The new OWC 1TB internal hard drive I installed in my Mac Pro 3,1 just died and cannot be recovered. This was presumably a good-quality brand new hard drive. Fortunately, I have multiple backups thanks to SuperDuper and Time Machine, so I only lost a few emails that arrived after the previous hourly TM backup and before the disk failure. I'm seriously thinking about a solid state drive to replace the dead internal hard drive. The startup volume that was just lost was this size:

Volume size: 601.61 GB (601,611,190,272 Bytes)
Available: 328.28 GB (328,283,320,320 Bytes)
Used : 273.33 GB (273,327,869,952 Bytes)

An unused, second volume occupied the remaining space available on the 1TB hard drive.

Barring some unforeseen development, I should be able to keep the used space below 300 GB, but I need some additional space for Photoshop scratch and OSX's virtual memory as well as ordinary OS operation. The main applications and their attendant documents and images occupy a lot of room, so it will be hard to downsize to squeeze into a smaller-sized SSD. They are Adobe CS4, MS Office, and VMware Fusion 3 with Windows 7 Ultimate. Also, documents and images have historical value and can't be reduced by much.

I'm interested in advice about what's available and practical in the world of solid state drives. An assist from you experts would be very helpful. As you already know, I'm not an IT expert by any measure, but I use a computer every day in my work. I'm best described as an experienced, bruised user who can't program anything now that MSOffice doesn't support macros.

imho SSDs are of very little value for their added cost. Why do you want an SSD? If you don't have a good reason, don't get one.

Shock resistance, read (not write!) speed, and power consumption are the only real advantages.

They have no moving parts. Doesn't that make them inherently more stable than hard drives?

I'm reacting to a mechanical failure that could have been a disaster if I didn't make backups.

Assuming no defect in manufacturing, won't they pretty much last forever and never break down, when operated within electrical tolerances? I.e., the power supply will wear out one day... but the "solid-state" part does not degrade (at least in terms of human lifetimes).

The fly in that ointment may be limited Write Endurance.

Bummer... ah well, thanks (sort of).
Your avatar seems pretty endurable.

They do boot faster too.

This discussion and the prices I encountered have scared me off for now. I've joined the wait and watch club regarding SSDs. Today, I ordered a replacement hard drive, but this time I ordered the version with a 5 year warranty (instead of the 3 year warranty version that failed) that is glowingly described as follows:

1.0TB Seagate Barracuda ES-2 Enterprise-Class SATA II 7200RPM 32MB Buffer Hard Drive W/ Perpendicular Drive Technology. New w/5 Year Seagate Warranty!

Thanks for the help in making a decision. One little twist worth mentioning is that I told the vendor of the failed drive, OWC, that the drive contained confidential and proprietary information of clients that I absolutely must protect. I told them I would personally mutilate the drive and send them the mutilated drive if they wanted it. They told me that there would be no warranty replacement unless I sent in the failed drive unscathed. I told them that my license to practice law would be in jeopardy if I did such a thing and that I would keep the drive and mutilate it. The safety of my clients' data was worth more to me than the price of a new drive. So, no replacement unless you're willing to trust them and Seagate to which they would ship the failed drive. The moral to the story is that there is no such thing as a warranty replacement if you're not willing to trust them with your data on a failed drive. I found that disillusioning, but I do see their point. They want to actually see that it failed and they also want to see why it failed.

In the past, I've always bought the "server quality" version of hard drives rather than the economy model and I've never had a hardware failure of a hard drive. My first venture into the lower-priced version bit me within 2 months of installation. Thanks again for the very helpful discussion.

Ah, a Seagate drive. I wondered about that. Virtual1 has often complained that Seagate has quality control problems. I have bought 6 Hitachi drives (all from OWC) and never had a problem with any of them.

Of course, if you buy a seventh and it crashes immediately, you will be able to be described as having experienced a 14% failure rate!

(I appreciate your confidence in Hitachi drives, but it would be nice to find an apples-to-apples comparison; your six drives versus Virtual1's extensive but non-quantified experience doesn't constitute a very scientific basis for making a buying recommendation.)

Yep, sadly, all Flash memory on the market has a finite lifespan. Each time a memory cell is written to, damage to the cell occurs; after a finite number of write cycles, that cell is destroyed. This affects all types of NAND Flash devices--camera cards, USB keychain drives, SSD cards, iPhones, you name it. In my opinion, that's a huge problem with Flash memory.

