Storage Urm.. Kingston, are you sure? (Now a discussion about OCZ SSDs)

http://www.digitimes.com/news/a20110210PD216.html

 

It could be worse.

Look at OCZ.

Intel; Toshiba and other players in the SSD market have decided to postpone the market launch of the new smaller (25nm) cells for SSD till they mitigate the problem with the durability.
The producers are in doubt whether they will be able to keep the warranty time.
34nm-5000 write cycles
25nm-max. 3000 cycles
SandForce drives are less heavy (better distribution) on the durability issue+more reserve cells should enable to launch 25nm products.

Still the producers are hesitating to shower the market till they further improve the whole package.

Only OCZ decided to quietly replace the 34nm with 25nm in their Vertex 2 drives; they kept the final price and pocketed the 50% price difference of NAND flash.

 

Yea reading OCZ's forums and they are littered with people having drive failures. The whole industry have been caught with their pants down; there's a HUGE 34nm shortage on the market right now as everyone was expecting a transition to 25nm. You literally can't buy 34nm NAND AT ALL.

When OCZ has just told all its investors that it will drop memory like a sack of bricks and put all its resources into SSDs, clearly the company is well and truly up **** creek. Apparently it ought a LOAD of 25nm stuff and basically it has to shift it.

 

All I can say there is ooops for OCZ

 

OCZ's doing is on the verge of customer fraud.
They are invited to sell 25nm flash when they like at any price they like; just write it on the package, clearly and plainly.
It's up to the customer to decide what, why and for how much.
OCZ took an excellent Vertex2 drive and silently changed the internals, leaving the package-box unchanged.

"aparently it ought a LOAD of 25nm stuff and basically it has to shift it"
They could fry the chips in palm oil!!
Ups, there is a shortage of that right now. Well, I can contribute a packet or two of coconut butter from my fridge. Free of charge!
That should turn the things crispy


"You literally can't buy 34nm NAND AT ALL."
What do you mean???
I'm constructing a new PC! (was, to be precise, till this SB flaw) anyway, I'll need a nice 120GB SSD till April! (possibly 34nm)
Intel said it's launching 34nm G3 drives with a hyper quick controller this month! ?????

 

OMG i just got an OCZ. (well its on order now cancelled hope i got there before it started its jorney out to me)

FFS where was the review on this Sandy Bridge is up there in lights while this is just here in the forums I missed this and now have a OCZ is 120gb on its way is there anyway of telling if mine is 25nm or 34nm.


120GB OCZSSD2-2VTXE120G VTX2E.

Crucial havent done this with the have they would look stupid canceling two orders in two days.

128GB Crucial RealSSD C300

 

Ummmm... Well, the difference with the OCZ forums is they don't filter or remove threads (other than having a ban on discussing competitors' products) & that's how they work their tech support -> if you have an issue, you post on the board & people advise &/or an RMA is arranged.

This means that you have a self-selecting set of people who are going there with problems which, on one hand, 'could' give a false perception of there being a high failure rate but, on the other, i would suggest actually shows that they are open with issues that individuals have - not least as it informs other users as to potential solutions.

So, whilst you're never going to get 100% reliability with any product, you have to look at the source of the data in order to make responsible claims about it - which you've clearly not done (for the reason stated).

it'd be like going to a football match, asking 200 people if they like football, & then making the assumption that, as the vast majority said "yes" that this was a representative percentage which could be carried across to everyone in the country - a very irresponsible & clearly biased way to use statistics.

Well, if you look at the indilinx bit of the forum (where obviously none of their products used 25nm nand as it wasn't available) you will see exactly the same pattern of there being a (seemingly) large number of threads with people having problems, but it's for exactly the same reason that that's how you sort(ed) out an issue with an OCZ indilinx SSD...

...& OCZ obviously didn't go bankrupt in having to replace all the indilinx drives as they clearly didn't fail in significant quantities because, again, it's a self-selecting group...


Then, as you 'should' know (being bittech staff then writing with 'authority'), 'if' all of the nand in a SSD were to use up all of it's r-e-w cycles, the actual symptom would be that the drive would become read only however, having read vast numbers of forum threads, i have never seen a single case where this has actually happened.

instead, all of the issues that have been failures (ie aren't down to errors on the part of the user or some incompatibility or other) are down to the controller itself failing, the f/w corrupting or the controller going into 'panic mode' - between the indilinx & the SFs there are some solutions that have been resolved with f/w updates (so issues that some people reported now don't reoccur), some that have solutions which don't require an RMA, & others that do need replacing; as is the case with most tech products.

