OCZ is developing junk

[SickBeast]

http://www.xbitlabs.com/news/storage/display/20111104123908_OCZ_Plans_Cheap_SSDs_Based_on_Triple_Bit_Per_Cell_NAND_Flash.html

1000 write cycles is not enough IMO.

I suppose 100gb for $20 would be tempting, but I highly doubt that this new process is really going to drop prices that low.

Perhaps there are servers out there that only read data, in which case a drive like this could work.

This thread has gone down the tubes. SickBeast, please feel free to make a new thread
-ViRGE

 

[shabby]

Lol disposable ssd's.

 

[frostedflakes]

If drives using NAND rated for 3k-5k write cycles can handle hundreds of TB of writes, 1k would probably be good enough. Wouldn't use one in a server or other heavy I/O environment, but for a desktop or laptop it would probably be fine. Will depend on pricing and how much cheaper they are, though.

To put things into perspective, I've had my 120GB Vertex 2 (one of the older 3Xnm models, not 25nm) for about a year, have written ~3.3TB to it over that period, and the MWI is still 100%. That means that at the current rate I've been writing to the drive, it would take at least 100 years before the P/E cycles of the flash would be exhausted.

Let's say the drive was using flash rated for 1k P/E cycles instead of 5k. Means it would only last for >20 years instead of >100. Yes it's quite a bit less durable, but even 20 years is overkill, so it doesn't really matter.

 

[groberts101]

just like the commercial says.. "it pays to discover". Obviously much more to it than you are giving credit for as all controllers are only getting better and better at reducing PE/c consumption.

Plus it helps to make it look like you know what you're talking about. lol

http://www.storagereview.com/ocz_provides_vision_next_generation_ssds_and_processors

PS. NVM based drives with anywhere even close to 3.2 million IOPS(hell.. I'll settle for a measily Million IOPS) would definately be "junk" that I would love to own.

 

[reynoldsjrmy]

:thumbsup:

That means that at the current rate I've been writing to the drive, it would take at least 100 years before the P/E cycles of the flash would be exhausted.

Hi 'flakes,

My Vertex LE suggests the same. However, I remember that with earlier versions of SF firmware a lot of folks were getting spooked by the FW reporting early threats to life expectancy. Magically, these threats dissappeared with later fw. So call me cynical, but I suspect that your straight line projection of 100 years is massively overstated.

I imagine that 3 or 4 years in we will suddenly see a spurt towards a much earlier death reported. I dare say though that our SSD's will live long enough to see us wanting to replace them with some form of 3D Holo or organic based storage technology :thumbsup:

Regds, JR

 

[lehtv]

To put things into perspective, I've had my 120GB Vertex 2 (one of the older 3Xnm models, not 25nm) for about a year, have written ~3.3TB to it over that period, and the MWI is still 100%.

How does one go about determining the amount of data written, and what does MWI stand for?

 

[StinkyPinky]

I've had my SSD for two years and it is still 99%. People are a bit anal about it this issue.

 

[StinkyPinky]

How does one go about determining the amount of data written, and what does MWI stand for?

Crystal Disk Info can do it I guess. MWI = Media Wearout Indicator.

 

[frostedflakes]

:thumbsup:

Hi 'flakes,

My Vertex LE suggests the same. However, I remember that with earlier versions of SF firmware a lot of folks were getting spooked by the FW reporting early threats to life expectancy. Magically, these threats dissappeared with later fw. So call me cynical, but I suspect that your straight line projection of 100 years is massively overstated.

I imagine that 3 or 4 years in we will suddenly see a spurt towards a much earlier death reported. I dare say though that our SSD's will live long enough to see us wanting to replace them with some form of 3D Holo or organic based storage technology :thumbsup:

Regds, JR

IIRC that was some bug that was causing it to incorrectly read MWI.

