Beta
×

Welcome to the Slashdot Beta site -- learn more here. Use the link in the footer or click here to return to the Classic version of Slashdot.

Thank you!

Before you choose to head back to the Classic look of the site, we'd appreciate it if you share your thoughts on the Beta; your feedback is what drives our ongoing development.

Beta is different and we value you taking the time to try it out. Please take a look at the changes we've made in Beta and  learn more about it. Thanks for reading, and for making the site better!

Samsung Launches SSD 830 Drive

samzenpus posted about 3 years ago | from the new-toys dept.

Data Storage 148

MojoKid writes "Although they haven't been big hits with enthusiasts, Samsung's solid state drives have been successful due to strong relationships with a number of OEMs, including Apple. With the release of their new SSD 830 Series Solid State Drives, however, Samsung appears ready to make inroads with enthusiasts as well. The SSD 830 tested here is 256GB model, with eight 32GB Samsung NAND flash memory chips, 256MB of Samsung DDR2 SDRAM cache memory, and a new Samsung SSD Controller. The Samsung controller features a 3-ARM core design with support for SATA III 6Gb/s interface speeds. Performance-wise, the Samsung SSD 830 Series drive offered the best Read performance of the group that was tested, even versus the latest SandForce-based SSDs, though the SSD 830 couldn't quite catch SandForce in writes."

Sorry! There are no comments related to the filter you selected.

Golden Girls! (-1)

Anonymous Coward | about 3 years ago | (#37508770)

Thank you for being a friend
Traveled down the road and back again
Your heart is true, you're a pal and a cosmonaut.

And if you threw a party
Invited everyone you ever knew
You would see the biggest gift would be from me
And the card attached would say, thank you for being a friend.

why haven't they "been a big hit with enthusiasts" (1)

Anonymous Coward | about 3 years ago | (#37508772)

I assume if anywhere there are "enthusiasts" here on Slashdot, so per the summary, why haven't Samsung's solid state drives "been big hits with enthusaists"? Whose drives have?

I'm thinking of sprucing up an old laptop with an SSD - any recommendations?

Re:why haven't they "been a big hit with enthusias (5, Informative)

fuzzyfuzzyfungus (1223518) | about 3 years ago | (#37508930)

Historically, Samsung's offerings have been relatively solid; but quite unexciting in performance terms, and pretty tepid in performance/dollar.

OEMs love 'em because, while mediocre, they have been comparatively reliable(no equivalents of the Jmicron controller debacle, firmware that makes them show up as only 8MB in size, assorted bleeding-edge weirdness and general "No, we really do have to offer these things under a 3-year warranty to get business customers"-stopping issues.)

The enthusiast-darling crown has changed hands a number of times. Intel was the one to have a little while ago, I think that they've been eclipsed by some of the newer Sandforce gear of late. There are rather more brands than there are chipsets, so brand enthusiasm tends to swing wildly based on cost and who is releasing the new hotness chipset this month.

I have a couple of 470s, partly chosen on price. (1)

Robert Frazier (17363) | about 3 years ago | (#37509112)

When I recently built a couple of boxes (one for me, one for my wife), I put in each a 64GB Samsung 470. In this neck of the woods, they were the amongst the first to break the 1GBP per gigabyte price (on sale). So, price was an important consideration. Admittedly, I was also terribly keen to get relatively solid, but quite unexciting performance.

(The only other SSD I have is an Intel.)

Best wishes,
Bob

Re:why haven't they "been a big hit with enthusias (5, Interesting)

JonySuede (1908576) | about 3 years ago | (#37508934)

Avoid SuperTalent like the plague they are.
Avoid anything used or refurbished.
Avoid any hybrid solution as they drain more battery.
If you don't need a lot space and need extreme reliability look at intel 311 series (those drives kick ass) or any SLC based SSD for that matter.
If you don't need extreme reliability, but don't want to play a game of Russian roulette with 3 bullets instead of one (like in the case of a SuperTalent drive), look and anything sand-force based.

Since you have an aging laptop, you do not need something that can saturate sata 6Gb/s so try to find something like an OCZ Vertex 2 1 drive or a Corsair Force 1 as in real life they are quite similar (you do not need the third edition (both drives have a V3) in an aging laptop).

Also bench the writing speed only one or two time as the more you bench the slower your drive get, you can usually bring some of it back by emptying the drive by formating it to ntfs in windows 7 and the use a force trim utility, wait about 15 minutes. After that you can reformat your drive to your file system of choice and the performance should be OK

1- In synthetic benchmark they differs a little bit but it is imperceptible in real life, unless your main workload is approximated correctly by the synthetic benchmark you were looking at.

Re:why haven't they "been a big hit with enthusias (0)

JonySuede (1908576) | about 3 years ago | (#37508956)

yay, the sup html tag is to hot for Slashdot a site for supposed nerd.....

Re:why haven't they "been a big hit with enthusias (4, Funny)

ArsenneLupin (766289) | about 3 years ago | (#37508984)

yay, the sup html tag is to hot for Slashdot a site for supposed nerd.....

And apparently, you are too sexy for preview...

Re:why haven't they "been a big hit with enthusias (1)

JonySuede (1908576) | about 3 years ago | (#37509460)

Hell, yeah, with my tone biceps and defined abs and my "I work on my backyard, with my torso naked, each weekend" tan , I am definitively to hot for that ;)

But, on a more serious note, I gave up on the preview system yesterday because I corrected every nonsensical phrases*1 in my post about Scala only to notice that the message slash posted was the one before all the corrections. It might not be a slah bug, it might be a Firefox 9 bug (it is a nightly build after all) but it made me gave up on it nonetheless.