So, imho, to randomly suggest (well, you're talking about nand & OCZ's forum) that the issue is down to the nand that's used when there is no evidence to back this up is somewhat dubious 'commentary' to say the least.


&, of course, you all appear to be forgetting that the nand cycles are based upon a fraction of a percentage failing within that lifespan - not that, as soon as x,000 r-e-w cycles happen then all the cells will become read only...

...though at least Synalar correctly pointed out that the SFs write to significantly less of the nand than other controllers, so are more able to maintain longevity despite a reduction in cycle ratings with nand shrinkage.


Now, that's not to say that, because they've recently added in the ability to use 25nm nand in the V2 f/w that i think it's justified to price at the same level when the nand itself is that much cheaper, but that's a completely separate issue.

Well, similarly, part of the premise of the SF controller was that, as it didn't need a DDR cache, the SSDs would be cheaper, but neither OCZ nor Corsair nor whoever else appeared to pass on the cost saving to the consumer.


Anyway, it's not that i have shares in OCZ or anything, so it doesn't matter to me one iota if anyone actually chooses to buy things from them or not, but for people not to do so based upon shonky statements & inferences that simply cannot be justified would be a shame.


[Edit]

it's just crossed my mind that i have pointed out what the OCZ forum is about before - a quick search & this is one occasion from October last year.

 

Happy with my Crucial C300. I still think they are the best for that price range.

 

OK first part: It's not fraud, OCZ can do what they like with the contents as long as it's warranted and the user experience is the same. I know 3000 writes is 60% of 5000, but they could change the firmware and it might have the same effect. Sony removed features from its PS3 via firmware - it's still a PS3. Intel generates a new spin of its chip, it's still the same CPU...

Unfortunately companies are not required to post the exact technical details of their parts - same as gfx card makers cheapening down their hardware over time, but at the cost of overclocking and/or life, maybe. It has happened.

WRT 34nm: there's no chips left on the open market. Unless you're Intel/Micron or Samsung making them, you can't buy them.

 

Up and about on Saturday 6:50?
http://www.youtube.com/watch?v=LuqEbRzy_t8

It's like reducing 100grams of chips to 60g and stating that it's the same, since Kcal/package value does not change-"we just added more grease". Right!
http://www.lorenz-snackworld.de/content/en/01brands/010crunchips/index.php?isPrintable=1&noFlash=1

So, the best thing is to go for the G3 of Intel SSDs at the end of this month.

What happened to other producers of 34nm NAND? They dismantled the old production lines to accommodate the new?

 

The G3 will be 25nm NAND too. Whether Intel is better than Samsung, I don't know, but I suspect Intel/Micron was out first.

Who does OCZ currently use in their rehashed Vertex 2E?

Synalar - a bit of that and everyone stocked up on 34nm once they saw 25nm was ****. They can't just change production overnight when they are invested on moving forward.

 

True; apparently there is going to be a G2.5 before it.


EMCREST- Intel readying new 510 series of SSD drives for February release
Intel continues to rely on 34nm MLC NAND for the SATA 6 Gbps drives, in 120GB and 250GB storage capacities.
The new 510 series is designed for the “enthusiast consumer client PC.” Read speeds on the new drives will leap to 450MB/s (from the X25-M’s 250MB/s) and write speeds will accelerate to 300MB/s (compared to the X25-M’s 100MB/s).
A 120GB drive will cost $279, while a 250GB version will set you back $579.
Sequential read speeds should match the new SandForce2000 controller.


The G3, 25nm
The new “320 series” with a “G3″ at the end of the model-strings would be the highly anticipated X25-M successors. A reduced-sized 310 SSD appeared in December, making the 320 name appropriate for the 2.5-inch models.

Today’s X25-M capacities; 40, 80, 120 and 160 GB are all bumped to 25nm NAND, and the series is extended with 300GB and 600GB models. According to previous info, Intel’s third-generation SSDs also brings lots of new features to the table, including hardware AES encryption and improved life span for the NAND. Performance improvements are also offered, particularly in 4K random writes, but also in sequential writes. (I think OCZ has got a wagon-load of Intel's NAND in the warehouse)

http://www.fastestssd.com/wp-content/uploads/intel-320.jpg

 

From a quick check OCZ 'normally' use Intel/Micron nand - though they have stated in forum posts over a long period that 'if' there is the opportunity to get higher r-e-w cycle nand that has the longevity & whatnot that they need then they will use it.