Modern SSD drives try to mitigate this by using smart controllers that spread out the data you write so that no single cell gets written to more than the others. So for example if the computer writes data to the same "track" of an SSD over and over, it will be recorded to different cells each time. That evens the wear and makes sure the drive doesn't fail prematurely become files (such as the disk directory and VM files) are written to more than others, but there's still a limited lifespan.

Using something like Photoshop on an SSD is probably something of a worst-case scenario, because SSD devices have slow write speeds and Photoshop does a LOT of writing.

Scientific, no; anecdotal, yes. I have read favorable comments about Hitachi drives in other forums so I'm not the only one who likes them. Don't you think that the experience of users is valuable, even if the information is anecdotal? When I contemplate purchasing something, the opinions of people who have actually used the product are most useful to me.

The experience of userS is very valuable. However, I've never before heard you mention "favorable comments about Hitachi drives in other forums;" that expansion of your very narrow experience to a more generalized anecdotal impression actually makes your perspective a lot more useful, in my opinion.

There aren't a lot of posts in which you criticize products you've used. DiskWarrior is "worth its weight in gold" (which, if you think about it, is actually a severe condemnation, since the weight of 5 MB of digital data is infinitesimal). Each new OS version, each piece of software performs "flawlessly" (in retrosepcct, anyhow). Longtime readers of MFIF/FTM likely understand that your frequent glowing recommendations of a small group of products and vendors you use is an expression of enthusiasm for things whose reliability you take pleasure in.

But I wonder if someone new to our troubleshooting community wouldn't be better served by recommendations which temper the extreme narrowness of individual experience with a little research into broader data—whether anecdotal or statistical?

Will an iPod weigh more when it's full of music?

Understanding that each SSD has its own"Write Endurance," have you run across any stats on expected SSD life such as has a computer ever outlived the "Write Endurance" of its SSD?

I suppose one could muse whether a 1 is heavier than a zero or not...?

Considering the speed of HD platters it could throw it off balance!

I'm sure there are stats out there, but it's hard to say whether or not the SSD will live as long as the computer given that different patterns of compute usage will have radically different effects on the SSD life.

For example, I am a very, very heavy Photoshop user. Photoshop does a huge amount of writing of very large scratch files, so given that I would expect to wear out SSD drives in a pretty short order--I would not at all be surprised if my pattern of usage would destroy an SSD in a couple of years. Users who don't use their computers the way I do might have their SSDs last a lot longer, and someone who used an SSD only for periodic backups might expect the drive to outlast the computer.

The type of SSD flash memory also has an effect. Some of the older style NAND Flash chips had a VERY short write endurance--as little as 10,000 writes would destroy the cell. Today, the most expensive high-end, high-performance chips have over a hundred times that much life, and enterprise-level SSD drives have nearly a hundred times that life, but the consumer-grade SSDs are still less durable than the highest-end chips.

Manufacturers usually don't give reliability figures for their flash drives directly; that information is considered proprietary. And they don't talk about the ways they compensate for Flash cell failure in detail; that's also considered proprietary. But I've read some papers that claim that the most expensive, high-end, server-level SSDs, which contain controllers that spread out writes and which contain spare cells to compensate transparently when some cells begin to fail, can have a real-world life expectancy of more than a decade, which is better than conventional hard drives and will certainly last longer than most computers do. Some SSDs, which contain multiple Flash storage elements in RAID-like arrays, have a claimed real-world life of more than 50 years, which is way better than you could ever expect from a hard drive.

These particular types of SSDs are still fairly exotic and hellishly expensive, but technology being what it is, I would expect to see the same performance from consumer devices in the not too distant future.

As is always the case, anything you read in Wikipedia requires further research, but this text about wear leveling and security is interesting:

http://en.wikipedia.org/wiki/Solid-state_drive

Thanks tacit and MMT3 for the added perspective.

Given that our LCD screens are also "solid-state", how come they don't wear out more easily. [actually, i still find it surprising to learn that SSD drives don't last a lot longer.]

The differing technologies in displays (LCD, LED, OLED, plasma, CRT) and in hard drive (SSD/MO/M) are so strikingly different in construction and behavior that they're impossible to objectively compare by construction differences. You have to examine each one's strengths and weaknesses independently and then compare them based on their final performance.

Hard drive wear is a topic about on even keel with comparing burn-in issues with various displays.

I've ran into countless mechanical drives with bad blocks, but have yet to play with an SSD or even a flash drive or flash memory of any kind that has had a bad cell/block... anyone have experience with that?