  1- I do not know why but when I turned 30 some years ago, coffee started to hit me more than any hard stimulant that I ever tried in college; now when I drink coffee, I write like a methhead

Re:why haven't they "been a big hit with enthusias (1)

ArsenneLupin (766289) | about 3 years ago | (#37509684)

Hell, yeah, with my tone biceps and defined abs and my "I work on my backyard, with my torso naked, each weekend" tan , I am definitively to hot for that ;)

Indeed, sounds hot. Any chest hair with that? Photo?

Firefox 9 bug

Oops, here I go off the internet for just five minutes, and when I come back, Firefox has jumped 3 major versions...

I do not know why but when I turned 30 some years ago, coffee started to hit me more than any hard stimulant that I ever tried in college; now when I drink coffee, I write like a methhead

Lucky you. For me, coffee seems to have less and less effect the older I get. Fortunately, there's Red Bull, which still seems to keep some of its power...

Re:why haven't they "been a big hit with enthusias (1)

JonySuede (1908576) | about 3 years ago | (#37509808)

Any chest hair with that?

Hell, Yeah, just the right amount and I wear an hybrid between the Disney villain mustache and the smug bastard two month later mustache as depicted on that page : http://www.fearlessrp.net/showthread.php?tid=120 [fearlessrp.net] .

Photo?

Hell no, my wife would kill me, in conjugal life, you have to choose your battles wisely!

What will happen when they die? (2)

Jane Q. Public (1010737) | about 3 years ago | (#37508806)

Does anybody have a backup plan for when their SSDs die? After all, unlike magnetic media, SSDs have a limited number of writes. AFAIK, none of them are rated yet for over a million writes, so they are bound to fail at some point.

When SSDs were newer, I argued here on /. (against vociferous claims to the contrary) that I could write a program that would break an SSD quickly. The wear-leveling is better today, but since then such applications have actually been written and tested, and they work.

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37508828)

And that has been proven and confirmed by the links you provided ?

Re:What will happen when they die? (4, Interesting)

Microlith (54737) | about 3 years ago | (#37508834)

AFAIK, none of them are rated yet for over a million writes, so they are bound to fail at some point.

That rating, mind you, is per cell. Virtually all SSDs do some form of wear leveling and are over-provisioned to ensure that no one erase block gets worn out early. And the "backup plan" is pretty much the same as for a regular hard drive: duplicates on RAID for reliability and backups for failure recovery.

I could write a program that would break an SSD quickly

Sure, you can deliberately and forcefully break an SSD. But the amount of IO required to do so tends to go above and beyond what even the average enthusiast will do. And if your typical IO pattern is one that will break an SSD, then you should plan for it and determine if the speedup is worth the cost.

Re:What will happen when they die? (1, Informative)

0123456 (636235) | about 3 years ago | (#37508858)

And the "backup plan" is pretty much the same as for a regular hard drive: duplicates on RAID for reliability and backups for failure recovery.

Mirroring an SSD to another SSD which is likely to fail at almost the same time doesn't seem a great plan to me :).

Re:What will happen when they die? (1)

Microlith (54737) | about 3 years ago | (#37508896)

SSDs aren't exactly inexpensive, are they?

Perhaps I wasn't clear: you'd keep copies on a RAID made of regular disks.

Re:What will happen when they die? (1)

0123456 (636235) | about 3 years ago | (#37508916)

SSDs aren't exactly inexpensive, are they?

Byte-by-byte they're about the same price as the 15k SAS drives we use in the RAID on the servers I maintain; and a lot cheaper than those drives were when first installed a few years ago.

Re:What will happen when they die? (2)

Microlith (54737) | about 3 years ago | (#37508942)

Depends on your perspective then, I suppose. Inexpensive for a server isn't exactly inexpensive for the average home user.

Re:What will happen when they die? (1)

Bengie (1121981) | about 3 years ago | (#37509696)

"Byte-by-byte they're about the same price as the 15k SAS"

If that's the case, what about power costs and gained IO performance?

I'm curious from someone who works with them in servers.

Re:What will happen when they die? (1)

Jeremi (14640) | about 3 years ago | (#37509770)

Mirroring an SSD to another SSD which is likely to fail at almost the same time doesn't seem a great plan to me :).

Assuming SSDs are likely to fail after a certain number of writes (as opposed to after a certain number of hours of uptime), a Time-Machine style backup system would work fine (since the number of writes to the backup drive would be much less than the number of writes to the primary). RAID-0 probably would be a bad idea, for the reason you mentioned.

Re:What will happen when they die? (0)

Jane Q. Public (1010737) | about 3 years ago | (#37508870)

"That rating, mind you, is per cell. Virtually all SSDs do some form of wear leveling and are over-provisioned to ensure that no one erase block gets worn out early."

I am aware of how they are constructed and how they work, thank you very much. None of that changes the essential point: they can and do wear out, and probably cannot be expected to last as long as a modern hard drive, depending of course on usage.

The amount of I/O required to break an SSD (if one is doing it deliberately) is nowhere near as much as you seem to think. One only has to do it intelligently.

I can envision a simple virus that could break SSDs willy-nilly, although its operation would be transparent to anyone who knew what to look for.

Re:What will happen when they die? (2)

DamonHD (794830) | about 3 years ago | (#37508910)

Running a busy USENET server (I think I hovered at ~#10 in the stats for while) used to wear out normal hard drives too, back in the day; SSDs aren't especially novel in that regard IMHO. It's really only a matter of how frequent and comprehensive your backups are.

And as my USENET data didn't last longer than about a week then I think I regarded backups as largely pointless except for some very low-traffic local groups and just threw away a drive when it died and let the new one fill up again!

BTW, I've been running a server entirely on a mixture of SD cards and USB Flash for a couple of years so far, so good. I have taken efforts to reduce spurious writes but I still make sure that I have backups of critical stuff elsewhere: http://www.earth.org.uk/note-on-SheevaPlug-setup.html [earth.org.uk]

Rgds

Damon

Re:What will happen when they die? (1)

Jane Q. Public (1010737) | about 3 years ago | (#37508994)

"BTW, I've been running a server entirely on a mixture of SD cards and USB Flash for a couple of years so far, so good."

