard rebbot behaviour, and power cut recovery on SSDs compared to standard drives. Our gut feeling is that the drives are more open to failure after these events.
 
G

Guest

Guest
The paragraph on the last page below the graph, and the graph itself, are so hilariously contradictory as to call the credibility of the whole article into question.
 

acku

Distinguished
Sep 6, 2010
559
0
18,980
[citation][nom]mooo mooo[/nom]The paragraph on the last page below the graph, and the graph itself, are so hilariously contradictory as to call the credibility of the whole article into question.[/citation]

I'm not sure what part of it is contradictory. The 3% that I referred to is an AFR rate. The chart calculates cumulative % based on afr, which is basically compound interest. Think about 100 drives. 1% failing every year. The second year isn't 102%. That's not how the math works.

Plus, this is cumulative % failing, which is why its more like a compound decay, because you have to take away from the original "subset" of numbers. That's why the 1% AFR continues to climb, but it does so as a constant curve. There is no reason to suspect SSDs to model a constant failure rate, but I provided the dotted line projections to show how the numbers relate to other studies.


Cheers,
Andrew Ku
TomsHardware.com
 

slicedtoad

Distinguished
Mar 29, 2011
1,034
0
19,360
[citation][nom]acku[/nom]All of the data is so fragmented... I doubt that would help. You still need to take a fine toothcomb to figure out how the numbers were calculated.[/citation]
Why would it be fragmented? The companies almost certainly look at their own numbers, and they would have reports and summaries as well.
 

acku

Distinguished
Sep 6, 2010
559
0
18,980
[citation][nom]slicedtoad[/nom]Why would it be fragmented? The companies almost certainly look at their own numbers, and they would have reports and summaries as well.[/citation]

There are a few relatively large populations of SSDs that we couldn't publish information on due to confidentiality. For those data sets, the reports were chaotic to sort through - unvalidated, validated, some data wasn't normalized due to rolling deployments. Then even the failure modes are an issue. Some manufacturers separate hard (like an IC going dead) and soft errors (firmware basically), and others combine them without and footnotes. And some companies ignore specific types of soft errors. Basically, "it's all about the numbers."

We define a "failure," as anytime the SSD has to be replaced. I really don't care what the manufacturer claims the cause is. If you can't get access or write data, that's a dud.

Cheers,
Andrew Ku
TomsHardware.com
 

jimr1234567890

Distinguished
Jul 29, 2011
1
0
18,510
Excellent work. Love the reference material, giving me lots to read.

I would have loved to seen cost benefit analysis of the SSD vs HHD. It is worth while for the 'mom & dad' user pay the extra bucks for a SSD, it is worth while for the hard core gamers, it is worth it over 5-10 years in a small 4 disk home RAID (less than 10TB)? It is nice to see the larger averages from data centers but how does this translate into the various home user levels?

Again great work, thanks for all the effort you put into this article.




 
G

Guest

Guest
Not sure what warsp1te69 is doing to their Intel SSDs, but I have 3 of them, in 3 different machines, and they all work great. One is a 80GB 1st gen "M" (My primary desktop), a 160GB 1st gen "M" (A public facing web/ftp/email server that gets hit HARD), and a 120GB latest gen "V" (My son's primary desktop). I accidentally blew out my original 80GB 1st gen by overvolting the memory/ICH too high, and while I found it hard to get ahold of Intel's SSD department, once I did, service was absolutely excellent -- cross shipped drive, and I had a replacement in my hands in under 24 hours (16 I believe) from when I talked to them. Couldn't ask for more, other than a better way to actually get through their customer support.

I also have an OCZ Agility 2 160GB drive, and it's still working great, but it's in my secondary machine, so it doesn't see all that much use.

I am definitely looking forward to Intel's next big SSD offering, and I'll NEVER own another PC without a SSD in it. It doesn't have to be only SSD -- not a single machine is SSD-only, but all boot from it, page swap to it, hibernate to it, etc.
 

bounty

Distinguished
Mar 23, 2006
389
0
18,780
Just a couple of thoughts.....

"Higher Reliability Through Fewer Devices"

This only benefits SSD's if you need high IO. If you need lots of storage it hurts SSD's. Think TB/s of backups for enterprise or your movie collection for end users. 1 HDD v.s. 3-4 SSD's.