Anyway, a 60GB SF-1200 drive is one of the ones being tested in the big SSD endurance thread on XS. It has 280TB written to it so far and is still chugging along. A 120GB SF-1200 has twice as much flash to spread wear over which means it should be able to handle at least 560TB of writes. At the rate of 3.3TB per year, that's 170 years to reach that amount of writes. Also note that all of the drives being tested have lasted far longer than the MWI suggests. It seems this is a worst-case scenario for write exhaustion, and in real-world usage the flash can handle far more P/E cycles than the manufacturer rates it for, which is what MWI appears to be based on. The Samsung 470, for example, was able to handle nearly 40k P/E cycles on 34nm flash that is only rated for 5k I believe.

http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

Also all the testing suggests MWI is pretty linear. Except with the SF-1200 drive being tested something weird happens, it decreased linearly between 100-10% but now seems to be stuck at 10%. Maybe once it gets to this point it starts using up some of the reserve flash or something to extend the lifetime of the drive? Or maybe 10%=0% on SandForce drives? Don't know, maybe somebody else has some insight on this. Might have been covered somewhere in the XS thread as well and I just missed it. The thread is big and there's a lot going on in it, even to miss stuff like that.

 

[reynoldsjrmy]

Thanks 'flakes,

Excellent information - I'll have a read of the thread over at xtreme on a rainy afternoon.

It looks as if all my SSDs are going to outlive me then

Regds, JR

 

[gevorg]

I can just image the myriad of bugs these OCZ drives will have.

"Thank you for installing Windows 8. You have reached 6 of your 5 writes. Goodbye!"

 

[Voo]

I can just image the myriad of bugs these OCZ drives will have.

"Thank you for installing Windows 8. You have reached 6 of your 5 writes. Goodbye!"

Considering that the firmware doesn't have much to do with what flash is used, that doesn't make much sense. Case in point: Afaik there's no difference wrt to stability/compatibility between the different fw versions that's available now because of different flash..

b2t: 1k write cycles for cheap SSDs? More than enough for the vast majority of users. Just do the math for a 60gb drive with 1k write cycles and a WA of 5 (that's a reasonable upper bound for basically any desktop usage and non perfect wear leveling algorithms)..

 

[gevorg]

Considering that the firmware doesn't have much to do with what flash is used, that doesn't make much sense. Case in point: Afaik there's no difference wrt to stability/compatibility between the different fw versions that's available now because of different flash..

b2t: 1k write cycles for cheap SSDs? More than enough for the vast majority of users. Just do the math for a 60gb drive with 1k write cycles and a WA of 5 (that's a reasonable upper bound for basically any desktop usage and non perfect wear leveling algorithms)..

you get what you pay for

 

[Itchrelief]

you get what you pay for

huh? if what he said was that it's good enough, then why pay more?

not that I agree or disagree with your position, but you aren't doing well arguing it.

 

[VirtualLarry]

I've posted before about this, but a certain flash-memory data-recovery company, found that empircally, there was a lot of 3LC flash out there on flash drives, that was only stable for 10 write cycles. That's not a lot, and a far cry from 1000.

I agree with the OP, SSDs based on 3LC are basically cheap disposable garbage. I expect to see them sold by cheap dollar stores for $20.

I know from personal experience that 3LC is crap. I remember reading on these forums a post from someone that claimed to have firsthand knowledge that Sandisk USB flash drives were some of the first to use 3LC. Well, I bought a 32GB recently. and in using it, it hung several times. I finally had to use vconsole.com's USB flash drive tester to write to all cells, which apparently helped to flush out the early-failure cells, because when I reformatted in in Windows, the "total capacity" in sectors had decreased quite a bit.

I suspect that these 3LC SSDs will be "ever shrinking" as they are used, and cells quickly wear out. Not a promising technology, it could set the reputation of SSDs back a few years. (Not like OCZ's countless bugs haven't alreasdy done that.)

 

[frostedflakes]

I'd assume like 2-bit MLC, there are different quality grades. Presumably the highest grade 3-bit MLC will be put in SSDs, and the junk will continue to be used in cheap thumb drives and stuff like that. Also the error correction on an advanced SSD controllers is probably quite a bit more sophisticated than the algorithms used by devices like thumb drives. Poor wear leveling compared to SSDs might also contribute to the relatively poor write endurance seen on flash drives.

http://www.recovermyflashdrive.com/articles/29F64G08CFAAA-29F64G08FAMCI-3Bit-MLC-high-failure-rate

Notice that they also claim 2-bit MLC becomes unreliable after only 1k P/E cycles, whereas empirical testing shows that the MLC in a good SSD can handle quite a bit more. Controllers and flash memory used in SSDs tends to be a lot better than the cheaper controllers and flash used in $20 thumb drives.