Unless you have some kind of RAID-style paralleling arrangement, that has to be slow as molasses.

Re:What will happen when they die? (2)

DamonHD (794830) | about 3 years ago | (#37509032)

It's entirely fast enough for my purposes: I have front-end mirrors and CDN to serve data quickly to end users. It also handles my mail (including many thousands of SPAMs per day), and SVN and so on.

And it does mean that I've been able to run the entire system off-grid, on a few solar panels propped up against a wall! B^>

Rgds

Damon

Re:What will happen when they die? (1)

wagnerrp (1305589) | about 3 years ago | (#37509748)

And here I just thought it would be slow as molasses because it's running on a "plug computer" with a marginal amount of memory. Just because ones duty cycle is only 5% doesn't mean you actually want to get by with only 5% the computer.

Re:What will happen when they die? (1)

DamonHD (794830) | about 3 years ago | (#37509818)

It works for me. Maybe it wouldn't do for you, I can't tell.

But as I actually enjoy the challenge of working with resource-constrained 'embedded' systems (my first job was designing and implementing a robotic OS and hardware 25+ years ago), this part of the fun.

Rgds

Damon

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37509680)

The amount of I/O required to break an SSD (if one is doing it deliberately) is nowhere near as much as you seem to think. One only has to do it intelligently.

I can envision a simple virus that could break SSDs willy-nilly, although its operation would be transparent to anyone who knew what to look for.

The reason you imagine these things are true is that you know much less about how SSDs work than you think you do.

SSDs with a true wear-leveling controller won't die quickly no matter what pattern of writes you do. If you know low level details about the algorithms used by a particular controller, you might be able to tailor a pattern to it to force it to operate with a higher write amplification factor than it would under ordinary loads. (Write amplification is an increase in the effective amount of data written, i.e. a factor of 1.3 would mean that for every 1.0 units of data written by the user, the drive writes 1.3x to the flash media. It's a way of measuring the overhead of wear-leveling.) But you're unlikely to grossly break it and force a ridiculously high amplification factor.

I'd guess you think that you can overwrite single sectors over and over until they fail, and use that to run the drive out of its spare capacity and thus induce failure. That might actually work on lots of USB sticks, SD cards, and so forth, but wear-leveling SSDs don't work that way. For them, the mapping between logical block addresses and physical media locations is completely arbitrary, and they change the mapping as needed to spread erase/write cycles evenly across the entire drive. (That's what write amplification factor is about; sometimes the drive must move data around to ensure that wear is even.)

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37509024)

Virtually all SSDs do some form of wear leveling and are over-provisioned to ensure that no one erase block gets worn out early.

Still a kludge. I'll be waiting for a technology that doesn't wear out at all - or at least not within a human lifetime. Flash memory is still half-baked IMHO.

Re:What will happen when they die? (5, Insightful)

lucifuge31337 (529072) | about 3 years ago | (#37509094)

Virtually all SSDs do some form of wear leveling and are over-provisioned to ensure that no one erase block gets worn out early.

Still a kludge. I'll be waiting for a technology that doesn't wear out at all - or at least not within a human lifetime. Flash memory is still half-baked IMHO.

So what exactly are you doing for data storage right now? Surely not a regular hard drive, because that doesn't meet your criteria either. Are you carving things into brass plates?

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37509690)

You miss the point - if I'm looking for a technology to replace hard drives (which, though imperfect, have been more or less your only choice until recently) I expect it to be *better* than hard drives in reliability, not just performance.

This is about upgrading; why upgrade to a crappy technology just because it's faster?

Re:What will happen when they die? (3, Insightful)

lucifuge31337 (529072) | about 3 years ago | (#37509742)

This is about upgrading; why upgrade to a crappy technology just because it's faster?

I see you haven't spent much time in the computer industry. Enjoy your Windows 95 and ham radio license.

Re:What will happen when they die? (1)

Kjella (173770) | about 3 years ago | (#37509850)

Sure, you can deliberately and forcefully break an SSD. But the amount of IO required to do so tends to go above and beyond what even the average enthusiast will do.

For me it took all of 18 months, I must say I didn't optimize my system to minimize writes and I used and abused it heavily - my OS was on it, it was running a freenet node and downloading incoming torrents, keeping it almost full so it had to work really hard to wear it levelly and so on. And it wasn't a premature failure either, the chips were rated to 10k writes and when it failed I had an average of 8.7k writes with the worst cells having almost 15k writes. If I'd taken the easy steps to reduce writes it'd probably last 5 years, with heavy steps maybe 15 years. But if you just don't care at all and use it for everything because it's so damn fast, you will run it into the ground in 1-2 years. Also in theory the cells will go read-only but file systems don't cope, I could recover most of it but the file system was corrupt.

In short, to me it's more for the user experience than the extremely high random read/write IOPS. If you use those - at least the writes - for any extended period of time you'll burn through its lifetime very quickly.

Re:What will happen when they die? (4, Informative)

0123456 (636235) | about 3 years ago | (#37508844)

Fortunately the Intel SSDs come with a 'wear indicator' showing how much life is left. Mine are all showing 99-100% life left, so unless I hit the Intel 320 8MB bug that randomly trashes the drive I don't see failure being a problem before I replace them.

Re:What will happen when they die? (1)

Jane Q. Public (1010737) | about 3 years ago | (#37508884)

That would seem to be a useful answer to the problem. It will be interesting to see how long they do indeed last under normal use.