Also, we need to come back to this in another couple of years, since we need to evaluate how well wear leveling works long term. We may see a flood of SSD's die on year 6 etc. You kind of addressed this a bit at the beginning of the article, but it's just theory and not real world. We don't know how well existing firmware algorithms work long term. It may also make a difference in enterprise (constantly working drives, lots of GB written) v.s. consumer.
 

flong

Distinguished
Dec 27, 2010
1,106
0
19,310
[citation][nom]acku[/nom]Actually that SSD 320 problem would have counted as a failure. When you can't accesses data, that's a big no no.Thanks for the kudos. But a few corrections. There is no data to suggest that hdd failures are greater than ssds. The projections in the graph assume a constant failure rate, which never occurs. I just put it in so that people could see how it relates to a AFR of 1%. For the moment, it's unclear if SSDs are more reliable. The initial 2 year data suggests otherwise.[/citation]

Thanks for the clarification. The graph lead me to believe that HDDs have a higher failure rate as they age, but you are correct in pointing out that may not be the case. Logically, because HDDs are mechanical, you would think they would tend to break down more over time, but that may not be true. The HDD in the computer I am using now is 8 years old and still functions brilliantly.
 

jeremiahpope

Distinguished
Feb 10, 2010
13
0
18,510
I have a seagate 320gb that has been my boot drive since 2006. I upgrade parts at a time and its the oldest part of the machine. It has had multiple mobo's, cpu's, video cards, memory, psu's and too many to count installations of xp, vista and 7. Still going strong, but I will say that I back everything up thats on it more often now. I would call it a pretty decent investment.
 

triclops41

Distinguished
Dec 6, 2009
12
0
18,510
Thanks so much for this article. This is a topic that is wholly ignored, and articles like this are needed to bring the truth out. We just accept that SSDs are more reliable. That may or may not be true, but this is the kind of stuff that will answer that question.
 

acku

Distinguished
Sep 6, 2010
559
0
18,980
[citation][nom]triclops41[/nom]Thanks so much for this article. This is a topic that is wholly ignored, and articles like this are needed to bring the truth out. We just accept that SSDs are more reliable. That may or may not be true, but this is the kind of stuff that will answer that question.[/citation]

I'm glad you enjoyed it. :)

Cheers,
Andrew Ku
TomsHardware.com
 

compton

Distinguished
Aug 30, 2010
197
0
18,680
As time goes on, I'm positive that SSDs will get the same kind of study that HDDs have over the years. It's too bad that the different controllers/nand in SSDs (something for I'm grateful -- variety) end up obscuring SSD reliability perception (is it a faulty controller, or nand?). Sandforce drives clearly have some issues other controllers don't. But the same is also true of Micron's controller, and it found it's way in the Intel 510 drive (which I'm using). Who can say for sure that an Intel drive with a non-Intel controller is more reliable? I know Intel picks the most reliable flash from it's NAND venture with Micron. How much this helps probably is only know to Intel itself. Ultimately some mlc flash is more reliable than others, but good luck figuring that one out. Some controllers are more reliable than others, but it's hard to tell which is which.

I recently dropped my laptop over five feet (while running) onto my hardwood floor. Since it was running an SSD, it was fine. Not even a scratch, but had a HDD been in there, it would have certainly been curtains. SSDs are clearly at home in laptops and mobile devices which would otherwise be an inhospitable environ to spinning media.

As for mlc endurance, I think it's not quite the non-issue it might appear from the first graph (even with the caveats). When measuring daily writes, it doesn't include the drive's background activity, or the fact that some cells will surely not have the same endurance as others. Clearly this still isn't an issue at 25nm, but what about at 20nm and below? You'd expect at this rate of endurance reduction that in a few years a cell might only last
 

boletus

Distinguished
Mar 19, 2010
69
0
18,630
Thanks for the article; it provides at least a part of what I need to know. Specifically, that *Intel* SSDs are *probably* reliable enough for my purposes. As stated in the article, it tells me nothing about any of the other vendors except the one mention of Super Talent, which apparently had dismal reliability for the one info source. Until real, objective numbers are available for the other vendors, I have no use for them. Here's why:

I will take issue with the statement that regular backups alleviate the reliability issues. Even with a full disk image created every day, a day's worth of data/work can be lost forever. Also, do not understate the hassle of getting an RMA from the supplier/vendor, pulling the drive and sending it back, waiting for a replacement, and reinstalling it in the machine.

In a perfect world, that will take about 2 weeks. If I don't have a spare of the same unit on hand, and I need to compute during that interval, I've got to set up another drive. Of course disk images from the SSD may not be compatible with the temporary replacement, and that drive may not be compatible with my SSD once I get it back, so now I need to spend some time doing file transfers and possibly an OS reinstall (with updates, drivers, security software, etc.), keeping my fingers crossed the whole time that nothing else goes wrong.
 

