Samsung puts the kibosh on SSD reliability worries

If you're one of the (apparently) many out there scared half to death not over what's residing in your closet at night, but that your recently purchased SSD will wear itself out in no time flat, here's a little reassurance. According to Michael Yang, flash marketing manager at Samsung, a flash device rated at 100,000 write cycles can write 100,000 times "to every single (memory) cell within the device," and of course, it won't continuously write to the same cell over and over thanks to a process dubbed wear leveling. In case you still needed a tad more evidence, he also stated that "a pattern could be perpetually repeated in which a 64GB SSD is completely filled with data, erased, filled again, then erased again every hour of every day for years, and the user still wouldn't reach the theoretical write limit." So, now that we've had that cleared up, why not tell us more about those 250GB MLC-based SSDs that could land before 2009 dawns?

[Anton @ Feb 23rd 2008 9:38PM]

SSD + Price Drop = WIN!

[E71 @ Feb 23rd 2008 10:15PM]

My main concern with SSD is write speeds. I hear SSDs are faster @ reading but slower @ writing -- great for my fileserver, not so great for my database server.

[mentalsticks @ Feb 23rd 2008 10:20PM]

actually, it's bloody great for your database server, as random read/write times are much better than HD's.
Then again, I don't know shit about this so I'd best shut up.

[Mike @ Feb 23rd 2008 10:29PM]

Sequential read write speeds are better on spinning discs after the read/write head has moved into position. Random read write speeds on SSDs is ridiculously better because there's no disc head to move. There's no generic 'this is good for all databases' answer, but if you have a lot of small random reads or writes, an SSD could be very good for you. It's really best that you evaluate it yourself through test cases and benchmarks, because there's no right answer for everyone.

[Sam Winter @ Feb 24th 2008 3:01AM]

@E71

Granted, MLC flash SSDs will be cheaper and their sequential write speed usually tops out at around 45MB/s, but very fast SLC flash SSDs are available and obviously will come down in price as time goes on. For a database server, these would be incredible. The SSD would come pretty close to matching an enterprise SAS drive on sequential, sustained write/read speed, AND at the same time it would absolutely blow it away in random read/writes because of the

[Sam Winter @ Feb 24th 2008 3:02AM]

@E71
damn comment system... it just cut off 1/2 of my post!!!! I'll repost:

Granted, MLC flash SSDs will be cheaper and their sequential write speed usually tops out at around 45MB/s, but very fast SLC flash SSDs are available and obviously will come down in price as time goes on. For a database server, these would be incredible. The SSD would come pretty close to matching an enterprise SAS drive on sequential, sustained write/read speed, AND at the same time it would absolutely blow it away in random read/writes because of the

[Sam Winter @ Feb 24th 2008 3:03AM]

WHY THE FUCK IS IT CUTTING OFF MY POSTS!!!!! Im going to post in the data without the links to reviews...

Samsung SATA II 64 GB SSDs
Sequential Read: 120MB/s
Sequential Write: 100MB/s

MTron 2.5" 32GB SSDs
Sequential Read: 95 MB/s
Sequential Write: 75 MB/s

BitMicro 2.5" 32GB-416GB SSDs
Sequential Read: 100MB/s
Sequential Write: 100MB/s

[why not the LS2LS7? @ Feb 24th 2008 3:15AM]

Weblogs Inc's system cuts off your post at the first less-than sign with no warning.

[James Cameron @ Feb 24th 2008 4:18AM]

Can you have these in RAID?

[Mr. S. @ Feb 24th 2008 4:35AM]

How much more efficient are these things? Am i going better batt life out of my laptop? I would think i should...

[Seth @ Feb 25th 2008 4:11PM]

@E71

Not all SSDs have slow write speeds.
Just buy one that has fast write speeds - like an MTRON.

[zunq @ Feb 23rd 2008 9:45PM]

Too bad most of our worries reside on the price of these things.

[Mark @ Feb 23rd 2008 9:55PM]

True that. I'd happily pay more for an SSD, but not the premiums they command today.

[mattclarkie @ Feb 24th 2008 10:28AM]

I would say that I would be willing to pay 100% more than the cost of an similarly sized HDD, but don't they sell for around 400%.