Re:What will happen when they die? (4, Informative)

TheLink (130905) | about 3 years ago | (#37509180)

See these (their usages might match slashdotters more):
http://www.codinghorror.com/blog/2011/05/the-hot-crazy-solid-state-drive-scale.html [codinghorror.com]

These rates are probably for "normal users" (as in normal users who buy SSDs ;) ):
http://www.behardware.com/articles/831-7/components-returns-rates.html [behardware.com]
http://www.behardware.com/articles/810-6/components-returns-rates.html [behardware.com]

Note the common failure modes are not very graceful, they're usually brutal and/or weird:
http://www.dslreports.com/forum/r25491097-Dell-Laptop-and-SSD-Time-warp-issue [dslreports.com]
http://www.ocztechnologyforum.com/forum/showthread.php?83778-Time-warp-drive-vanishing-after-3-days-data-gone-on-reboot...I-need-3-to-5-users-with-this-issue-to-help [ocztechnologyforum.com]

http://www.techspot.com/news/44694-intel-confirms-8mb-bug-in-320-series-ssds-fix-available.html [techspot.com]
http://en.wikipedia.org/wiki/X25-M#Past_bugs [wikipedia.org]

In contrast with most (not all of course) of the HDD failures I've seen you still can get a lot of data out.

Re:What will happen when they die? (1)

bbn (172659) | about 3 years ago | (#37509634)

It is worth noting that most of these failures do not seem to be wear related. There must be some severe quality issues where they build these things.

Re:What will happen when they die? (1)

ShakaUVM (157947) | about 3 years ago | (#37509736)

>>It is worth noting that most of these failures do not seem to be wear related. There must be some severe quality issues where they build these things.

Exactly. Before buying my SSD, I read thousands of comments on Newegg, looking both at the highest average ratings and also filtering down to the "1 egg" rating to see what the reasons were for the low scores. If a pattern emerged ("Drive suddenly died" or "Firmware update destroyed data on the drive, and drive dies without firmware update") then I'd skip it and move on.

Normally I only care about performance in these sorts of things, but since SSDs have such a high failure rate, it became a primary criterion for me.

Re:What will happen when they die? (2)

aztracker1 (702135) | about 3 years ago | (#37509440)

Have had that on two Intel drives now... :( The one in my laptop is still going, and I used crucual & corsair drives last turn around... won't go back to HDD for boot devices, too big a difference.

Re:What will happen when they die? (4, Informative)

vadim_t (324782) | about 3 years ago | (#37508852)

So? HDDs also die. They're guaranteed to in fact, since they have plenty moving parts that will wear out eventually. I've had quite a few drives die on me.

SDDs at least in theory wear out in a predictable manner and can deal with the effects without data loss. Since flash fails on write, a SDD conceivably could (I don't know if any do that) reach a point where it says "that's it, no more redundancy left, read only access from now", which is a whole lot better than a head crash.

Everybody should have a backup plan, regardless of storage tech.

Re:What will happen when they die? (1, Troll)

0123456 (636235) | about 3 years ago | (#37508868)

So? HDDs also die. They're guaranteed to in fact, since they have plenty moving parts that will wear out eventually. I've had quite a few drives die on me.

HDDs usually fail gracefully starting with a few bad blocks, giving you time to get the data off. SSDs have a marked tendency to fail catastrophically and lose everything.

Re:What will happen when they die? (1)

DamonHD (794830) | about 3 years ago | (#37508928)

I don't think you've been doing it right and having enough fun! I've experienced plenty of catastrophic HDD failures with little warning or possible recovery, including my last MacBook's internal HDD killed pretty abruptly by static AFAIK. (I managed to recover my SSH key, but that was about it.)

Rgds

Damon

Re:What will happen when they die? (1)

Terrasque (796014) | about 3 years ago | (#37508964)

HDD's usually fail by not spinning up, or just stop answering commands properly, in my experience. Which is from 100 to *Crash* before you know what's happening.

On my fileserver I've had two disks stop working, with notices like :

[1995429.300714] sd 11:0:0:0: [sdi] Unhandled error code
[1995429.300718] sd 11:0:0:0: [sdi] Result: hostbyte=DID_BAD_TARGET driverbyte=DRIVER_OK
[1995429.300723] sd 11:0:0:0: [sdi] CDB: Read(10): 28 00 00 00 00 00 00 00 08 00
[1995429.300738] end_request: I/O error, dev sdi, sector 0

Because of RAID, I didn't lose any data, but the funny part was that even after it stopped working properly, SMART still reported it as fully functional.

Re:What will happen when they die? (1)

0123456 (636235) | about 3 years ago | (#37508988)

HDD's usually fail by not spinning up, or just stop answering commands properly, in my experience.

Whereas I've never had any of those happen. Every hard drive failure I've seen has been easily predicted by looking at the SMART data for reallocated blocks.

In fact, no drive has ever actually stopped working, probably because I get them replaced within a few days of the bad sectors appearing. Even my old laptop drive that had been sitting around in a box for a decade still worked when I plugged it in, though it had developed a bunch of bad sectors over that time.

Re:What will happen when they die? (1)

thetoadwarrior (1268702) | about 3 years ago | (#37509456)

Both of you are giving personal experiences which don't really mean anything. I've had HDs slowly die and I've yanked the data off first and some that just decided one day to not work at all. Luckily I back things up so I've not really lost anything of value since probably about 1995 when I tried transferring data from one computer to another via a big ass stack of floppies and stupidly deleted the source data after putting it on floppies only to find out a couple were bad.

Re:What will happen when they die? (1)

hedwards (940851) | about 3 years ago | (#37509394)

Google did a study awhile back and came to the conclusion that when you start seeing SMART errors that the disks are 10x or so more likely to fail than ones without. But, when it comes to HDD, or really any storage medium, when you stop having complete faith in the unit, it's time to get a new one.

I have a few HDD that probably will work for some time, but since I don't trust them and I can't prove them to be reliable, they're going to be recycled.

It kind of sucks, but the disks are a lot less expensive to replace than they are to do data recovery on.