Now, the only way for a consumer to know what brand of nand is used appears to be to open the SSD up - though this voids your warranty...


Whilst i agree that OCZ 'should' put a label on the packaging or whatever (though no manufacturer did this when the shrink from 50 to 34nm nand occurred in the middle of the indilinx, etc G2 drives) they're basically saying that, because intel/Miron are hanging on to whatever 34nm nand they have for intels drives, either they use the 25nm or they close up shop until the 6Gb/s drives are out there.


From a quick scan read, to tell whether you've got 34(+) or 25nm nand, on the sub 180GB drives (logically it appears that they're keeping the remaining 34nm nand they have for the 180GB+ ones), you can do a google search for the formatted capacities & 'if' the formatted capacity is ~4GB less then you've got the latter.

[NB this is d.t. the RAISE tech -

"RAISE requires 1 unit of the Nand IC, on 34nm this unit was 32Gb or 4GB over the whole drive. Now you have 25nm that is 64Gb or 8GB over the whole drive."]


Otherwise, on the sub 100GB drives, there is a decrease in sequential compressed write speeds on the 25nm nand versions.

This is simply because 25nm nand is slower & so, as the lower sized drives use fewer nand channels, that's life.

& for the 6Gb/s drives, there will simply be an increase in the number of nand channels to both compensate for this &, obviously, to get to the <=6Gb/s speeds.


Yeah, this is all the info i can find about the situation & (hopefully) shows that i'll point out the sides of OCZ that aren't as great for the consumer, rather than just randomly defending them...

...well, the point of my earlier post wasn't to say that it was right to reduce nand specs without notifying (other manufacturers doing the same isn't an excuse imho) or that there shouldn't be a significant reduction in price d.t. using cheaper nand - simply that what was being given as fact & the inferences that were made didn't stand up.


Otherwise, excluding brands which make their own nand so can retain stocks, are we actually sure that this is only an issue (albeit a temporary one) with OCZ?

Simply that, as OCZ are allowing open discussion on the forum then it's apparent that this is where things stand with them, but can we assume that the silence by the other manufacturers is actually d.t. them having all the 34nm nand in the world or that they're just keeping quiet about it?

Yeah, forgetting anything else, it is a reason why i like OCZ as everything is open for discussion.


[Edit]

Oh, & in looking for specific info i'd missed something interesting -

OCZ are offering a part exchange for 25nm drives for 34nm ones, whereby you pay the difference in cost between the nand...

...i've no idea how much this is (as my V2s obviously pre-date), but it does give the consumer the option to pay a premium to get the better nand on the 60/120 E drives.

 

just a selected few flower power extractions from OCZ's customer support forum:

A.
Posted by OCZ:
"if a retailer feels they can make a few extra points from a customer they will, regardless if the drives are faster/ slower or the same, its called business."

Customer's answer:
Look how he calls dollars points. People, you got screwed but they won't tell you that. That's called business.


B.
OCZ
As the industry transitions to the 2Xnm process, OCZ is notably the first to market with this
technology that aggressively reduces the costs of our SSDs. Most of our
customers are aware that as the NAND flash technology process nodes are shrunk, the price of
NAND comes down substantially. OCZ continues to focus on delivering the highest performance
and highest reliability drives available at a lower price point…paving the way for SSDs to
become more accessible to the complete range of consumers and to take the place of
traditional mechanical hard drives over the next few years.

Customer:
What I don't understand is I bought a 25nm 120gig Vertex 2 unknowingly for the same price as you sell the 3xnm model. Those of us who bought a Vertex 2 near the end of January were not aware of the differences. It wasn't advertised on the packaging nor on the websites of the distributors you use. You change your products without public notice, making them slower and smaller - and now I should pay additional money to get what I expected (35nm) ?!

OCZ:
SF2000 will use 25nm just fine, it has double the amount of nand channels over SF1200 so speed even at sata 2.0 speeds should not be an issue.