So I have no reason currently not to believe the 1k P/E claim by OCZ. If they can achieve that kind of write endurance, I think 3-bit MLC SSDs will be a good option for users. I wouldn't want to see 2-bit MLC SSDs go away, but it would be nice to have the choice between the two. For example a cheaper 3-bit MLC SSD would be perfect for a netbooks/ultraportable IMO, something that isn't your primary system and wouldn't have as many writes as your primary desktop may. I had an Intel 320 in an X120e and only wrote like 2-3GB/day to it on average compared to about 9-10GB/day with the SSD in my desktop.

 

[Dadofamunky]

I've posted before about this, but a certain flash-memory data-recovery company, found that empircally, there was a lot of 3LC flash out there on flash drives, that was only stable for 10 write cycles. That's not a lot, and a far cry from 1000.

I agree with the OP, SSDs based on 3LC are basically cheap disposable garbage. I expect to see them sold by cheap dollar stores for $20.

I bet a lot of that is on thumb drives. Can't understand anyone making internal drives with that level of quality. Not even OCZ.

SSDs, when they work well, are a significant performance booster and a significant advance in storage technology, the first one we've had a in a long time. I doubt we'll see more than incremental changes in that area again for quite a time.

 

[Elixer]

The problem with all these SSDs is that nobody really knows what you are getting, since if you open the SSD to look, you void the warranty.

1k writes is pretty much bottom of the barrel, I expect those to have a 1 year warranty max, but still, it is faster than a HD, so for a OS only drive, it will work out pretty well for them.

 

[groberts101]

some also seem to be forgetting that a cheaper 512GB SSD using 3bit archetecture will not be much shorter lived than a 64GB 2bit drive with 5000 PE/c ratings(if at all). Capacity alone will do much to offset these nand PE/c losses.

Plus, pretty much everything that frostedflakes say above is true and newer firmware/improved ECC/larger amounts of OP reserves will help make up for lower PE/c nand. Why pay big bucks for an SSD that can last for 5-10 years under heavier write loads.. for a drive that is going to be used primarily for read only storage duty? May as well put an SLC drive for your OS volume then.. as it would be similar in lifespan, no doubt. In the end.. these drives can help bridge the gap between HDD and SSD based storage if looked at correctly.

Not to mention.. none of these damned things are wearing out anywhere near the advertised PE/c numbers anyways. You don't think that suddenly 1000 PE/c nand will only write to 1000 PE/c do you? That would be laughable speculation, all on its own.

 

[LOL_Wut_Axel]

I wouldn't touch an OCZ SSD with a 10-foot pole. Not for now, at least.

 

[groberts101]

I wouldn't touch an OCZ SSD with a 10-foot pole. Not for now, at least.

lol.. I "touch" 16 of them on an almost daily basis. Although.. I don't use "a pole" to do it.

as long as they work and remain faster than the ones you guys bought?.. I'm happy enough. :biggrin:

 

[LOL_Wut_Axel]

lol.. I "touch" 16 of them on an almost daily basis. Although.. I don't use "a pole" to do it.

as long as they work and remain faster than the ones you guys bought?.. I'm happy enough. :biggrin:

They don't on many occasions. That's the problem.

Look at all the issues the OCZ 3 Series (Solid, Agility, Vertex) is having. Let's not even mention the Vertex Plus. Reliability>>>>>>>>>>>>>>>Speed, and that's where OCZ falls flat on its face. A $10-15 difference in comparison to a Crucial M4 isn't gonna be enough to sway me towards it if it's shown to be a lot less reliable.

OCZ has been a 'meh' company for as long as I can remember. It's the perfect word to describe most of what they've done in the past five years, including their memory, power supplies (StealthXStream and ModXStream), and now their SSDs.

 

[aphelion02]

To me OCZ is a brand that consistently scraps the bottom of the barrel in terms of reliability for a few cheap bucks. It'd be interesting to see what they make of this, but I won't be buying their products again anytime soon.

 

[groberts101]

They don't on many occasions. That's the problem.