Re:What will happen when they die? (1)

aztracker1 (702135) | about 3 years ago | (#37509470)

I've seen Raid-1 fail in such a way that the second drive fails within a few days of the first drive (twice)... IMHO it's urgent to replace that first failed drive quickly.. as the second is likely to go soon. The first time the second drive died before the RMA replacement made it back... for my NAS box, It's raid-5 with a spare drive sitting next to the box, just in case.

Re:What will happen when they die? (1)

rev0lt (1950662) | about 3 years ago | (#37509658)

As usually the disks are from the same provider, and often from the same lot, they share the same problems. If you're building a raid-1/10/3/4 setup, you should buy your disks (or disk sets) with the same specs, but from different manufacturers. Today it is easier said than done on server-grade stuff, but still possible on homemade builds.

Usually? (1)

SuperKendall (25149) | about 3 years ago | (#37509002)

HDDs usually fail gracefully starting with a few bad blocks

I have never had a hard drive fail in this way. I have never seen a SMART status go bad before I had a very sudden loss.

I have had several drives that simply would not spin up one day (even after a trip to Mr Freezer). I have had large swaths of data oct in catastrophic storms, barley able to recover half my data after multiple passes with recovery tools.

I've been fortunate in that I've lost almost no data, primarily because I was able to recover recent work. But over time I've become more and more paranoid to the point that I will not go daily without at least one backup if possible. So I could care less if an SSD fails suddenly, even though it seems like that is less likely with an SSD than a mechanical drive.

Re:Usually? (2)

TheLink (130905) | about 3 years ago | (#37509358)

I have never had a hard drive fail in this way. I have never seen a SMART status go bad before I had a very sudden loss.

What do you use to monitor SMART on your drives?

Re:What will happen when they die? (1)

Anonymous Coward | about 3 years ago | (#37509044)

Hard disks fail in many ways. I've had hard disks fail with a few isolated bad blocks rapidly spreading to the rest of the drive. I've seen a broken head, where a quarter of the capacity instantly became unreadable. One drive went from 100% OK to absent from the bus over night. I've had a drive that wouldn't spin up anymore until I (literally) kicked it. I once struggled with a drive that would work fine for hours and then randomly corrupt data (and it wasn't a power-of-2 overflow issue).

Needless to say, I have actual backups, not just plans to make backups, knock on wood.

Re:What will happen when they die? (1)

isama (1537121) | about 3 years ago | (#37509244)

One of the strangest failures I've ever seen is a drive which will work perfectly when held vertically and do nothing when horizontal. I still use it, as a backup of backups... Just for fun.

Re:What will happen when they die? (1)

saleenS281 (859657) | about 3 years ago | (#37509416)

If you aren't taking backups, or at the very least using RAID, expect to lose data. "A few blocks" if they're the wrong blocks is everything.

Re:What will happen when they die? (1)

Redlazer (786403) | about 3 years ago | (#37509468)

Psh. I've NEVER had HDD drive gracefully, and I've lost at least 20.

I had a laptop drive right in the middle of booting into Linux once. That was cool. Typed in my password, boomheadcrash.

"What was that noise? Why is it taking so long to load?"

Re:What will happen when they die? (1)

Tapewolf (1639955) | about 3 years ago | (#37509184)

SDDs at least in theory wear out in a predictable manner and can deal with the effects without data loss. Since flash fails on write, a SDD conceivably could (I don't know if any do that) reach a point where it says "that's it, no more redundancy left, read only access from now", which is a whole lot better than a head crash.

My boot drive failed by destroying half the Linux partition. I was able to copy off /etc, kernel config and a bunch of useful scripts and things, but most of it was just a bunch of unreadable sectors. Shortly afterwards the drive failed completely and was no longer recognised as a disk. It was just under a year old - I used the noatime option, swap was on an HDD and it was only about 3/4 full.

Re:What will happen when they die? (2)

izomiac (815208) | about 3 years ago | (#37509220)

Since flash fails on write, a SDD conceivably could (I don't know if any do that) reach a point where it says "that's it, no more redundancy left, read only access from now", which is a whole lot better than a head crash.

That's been my experience exactly. Every PC I've owned has "died" from a HDD crash, usually sudden. The last SSD I had hit its erase limit in about two years (small SSD and I'm prone to reinstalling various OSes monthly). The lovely thing was that I could run a maintenance tool and see exactly how many erases were left on each cell (BTW the wear leveling was only 1% from mathematically perfect). This allowed a simple extrapolation down to the day some cells would start hitting their advertised capacity, although the drive held out overall a bit longer.

The initial symptoms were Windows blue screening on boot (1,000,000 writes per boot, so no surprises there), so I quick formatted the disk and reinstalling thinking it was probably Windows sucking again. From there the drive lasted a couple more days and became completely read-only. I keep good backups so I didn't need to salvage anything, but even now I can throw it in a USB enclosure and get my data off of it.

Re:What will happen when they die? (1)

klui (457783) | about 3 years ago | (#37509960)

Predictable in theory but not necessarily in practice. When your metadata goes RO on you I think that's why some responses are in the area of a corrupt file system or a drive that is not recognized at all. The ones about partitions being corrupted is a much bigger issue. You would think one wouldn't normally mess with partition metadata.

Probably the file systems and low level utilities need to be updated to take into account SSD-type failures.

Re:What will happen when they die? (3, Informative)

SiMac (409541) | about 3 years ago | (#37508854)

Does anybody have a backup plan for when their SSDs die? After all, unlike magnetic media, SSDs have a limited number of writes. AFAIK, none of them are rated yet for over a million writes, so they are bound to fail at some point.

Buy a new SSD? SSD failure is predictable. If you're lucky, your firmware will not try to write to blocks that are past their rated # of write cycles and so when your SSD reaches the end of its lifespan, your data will become read only. Even if not, you can still tell very easily if you're approaching end of lifespan using SMART status. I suspect that SSD death is much more predictable than HD death...