Please when you are discussing SSD first research how SSD work, Over provisioning, RAISE or similar technologies, Nand used on those SSD's and PE/c / Density and then lastly the price Vs GB ratio...this GB Vs price ratio really is something dreamt up by reviewers who have placed it high on the 'must read" list but in reality it counters how SSD work and ultimately can cause issues. OCZ will produce drives which the market are pushing for, if the market wants 64GB drives with No OP we could make them....even though many may hate them. If you want a fast SSD be aware of what makes SSD fast and choose a drive which has those features...it may cost more though.

Present 25nm drives

The 60GB 32Gbit drives feature 16x32Gbit (4GB) Nand IC with 7% OP
The 60GB (55GB formatted) drive features 8x64Gbit Nand IC with 7% OP
The 100GB 64Gbit drive features 16x64Gbit (8GB) Nand IC with around 25% OP
The 120GB 64Gbit drive features 16x64Gbit (8GB) Nand IC with around 7% OP
The 50GB drives I will need to double check.

25/28% OP drives recover faster than 7% OP drives, you can use them nearer to full capacity without them slowing down.

What is the fastest drive?...the 100GB model, what is the best selling drive the 60GB (the one you all hate) as it has the best price point, the 60GBx16 is more expensive so was ignored by retailers so we pushed it for enterprise use.

 

i'm not 100% sure what your point is Synalar?

Are you agreeing with me that OCZ do allow open discussion on their forum or that they're being underhanded in some way?


Well, A is simply that they've (slightly randomly) used "points" rather than "pounds" or "dollars" or "yen" or whatever - though they're blaming the retailers for keeping the price the same (i have no idea if this is true).

Though obviously, since there's a general agreement within this thread that the actual cost for 34nm nand has sky rocketed (unless you're someone like intel & effectively make your own), the other option, other than not making any drives for a couple of months, would have been to drastically increase prices.


B, within the context of A, it isn't that OCZ are selling them for a price to the consumer but that the retailers are...

...&, as pointed out in my edit, 'if' a consumer is prepared to pay the premium for 34nm nand, they can trade in a 25nm one (which have only been available since Jan afaik) for the differential cost...


& C is -

(i) simply reinforcing what i've said for a long time on here about OP, &

(ii) as is the same with all SSDs of varying capacities (unless the manufacturer deliberately slows them down), that the smaller drives with less OP will tend to be slightly slower than the larger drives d.t. fewer nand channels.

[NB obviously this post is talking about the 25nm ones, but my use of 4x 50/60GB V2s in a single R0 array was accepting that, whilst they are individually slightly slower than a 100GB one for (almost exclusively) compressed writes, the the speed gains by having 4 smaller ones vs 2 larger ones vastly outweighed the difference... ...there only being so much that i could afford to spend.]


[Edit]

Oh, & whilst it's not an answer at all to this, imho anyone looking at buying a SSD would be damn foolish to spend money now on any of the current models.

Well, we're looking at <=2 months before at least the shonky (unless they've learnt from the P3 & improved them) C400, the Corsair P3 & the V3 will be out.

Now this means that either you'll be able to pick up the current models that much cheaper or, whether or not you actually have a 6Gb/s sata disk controller, the new models will be dramatically quicker...

...obviously, on a 3Gb/s controller, you wont get the max sequential speeds from a 6Gb/s SSD d.t. the bandwidth limitation, but all of the smaller r/ws & the iops will dramatically improve.


The same holds true with the pcie & HDSL drives, as faster ones are also going to be appearing from OCZ around the same time.

 

PR I just received:

 

Not 100% sure why it's amusing...

Well, haven't Crucial already said much the same thing about using 25nm nand in the C400???

Okay it was actually Micron (who make the controller) -

"Despite using IMFT’s 25nm NAND (at roughly 3,000 program/erase cycles per NAND cell), Micron claims the C400 will last for at least 72TB of writes over its lifetime which is identical to how long Micron claims the C300 will last. I asked Micron why the C400 didn’t incur any endurance penalty with the switch to 25nm, its answer was simple: the C300 was conservatively rated. Even though the C300 NAND cells will last longer, both it and the C400 will at least make it to Micron’s 72TB rating. Micron also mentioned to me that it writes more than 72TB of data to its drives during development, so this is a verified rating."

- & they of course don't have the advantage of being able to use fewer cells per write by on the fly compression which the SFs do.