acku

Distinguished
Sep 6, 2010
559
0
18,980
[citation][nom]boletus[/nom]Thanks for the article; it provides at least a part of what I need to know. Specifically, that *Intel* SSDs are *probably* reliable enough for my purposes. As stated in the article, it tells me nothing about any of the other vendors except the one mention of Super Talent, which apparently had dismal reliability for the one info source. Until real, objective numbers are available for the other vendors, I have no use for them. Here's why:I will take issue with the statement that regular backups alleviate the reliability issues. Even with a full disk image created every day, a day's worth of data/work can be lost forever. Also, do not understate the hassle of getting an RMA from the supplier/vendor, pulling the drive and sending it back, waiting for a replacement, and reinstalling it in the machine. In a perfect world, that will take about 2 weeks. If I don't have a spare of the same unit on hand, and I need to compute during that interval, I've got to set up another drive. Of course disk images from the SSD may not be compatible with the temporary replacement, and that drive may not be compatible with my SSD once I get it back, so now I need to spend some time doing file transfers and possibly an OS reinstall (with updates, drivers, security software, etc.), keeping my fingers crossed the whole time that nothing else goes wrong.[/citation]

Well, I'm not going to lay the blame solely at the feet of SSD vendors. It's a storage problem in general. Hard drives have been around for 50 years, but there is no actual study on which brands are the best. Google did the closest thing to a study with consumer grade PATA and SATA drives, but they refused to publish the information (wait... isn't Google's mission is to organize the world's information and make it universally accessible and useful?)

There's more pressure on SSD vendors given their price premium, but I'm not going to hold my breath.

And as we stated in the article. All of our data is centered on unvalidated failures. If you have to replace it, it's a dud. As a customer or IT professional, it doesn't matter what the drive manufacturer says when the drive is diagnosed. If it doesn't work in the field, there's a problem. Simple as that.

Actually, if we limited ourselves to vendor supplied data, often times it only contains "validated" failures, and those figures would be 1/2 or 1/3 what we're reporting here. Any data that we get from the vendors will have to be scrutinized. Page 8 is a good explanation of how the numbers that you get in press documents can overstate reliability.

Cheers,
Andrew Ku
TomsHardware.com
 

boletus

Distinguished
Mar 19, 2010
69
0
18,630
Which is why we, as consumers, must rely on somewhat anecdotal evidence, such as user ratings on tech websites. However, even knowing that those numbers are likely skewed (users are more likely to provide feedback for products that fail, and some don't know how to install it correctly), a decent sample size that shows >20% of owners rate it as poor or awful is plenty enough to scare me away.

 

PreferLinux

Distinguished
Dec 7, 2010
1,023
0
19,460
[citation][nom]spookyman[/nom]Actually SSD's are still in there infancy. I remember when FM and MFM drives came out in the 80's. Talk about high failure rate. You actually had to use a mapping program to mark bad sectors on a hard drive. But it time hard drives improved in quality and reliability. Average user can get 3-5 years out of a HDD. In time SSD's will be just as reliable if not more. You do not have the mechanical factor to worry about. That is a big plus.[/citation]
In a way. They (solid state drives) have been around longer than HDDs. But if you are talking about modern SSDs (which we know you are), then that is a different matter.
 
G

Guest

Guest
I would hate to spend a lot of money on a large SSD only to have it die in three years. Maybe a lot of these speed addicts are in for a surprise?
 

AnUnusedUsername

Distinguished
Sep 14, 2010
235
0
18,710
It's fairly obvious that even the slowest SSD is going to be faster than the best HDD. If that's the case, why do companies seem to focus on trying to make "the fastest" SSD? For anything but enterprise use, how fast your SSD is isn't going to make much difference. How stable and reliable it is is what's going to make a difference, and you'd think companies would focus on that instead of speed. Maybe it's just a matter of companies not knowing how to develop more stable/reliable drives because they don't know why, how often, or for what reasons existing SSDs do have problems that could be overcome by better design. Maybe they could take some lessons from previous uses of solid state memory, like BIOS storage, flash drives, and various other places. It's similar technology, and probably could provide at least some insight into what makes it last and what's likely to make it fail.
 

boletus

Distinguished
Mar 19, 2010
69
0
18,630
[citation][nom]AnUnusedUsername[/nom]Maybe it's just a matter of companies not knowing how to develop more stable/reliable drives because they don't know why, how often, or for what reasons existing SSDs do have problems that could be overcome by better design.[/citation]

I think that's the crux of the matter (for everyone except maybe Intel), but I think (hope) enough market saturation is occurring now so manufacturers are seeing where their problems are. In another year or so, most of the problems should be ironed out (sigh). Until then, it is still a bit of a crap shoot. If the increased performance is enough to make up for the cost of replacements, or if your an enthusiast who has lots of time to play with your PC, SSDs make sense now. Otherwise, Intel seems the only relatively safe choice. I hate to tout one company, but I assume they're smart enough to avoid shooting themselves in the foot with the 5-year warranty (still just an assumption though), and there is less anecdotal evidence (e.g. website user ratings) of wiidespread problems.
 
Status
Not open for further replies.