Re:What will happen when they die? (1, Flamebait)

Jane Q. Public (1010737) | about 3 years ago | (#37508986)

"If you're lucky, your firmware will not try to write to blocks that are past their rated # of write cycles"

You would have to be very lucky, since such a creature does not exist.

Maintaining a count of how many times any given cell has been written would take a lot more memory (not to mention processing power) than these devices contain.

Instead, what they do is over-provision, so that a detected bad block is replaced with a spare. (Most hard drives do much the same thing.) However, there are only so many spares.

As someone else mentioned: with any real luck your firmware might report what percentage of those "spare" cells are left. If it doesn't, then you are left with sudden unexpected failure when the last of them is used up and another cell goes bad.

Re:What will happen when they die? (4, Informative)

vadim_t (324782) | about 3 years ago | (#37509080)

Maintaining a count of how many times any given cell has been written would take a lot more memory (not to mention processing power) than these devices contain.

Bullshit.

SSDs erase in extremely large blocks, like 256K. Having a counter per block is not a problem. It works out to 16K of memory per GB for a 32 bit counter per block.

It probably doesn't even take an extra space, since a block probably already contains metadata and ECC, so a simple counter probably fits in there nicely, It won't even cause any extra wear because the only time you want to change the counter is when the block is being rewritten anyway.

Predictable? (2)

TheLink (130905) | about 3 years ago | (#37509132)

SSD failure is predictable.

That's bullshit. You call the following predictable?
http://www.dslreports.com/forum/r25491097-Dell-Laptop-and-SSD-Time-warp-issue [dslreports.com]
http://www.ocztechnologyforum.com/forum/showthread.php?83778-Time-warp-drive-vanishing-after-3-days-data-gone-on-reboot...I-need-3-to-5-users-with-this-issue-to-help [ocztechnologyforum.com]

http://www.techspot.com/news/44694-intel-confirms-8mb-bug-in-320-series-ssds-fix-available.html [techspot.com]
http://en.wikipedia.org/wiki/X25-M#Past_bugs [wikipedia.org]

I might buy a Samsung SSD. The rest (except for Intel) don't have such a great track record even when compared to hard drive failure rates (and Intel's failures haven't been very confidence inspiring).

http://www.behardware.com/articles/831-7/components-returns-rates.html [behardware.com]
http://www.behardware.com/articles/810-6/components-returns-rates.html [behardware.com]

For some people the failure is predictable in that they can almost bet the drives will fail within a year! http://www.codinghorror.com/blog/2011/05/the-hot-crazy-solid-state-drive-scale.html [codinghorror.com]

But I don't regard that sort of predictability of failure as acceptable, unless the manufacturer is paying me to use their products and gives me plenty of spares.

Re:Predictable? (1)

orange47 (1519059) | about 3 years ago | (#37509532)

it is obvious they are referring to one kind of SSD failure that is predictable.
a lightning strike for eg. is quite unpredictable, and it might cause SSD/HDD failure.

Re:Predictable? (3, Insightful)

vadim_t (324782) | about 3 years ago | (#37509788)

Different kind of failure. You're linking to firmware bugs. HDDs have those as well [mswhs.com]

In this thread we're discussing wear induced failure.

Re:What will happen when they die? (1)

ShakaUVM (157947) | about 3 years ago | (#37509754)

>>Buy a new SSD? SSD failure is predictable.

Write failures are predictable and reportable.

Unfortunately, the more common SSD failure mode is turning it on one day and it doesn't work. As I mentioned elsewhere, I read through thousands of comments on Newegg looking both at the overall score and the 1-egg ratings, to see if failure patterns emerged. One I remembered was a drive that would arbitrarily corrupt its data without a firmware update, but the firmware update also would wipe the drive...

Re:What will happen when they die? (1)

Eluan (788868) | about 3 years ago | (#37508894)

So... A virus could force my ssd to die within days or months? Interesting, I hadn't tought about this!

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37508900)

SDD's can die in two ways, one is when they loose the table to map the actual data to how its stored on the chips and the other is when you can't write anything anymore, but can still read everything. However, today, it's mostly the table that gets lost, so you have a black hole effect... usually with the sand force or other controllers aimed at speed instead of reliability.

Samsung goes for reliability instead of speed, so you should be good with regular backups like usually.

Re:What will happen when they die? (1)

hedwards (940851) | about 3 years ago | (#37509412)

Personally, that's fine by me, since I don't consider SSDs to be cost effective for a desktop, the ones I get always go in laptops for that reason.

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37508924)

Luckily, I don't plan to run your poxy program - so my backup plan is the same as it's always been.

a) most of the stuff on my disk is cache, or close enough to - i.e git repositories that are frequently pushed elsewhere, mail that is just a replica of the IMAP server.

b) anything else is backed up to another machine that actually has RAID on a regular basis.

But my laptop just runs a whole lot better with an SSD, and is nicer to use. The SSD (x25m 80Gb in my case) has been running happily for over a year - that's a pretty good run for something that makes me safe against basic power failures and is blindingly fast.

Re:What will happen when they die? (2)

0123456 (636235) | about 3 years ago | (#37508944)

The SSD (x25m 80Gb in my case) has been running happily for over a year - that's a pretty good run for something that makes me safe against basic power failures and is blindingly fast.

I believe that 'basic power failures' were the primary cause of the Intel 320 8MB bug; from what I've read it seemed that when the power went out it didn't update the mapping table properly so the drive was toast when you rebooted.

Re:What will happen when they die? (2)

Terrasque (796014) | about 3 years ago | (#37508938)

This might be of interest to you :

SSD Write Endurance 25nm Vs 34nm [xtremesystems.org]

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37508952)

After all, unlike magnetic media, SSDs have a limited number of writes.