Look at all the issues the OCZ 3 Series (Solid, Agility, Vertex) is having. Let's not even mention the Vertex Plus. Reliability>>>>>>>>>>>>>>>Speed, and that's where OCZ falls flat on its face. A $10-15 difference in comparison to a Crucial M4 isn't gonna be enough to sway me towards it if it's shown to be a lot less reliable.

OCZ has been a 'meh' company for as long as I can remember. It's the perfect word to describe most of what they've done in the past five years, including their memory, power supplies (StealthXStream and ModXStream), and now their SSDs.

I can see the reasoning for being cautious. But after buying, testing, and returning(eventually settling on final purchase of 16) more than 40 OCZ SSD's?.. I will say that there is some massive confusion as to the actual cause of these "failures". Far too much lumping of every issue known to man going on here.

Consider that a "faulty Vertex 3 SSD" can be swapped to another mobo.. have another OS version installed.. a different driver or even bios revision.. and find 100% stability all of a sudden/.. as the result of more variables than most will ever understand(or admit to) and it's pretty easy to see the reasons for this confusion.


That being said.. at the end of the day.. the one's(OCZ SSD's) that really matter to me are the 16(21 total when including family/friends) that I still have running perfectly from day 1(and I run many in raid too.. so there's that). So if you're saying I should look down on a company because of everyone else's issues?.. not gonna happen. Who the hell in their right mind would do that with so much consistent success? I mean.. seriously.

I read it.. I study it.. and learn how to effectively implement them on systems that I have researched to be compatible and go on with my life feeling sorry for those that are not so fortunate as most anyone else would do. And "I must be doing something right" is all I can really say based on personal experience. I know it sure ain't luck.. cause mine's never that good with anything else in life. lol

So keep on speculating what you feel that you must around here to "protect others from making OZC based mistakes"?.. and I'll keep on posting of the successes that I have with them. Don't know about anyone else here.. but firsthand experience's with a product will always trump speculatory flame posts. Many of which come from folks who have never even owned the products being discussed. :whiste:

 

[LOL_Wut_Axel]

I can see the reasoning for being cautious. But after buying, testing, and returning(eventually settling on final purchase of 16) more than 40 OCZ SSD's?.. I will say that there is some massive confusion as to the actual cause of these "failures". Far too much lumping of every issue known to man going on here.

Consider that a "faulty Vertex 3 SSD" can be swapped to another mobo.. have another OS version installed.. a different driver or even bios revision.. and find 100% stability all of a sudden/.. as the result of more variables than most will ever understand(or admit to) and it's pretty easy to see the reasons for this confusion.


That being said.. at the end of the day.. the one's(OCZ SSD's) that really matter to me are the 16(21 total when including family/friends) that I still have running perfectly from day 1(and I run many in raid too.. so there's that). So if you're saying I should look down on a company because of everyone else's issues?.. not gonna happen. Who the hell in their right mind would do that with so much consistent success? I mean.. seriously.

I read it.. I study it.. and learn how to effectively implement them on systems that I have researched to be compatible and go on with my life feeling sorry for those that are not so fortunate as most anyone else would do. And "I must be doing something right" is all I can really say based on personal experience. I know it sure ain't luck.. cause mine's never that good with anything else in life. lol

So keep on speculating what you feel that you must around here to "protect others from making OZC based mistakes"?.. and I'll keep on posting of the successes that I have with them. Don't know about anyone else here.. but firsthand experience's with a product will always trump speculatory flame posts. Many of which come from folks who have never even owned the products being discussed. :whiste:

Yeah, yeah, sure. OCZ has no problems. 50% of the reviews on all websites are lying, as are forum posters talking about their issues even though OCZ themselves recognized them. Sure...

OCZ SSDs are unreliable junk in comparison to most everything Crucial, Intel, Samsung, and Kingston have put out. Until they fix all their issues I'm not gonna be getting anything with a SandForce controller. I'd rather get something that actually works for the vast majority of people and go with a controller by Marvell, Intel, Samsung, or Toshiba. I'd advise everyone else do the same; we shouldn't be Beta testing for OCZ/SandForce.

It's your testimony against that of the other half of users. Besides, why the hell would you buy them anyway? They're only $10-15 cheaper.