[Ken @ Feb 23rd 2008 10:10PM]

Lets see, my laptop has a 160GB 7200RPM drive. How long until a 256GB SSD will be $300?

[Blaktornado @ Feb 23rd 2008 10:12PM]

A looooooooooooooooooooooooooooong time.

[Ken @ Feb 23rd 2008 10:21PM]

Yeah, even if they COULD make them cheaper, I think they will keep the prices high for some time.

[Chuckles McGee @ Feb 24th 2008 12:46AM]

@Ken

I think companies will charge whatever price will maximize net profit for them.

[Dualboot @ Feb 24th 2008 1:08AM]

There's a little thing called competition which will help drive the price down, especially since we will have SSD manufacturers competing with HD manufacturers and computer makers trying to decide which way to go...

[Jason @ Feb 23rd 2008 10:17PM]

What idiot was worried about SSD reliability?

[Ken @ Feb 23rd 2008 10:22PM]

The same person who reads Consumer Reports before buying an electric can opener.

[Jason @ Feb 23rd 2008 10:34PM]

Sir, where I come from, talking about one's mother is fighting words, and I don't appreciate your comment.

[mentalsticks @ Feb 23rd 2008 10:18PM]

but can it play Doom?

[Joshua Walters @ Feb 23rd 2008 10:23PM]

Can you play Doom?

Im guessing the answer is no, since you are so concerned with making sure that EVERY device can.

[mentalsticks @ Feb 23rd 2008 10:55PM]

Actually, I hate Doom. it makes me nauseous.

[JuggleNuts @ Feb 23rd 2008 11:30PM]

Especially those goddamn imps!

[Ghen @ Feb 24th 2008 12:22PM]

Since you hate doom you're just perpetuating a meme which is a crime punishable by castration.

[Thinker @ Feb 23rd 2008 10:18PM]

Assuming you write 1 GB worth of data each day, it will take 64 days to fill the card. So one ("1") write cycle = 64 days. A 100,000 write cycle will take 100,000 x 64 = 6.4 million days of usage at 1 GB per day before the limit is reached. This is the equivalent of 6,400,000/365.25 = 17,522 years of usage.

[austin @ Feb 23rd 2008 10:47PM]

summary- SSDs DO NOT WEAR OUT

thank you for that math.

[AJ in the East Bay @ Feb 23rd 2008 10:51PM]

So basically the future is here.

[allen_singer @ Feb 23rd 2008 11:21PM]

methinks u suffer from some faulty logic right there.

some parts of the drive will get written to much more than others - say for applications and OS files (like your pagefile). A few clusters may even be written to once every few hundred clock cycles - like if the clusters contain a log file/parts of a pagefile.

What does that mean? Over the course of the first few months, random parts of your drive will wear out. In a year, the drive will be useless.

Trust me. Your back-of-the-envelope calculations aren't realistic.

[Carl @ Feb 23rd 2008 11:28PM]

allen_singer: RTFA... wear levelling prevents any one sector of the drive being written to significantly more than any other, it will "move" the page file to a different sector if it is being continually written to.

[Hardcore @ Feb 23rd 2008 11:43PM]

Damn, and I was hoping to leave some MP3's so that the RIAA can sue my great x 1000 grand child in the year 19531.

[chuck @ Feb 24th 2008 12:22PM]

Too bad, one movie (.avi or .mp4) averages out to about 700MB - 1GB... I like your math and it makes sense, its just your avg daily data usage of 1GB seems too low.

[Ghen @ Feb 24th 2008 12:26PM]

Chuck, so assume its 100GB daily. that means 170 years of use. Slightly better than the 4-10 year lifespan of a spinning disc HD.

[Seth @ Feb 25th 2008 4:12PM]

I think the only consumers got all chicken little about SSD reliability to begin with is that people don't like to hear they have a 'write limit' - they feel much more comfortable with 'MTBF' - even though your average hard drive is certainly going to fail long before any given flash product.

It's kind of a funny psychological quirk.