This is just plain wrong. Magnetic media also have a limited number of writes but due to the usage of solid state memories in critical applications it is far better specified. Good luck finding out how many write your magnetic media will tolerate before worn out.

When SSDs were newer, I argued here on /. (against vociferous claims to the contrary) that I could write a program that would break an SSD quickly.

I do not doubt that at all, the process if fairly straightforward for magnetic drives, I don't see why the same shouldn't apply to SSDs.

Re:What will happen when they die? (5, Informative)

fuzzyfuzzyfungus (1223518) | about 3 years ago | (#37508960)

Because they are somewhat more expensive, an SSD failure is a little more painful than an HDD failure; but the basic rules of "don't trust a hard drive" really haven't changed.

The mechanicals sometimes last a decade if you get lucky, or die within days of install if you don't. Moral of the story: If you store anything on a hard drive, you don't love that something very much. You'd better have backups.

The shape of the failure probability/time graph is likely a bit different for SSDs; but the "You'd better have backups" message, and the available means of taking those backups are pretty much exactly the same.

Again, because of the somewhat higher cost, burning your way through SSDs is a little more painful than burning your way through HDDs; but anybody whose plans involve just trusting a hard drive has always been doomed.

Re:What will happen when they die? (1)

jovius (974690) | about 3 years ago | (#37509214)

What exactly happens when an SSD dies. Are the cells just read-only then or complete garbage?

Re:What will happen when they die? (1)

Bengie (1121981) | about 3 years ago | (#37510138)

If it's a write cycle death, it should die gracefully in a read-only state. If it dies for defect reasons, then it's a up to luck.

Re:What will happen when they die? (0)

Anonymous Coward | about 3 years ago | (#37509252)

They go to SSD heaven of course!

Re:What will happen when they die? (1)

Anne Thwacks (531696) | about 3 years ago | (#37509766)

Unlike HDDs which, in my experience, generally go strait to hell! (I know this because: there is often an earsplitting scream of pain as they go!)

Re:What will happen when they die? (1)

aztracker1 (702135) | about 3 years ago | (#37509396)

I only have applications and the OS for the most part on my SSD, On my laptop anything important is in my dropbox directory, that syncs to my desktop... My desktop gets daily backups (+ whatever is in the dropbox), I have a few HDDs as well in my desktop all also backed up. I have a 4x1TB nas box (synology), but have out grown it, so will be building an (up to) 11 drive nas probably based around FreeNAS. within the next few months.

I've been a bit tepid in doing the upgrade to the homebrew nas solution, as 3TB drives still aren't at a reliability level I would like, and FreeNAS didn't support the 2-drive redundancy version of ZFS as recently as a few months ago... plan is an initial 5x3TB drive configuration with ZFS(2-drive redundancy) giving me 14.5TB of usable storage and when I need it, adding a lower cost raid card for another 6 drives (with 8 supported) with a similar config, for another 17.3TB usable storage. With FreeNAS installed to a USB thumb drive mounted internally.

My advice to anyone going SSD, is if you run windows, go for something >=80GB as once you get apps installed, it tends to fill up quickly. Get familiar with the mklink command (symbolic links), as it's really helpful for mounting your media to an HDD, or relocating rarely used games and programs, while keeping the mount point where it needs to be. Do this with GameTap, Steam and similar programs. If you are in a laptop, go at least 160GB as you will want that bit extra if you intend to carry any media with you or do VMs. If you're running a lighter linux distro without a crap ton of games/media, then 40-64GB may be suitable.

Re:What will happen when they die? (1)

aztracker1 (702135) | about 3 years ago | (#37509402)

grr... 8.5TB and 11.3 respectively usable... Need to proofread better.

Re:What will happen when they die? (1)

thetoadwarrior (1268702) | about 3 years ago | (#37509404)

No drive has unlimited writes. They all die eventually so the back up plan is the same as before. You either back-up nothing and lose it or back-up often and don't worry if it dies.

Re:What will happen when they die? (1)

couchslug (175151) | about 3 years ago | (#37509422)

"Does anybody have a backup plan for when their SSDs die? "

Same plan as ever. If it matters, burn it to DVD at slow speed. If it's large and matters, copy to different computers and external hard drives.

Re:What will happen when they die? (1)

Bengie (1121981) | about 3 years ago | (#37509668)

"Does anybody have a backup plan for when their SSDs die? After all, unlike magnetic media, SSDs have a limited number of writes"

Most 128GB drives are good for 10TB+ per day for 5 years. Even when you hit the write cap, you can still read from the drive, it just turns read-only.

Re:What will happen when they die? (1)

Surt (22457) | about 3 years ago | (#37510246)

I have bad news for you about your magnetic media, which also has a limited number of writes.

Big questions (4, Informative)

rsborg (111459) | about 3 years ago | (#37508832)

1) How does this Samsung chipset compare vs latest Sandforce2 in terms of compressed read/writes?
2) TRIM support?
3) OSX friendliness?
4) Cost?
5) Size max?

So far I've identified 2 use cases that have very nice sweet-spot answers - a) For a desktop with PCI-e, the OCZ Revodrive3 X2 just gives amazing performance, completely bypassing SATA and delivering unbeatable performance/cost ratio. b) For a laptop solution, I'm more interested in max storage/price/performance, and the 512GB Crucial m4 seems unparalleled in delivering this (expensive at $700, but can completely replace an laptop HDD).

It will be interesting to see if Samsung is ready to challenge this market.

Re:Big questions (1)

Anonymous Coward | about 3 years ago | (#37509028)

Home users do not need large capacity SSDs. Putting the OS and applications on the SSD gives a massive performance boost. Putting media on a HDD is more than adequate because you never read it beyond playback bandwidth requirements, which is trivial compared to SATA rates.

Re:Big questions (1)

tepples (727027) | about 3 years ago | (#37509070)

Putting the OS and applications on the SSD gives a massive performance boost. Putting media on a HDD is more than adequate because you never read it beyond playback bandwidth requirements

Are video games "applications" or "media"?