This 'may' be why the Crucial P3 is still using 34nm nand, & separately would help to explain the better specs than the C400.

it'll be interesting to see if the C400 & the 25nm V3 come in significantly cheaper (as they obviously should do) or if some profiteering goes on.

 

Crucial are the commercial arm of Micron. Marvell makes the C400 controller - as it did the C300

What's funny is that OCZ is having a world of issues with its 25nm and even Intel and Micron have put back their roll out of new SSDs because of it.

I worry that if this is the trend for NAND as it continues to shrink, what are they going to do when 1xnm arrives??

 

Sorry - middle of the night & wasn't thinking about who Micron are properly... Of course you're correct on that - my bad.


Anyway, other than the 'somewhat' poor decision to not clearly label the 25nm V2s as being so (as they're slower than the 34nm V2s for (esp) writes then it 'should' be stated) & not reducing the cost, where is this 'world of issues'?

Well, the change in nand has not made drives fail or unusable or anything so...?


Now, i really don't see what impact this has on the 6Gb/s drives - well, sticking with OCZ, whilst the 1x00 controller drives are obviously going to be slower if you use slower nand, the 2x00 controllers that are going to be used in the 6Gb/s ones are being built around using 25nm nand in the consumer market so whatever speeds they're aiming for aren't going to be magically compromised.

Well, if you know that there's a limitation in nand speed d.t. the reduction in size then you add extra channels to compensate - obviously simple to do when creating a product from scratch rather than when using it in an existing product as a stop gap measure for a couple of months.

Hence my earlier point that anyone would be foolish to buy a SSD now from any manufacturer when new & superior (at least in terms of speed) consumer products are a stone's throw away.


& of course there's a choice, as there is now - well, we could all have refused to use MLC nand d.t. its lower lifespan &, as the market wouldn't have developed to anything like the same level (economies of scale & the allowance for new entrants into an expanding market & all), almost none of us would have SSDs...

So, with the V3s, (according to the press release you quoted) it'll be possible to choose 34nm or 25nm nand, in the same way as it'll be possible to choose SLC, eMLC or MLC nand, or SATA, SAS, pcie or HDSL interface SSDs as options.

Thus, there will be a differentiated marketplace & you accept that there's compromises to be made if you want (or can only afford) the cheapest option.


Again though, none of this means that there's an inherent flaw in using 25nm nand or that further die shrinkage to 18nm nand (or further) can't happen...

...though i'd imagine that it'll be down to turning to eMLC as that has ~3x the cycles of normal MLC to make viable products - though there would be a slow down in write speeds as a result (that's the 'cost' of eMLC vs the 'benefit' of longevity), but unless another solution can be found or more channels are used then such is life.

(there's also eSLC which i believe has ~3x that of SLC, though i'm not sure that's coming to market this year as there's not yet the same demands on longevity for SLC yet.)

i guess the other option would be to reduce the cost to the point that they were short lived & disposable, though it wouldn't exactly be a great premise to work under.

 

OCZ trumpets successful transition to 25nm SSDs, consumers disagree
mass media wide

With its memory business cast aside, OCZ is pouring all its efforts into the SSD market and has today proclaimed itself as "the first SSD manufacturer to successfully complete the transition to 2Xnm NAND flash-based storage solutions".

few important details missing from OCZ's press release:

-the new 25nm models are in fact smaller in more ways than one
Despite outwardly appearing identical, a 25nm drive will provide consumers with less capacity than a 32nm model and performance

A.provisioning-by doing so, the available drive capacity decreases significantly
new high-density 25nm NAND flash is cheaper to produce but offers less write-cycle endurance

B.the new drives have fewer modules interfacing with the SandForce controller and should, in theory, benchmark slower than their 32Gbit, 32nm counterparts.

As expected, the confusion has led to an uproar from OCZ's customers, many of whom are disgruntled by the fact that OCZ is marketing the seemingly-inferior new drives with the same model numbers, capacities and speeds as the superior 32nm Vertex 2.
60Gb that have in reality 50Gb are sold for the full price (no telling the number of Gb after 5 years).

As a response to the crowd readying to burn OCZ to the ground, OCZ introduced a program through which customers can trade in their 64 Gbit die-based drive (25nm) and receive a credit toward the more expensive 32 Gbit die-based drives (32nm). Trouble is, it's currently difficult to tell which version you've purchased without opening the drive and voiding the warranty in the process.