[lowbot @ Feb 24th 2008 3:11PM]

Look, this isnt about downloading or moving data. This is everything your computer does. Does it write a 1kb file every .5 seconds because you have logging enabled on some application? Does windows use the swap file even when it has memory? How much does it move?

You probably write many hundreds of gigabytes per day, easily. I would guesstimate a power-users computer writes many terabytes of random junk.

I'm not saying that SSDs are not ready for the market, but I wouldnt be surprised if there is a class of users who will burn theirs out early and normal usage will not last longer than a typical modern hard drive. I would also be concerned about server usage. I doubt we'll be seeing serious servers with this stuff until theyve improved the process, if thats even possible.

[sholt @ Feb 25th 2008 10:52AM]

To give you a better idea:

Lifetime(min) = (Drive Capacity)/(Max write speed) * max lifetime cycles

So, for a 64GB SSD with a 45 MB/s write speed and 100,000 lifetime cycles, the drive will last, at minimum,

(64 GB / 45 MB/s) * 100,000 = ~4.6 years.

So, under *constant* writes, a 64GB drive will last over 4 years. Because of the wear leveling, a larger capacity drive will last longer.

[Octothorpe @ Feb 25th 2008 3:07PM]

Thank you sholt for the correct math.

@45 MB/s write you get 4.51 years

@100 MB/s you get a measly 2.03 years.

Granted, given average use, you would still end up with far more than that. These minimum calculations equate to rewriting the entire contents of the drive every 23.7 minutes @ 45 MB/s or every 10.67 minutes at 100 MB/s for its entire lifespan (this also shows you how fast these things are).

[monkeypoop @ Feb 23rd 2008 10:45PM]

test

[Arnie @ Feb 23rd 2008 11:06PM]

results are out you failed!!

[egotman @ Feb 24th 2008 11:57PM]

Was the test for what was coming out of your monkey butt?

I kid, I kid....

[Mark @ Feb 23rd 2008 11:10PM]

Call me when I can buy a 500 gigabyte SSD drive for around $150..

[Infamouship @ Feb 23rd 2008 11:26PM]

Also add a 10 MP camera, 4 gig ddr2 ram, biometric scanner, M80's and streamers. I'll definitely call you bro!!!!

[Dan @ Feb 23rd 2008 11:43PM]

Okay, so let's assume you write 4 GB worth of data each day, including the swap file/memory paging, etc. It will take 16 days to fill the card. So one ("1") write cycle = 16 days. A 100,000 write cycle will take 100,000 x 16 = 1.6 million days of usage at 4 GB per day before the limit is reached. This is the equivalent of 1,600,000/365.25 = ~4,380 years of usage.

I don't think it'll be dying any time soon?

[Dan @ Feb 23rd 2008 11:45PM]

sorry, posted the response in incorrect location.

[Sam Winter @ Feb 24th 2008 3:07AM]

Call me when you are not such a cheap ass, and I'll happily sell you good hardware for a good price.

[TIMMAH! @ Feb 23rd 2008 11:11PM]

"Assuming you write 1 GB worth of data each day, it will take 64 days to fill the card. So one ("1") write cycle = 64 days. A 100,000 write cycle will take 100,000 x 64 = 6.4 million days of usage at 1 GB per day before the limit is reached. This is the equivalent of 6,400,000/365.25 = 17,522 years of usage."

I think you can easily "write" 1GB of data a day if you take into account the underlying memory paging that's going on in the OS. If you're only using this as auxiliary file storage then your example is correct.

[DarkLightConnection Unbanned @ Feb 23rd 2008 11:33PM]

RAM drive for the OS, SSD for storage... And dump to/from the SSD at start-up/shutdown ONLY...

I'd call that a "blazin' fast setup", if it wasn't because it's running on a Pentium III....

[Dan @ Feb 23rd 2008 11:45PM]

Okay, so let's assume you write 4 GB worth of data each day, including the swap file/memory paging, etc. It will take 16 days to fill the card. So one ("1") write cycle = 16 days. A 100,000 write cycle will take 100,000 x 16 = 1.6 million days of usage at 4 GB per day before the limit is reached. This is the equivalent of 1,600,000/365.25 = ~4,380 years of usage.

I don't think it'll be dying any time soon?