Re:Big questions (1)

Anne Thwacks (531696) | about 3 years ago | (#37509776)

Are video games "applications" or "media"?

No. Next question please.

Re:Big questions (0)

MemoryDragon (544441) | about 3 years ago | (#37509154)

Trim is relatively pointless if you have a good garbage collection. Most Sandforce drives have a gc mechanism which works like a deferred trim.
(Aka once a block should be overwritten and already is allocated it is stored and marked as deletable, the new data is written elsehwhere and when the drive idles the block is erased.

Samsungs however in the past relied on filesystem features and hence only worked with NTFS properly in this area. It would be more interesting to see if Samsung finally has resolved that issue.

Re:Big questions (2)

Nahor (41537) | about 3 years ago | (#37509796)

A good garbage collection can't replace TRIM. Overwritten blocks are not the only ones that can be GC'ed. There is also the blocks from deleted files. And unless the FS uses those blocks first (which any recent filesystem will avoid to do to prevent fragmentation on hard-drives), they will eventually use a significant amount of the SSD and kill the drive performance.

And worse, some modern filesystems use "copy-on-write", so no data is ever overwritten (from the SSD point of view) the SSD performance will drop even more quickly.

Re:Big questions (1)

Nemyst (1383049) | about 3 years ago | (#37509896)

AnandTech has a review [anandtech.com] up. For your questions, the drive has TRIM support, Samsung has been the SSD manufacturer of choice for Apple so I'd say OSX support is a given, costs will be in line with the SSD 470 which is within the range of the OCZ Vertex 3, Crucial m4 and Intel SSD 510 and it goes up to 512gb, which is the sample that has been reviewed.

Notably, they say that this is the first really exciting release by Samsung. Apparently, garbage collection is delayed to moments with low IO activity, making torture tests dip down to as low as 50mb/s, but on the flip side this boosts peak speeds. In normal operation such issues are not quite as problematic. Another thing to note is that this SSD is less dependent on compression. SandForce based drives like the Vertex 3 suffer tremendously with incompressible data, whereas Samsung's offering doesn't dip that much.

Price??? (3, Informative)

Rick Richardson (87058) | about 3 years ago | (#37508856)

"It's really hard to rate a solid-state drive (SSD) without knowing its exact pricing, and that's just what we had to do with the Samsung 830 series. Samsung has been very tight-lipped about how much the 830 costs and will not reveal that until the drive is available for purchase in October." - CNET

Re:Price??? (1)

d4fseeker (1896770) | about 3 years ago | (#37508882)

In other words; it's gonna be very expensive and they dont want to trash all their good pub right now/

Re:Price??? (3, Insightful)

timeOday (582209) | about 3 years ago | (#37509118)

New high-end products reduce prices by devaluing older products, not by being cheap in themselves.

Re:Price??? (1)

klui (457783) | about 3 years ago | (#37510096)

From Anandtech http://www.anandtech.com/show/4863/the-samsung-ssd-830-review [anandtech.com] :

"The Samsung SSD 830 will be available to consumers starting in mid October. Although Samsung isn't announcing pricing at this time, I've been told to expect the drive to be priced around where the SSD 470 is today."

Dell OEM Samsung drives are bad (4, Informative)

DigiShaman (671371) | about 3 years ago | (#37508880)

If you're purchasing a Dell, stay away from the Samsung SSD option as they're OEM. The drives are absolute shit. Most likely a firmware issue, often Windows will just freeze because writes cannot be further committed. I've been through two different Dell laptops models and they experienced the same issue using this same drive. Only when we swapped drives did the issue go away. And that was after Dell decided to swap the motherboard, ram, CPU, and video card. Nice.

Re:Dell OEM Samsung drives are bad (0)

Anonymous Coward | about 3 years ago | (#37509320)

I have excactly that issue, if anybody knows any more details, why/what, then please spill!!
My firmware is VBM24D1Q.

Re:Dell OEM Samsung drives are bad (0)

Anonymous Coward | about 3 years ago | (#37510082)

I've read something about it being due to the owner's immense faggotry.

who makes it ? (-1)

Anonymous Coward | about 3 years ago | (#37509050)

If you are going to write a slashvertisment please dont write your company name 8 times in a 60 word summary

Re:who makes it ? (-1)

Anonymous Coward | about 3 years ago | (#37509392)

Slashvertisements are only good for Makerbot, right?

External drives (0)

Anonymous Coward | about 3 years ago | (#37509718)

These are probably best suited for two case scenarios:
1) Read-only/Read-Mostly, external USB drives/eSata/Thunderbolt where the majority of the usage is reading, like with video and photographic work.
2) Archives, where backups are rotated. In this scenario the fast read and write speed is desireable, but cycled through infrequently, it's more energy efficient than tape and conventional hard drive backups since they can be powered off and on with no wear.

For everyday use, I'd not recommend SSD's except as a boot/application installation for launch speed, however the OS and software (eg photoshop) need to be configured to use a conventional drive for scratch/swap file space since they will rapidly wear out a SSD.

Re:External drives (1)

Z34107 (925136) | about 3 years ago | (#37510174)

Why on earth would you backup to a solid state drive? I think the only way you could spend more money per gigabyte of backup involves robots.

For everyday use, I'd not recommend SSD's except as a boot/application installation for launch speed, however the OS and software (eg photoshop) need to be configured to use a conventional drive for scratch/swap file space since they will rapidly wear out a SSD.

No, they won't.

fusion-io drives 10000x better (0)

Anonymous Coward | about 3 years ago | (#37509798)

fusion-io still beats this drive by completely skipping the SATA layer. It's like having a SAN on a board.
770 MB/s

http://www.fusionio.com/products/iodrive/

Load More Comments
Slashdot Login

Need an Account?

Forgot your password?