The specifics of the problem with Flash memory has to do with the type of electronic circuit in it.

Flash memory uses a basic storage element called a "floating gate transistor." A floating gate transistor can basically be thought of as an on/off switch; leaving technical details aside, the transistor either conducts electricity (on) or it doesn't (off), and it will stay in whatever state you put it in even when there is no power going to it.

It's called a "floating gate" transistor because each switch has a tiny bit of conductive material, a few microns wide, that is separated from the rest of the transistor by an insulator. The little sliver of conducting material can be charged full of electrons, and when it is charged, it holds the charge for a very long time, because it's surrounded by insulator on all sides. When it's charged, the rest of the transistor "senses" the charge and it acts like a closed switch. When it isn't charged, it acts like an open switch.

Make sense so far? Here's where the funky bit comes in:

You charge or discharge the gate by pushing electrons into it or draining electrons out of it. Now, the gate is surrounded by insulating material, in a very thin layer, so you would normally think there is no way to get electrons in or out, since the electrons can't pass through the insulator.

But electrons have the ability to teleport, jumping from one place to another without passing through the space in between. They can only do this over very tiny distances; the phenomenon is called "electron tunneling," and it happens because electrons are not hard round little balls like billiard balls. They are quantum particles, and as such they do not have definite, precise position in space--they have a wavelength, and they can appear to act as through they are located anywhere within that wavelength. (That's the quick and dirty overview, anyway; the reality is a bit more complicated, but that is the basic gist.)

Because electrons can tunnel, they can be made to pass through the insulator and into or out of the gate without actually moving through the space in between. You can force electrons to tunnel into the gate or tunnel out of the gate even though it's surrounded by insulating material, because the width of the insulating material is smaller than the wavelength of an electron.

When you do this, though, you end up damaging the insulator. You have to provide a pretty good push on the electron in order to get it to tunnel--it takes more power to write to a Flash cell than it takes to read one--and applying a higher voltage field across the insulator causes it to degrade.

Eventually, the insulating layer degrades so much that it no longer acts as an insulator. That particular cell shorts out; the floating gate is no longer floating and no longer stores a charge.

Things like LCDs aren't made from floating gate transistors. They're made from thin-film field effect transistors, which are entirely different in their structure and in how they work. Field effect transistors don't have a charged floating gate, which means they don't retain their state when you remove power to them but also means that they aren't damaged by the process of changing states.

(thank you for the explanation of how flash memory works)

In a parallel universe, Tacit writes for Scientific Parallel Universarian.

Thanks for that.

Here are two useful links i discovered while googling items in your explanation. I post them here for convenience to both others and myself (so i'll know where to find them in the future, if need be):

1. Semiconductors Without the Quantum Physics (or not so much that you would notice)
   
   
2. Solid State Drives Data Reliability and Lifetime (PDF whitepaper)

--

It seems that the problem "insulating" material is SiO2 (silicon dioxide, aka silica), a special glass of some sort. So basically, the semiconductor components (FETs) remain more or less intact... but the loss of insulation (due to *writing* any data which toggles zeros and ones) renders the overall device increasingly useless for non-volatile storage. What a shame.

Fascinating, and even more fascinating after thinking about it for a bit.

Thanks.

In this context the following announcement may be of interest: DriveSavers' iPhone app to simulate SSD failures.


UPDATE: AFAICT, the SSD update only adds a simple animation with a single popup text box. Not all that useful beyond suggesting things can go poof with SSDs too...

I downloaded that and was NOT impressed. Most of them sounded the same, and it contained a very small portion of all the sounds I've heard. And it's really gimmicky to tell you the exact model of drive and what's wrong with it. Do we really care if it's a head crash or a bad spindle motor? Either way it don't work and we can't do anything about it short of mail it to them.

fwiw, my collection has been slightly updated, and features not only bad hard drives, but good ones too, so you can compare. I tried to weed out the ones that were too similar sounding. Some of them are doing some really wild things. (some failures are a LOT more common than others, hence trying to dedup)

I have to agree on both counts. That said, do you think the app is even halfway useful to the 'naive' user confronted with a noisy and malfunctioning HD? This type of user isn't likely to have immediate access to your collection of drive noises, and probably doesn't even know something like it exists.

it's of no practical value. it's an advertisement is all. I was amazed to not even hear the enormously common identical taktaktak taktaktak taktaktak of the bazillion of seagate laptop drive deaths as of recently. That alone probably accounts for 1/4 of all recent failures I've ran into.

I had almost 30 of them available for sample.