Next generation of SSDs to have a shorter lifespan?

Well DDR3 has quite low memory requirements, especially at lower clocks/speeds. Remember, Intel X25-M is still using PC133 SDRAM (not even DDR) in their SSDs; so performance is not paramount here but rather capacity of the DRAM. The G1 even used 166MHz SDRAM is i recall correctly, so the G2 has slower DRAM but is faster. ;-)

1GB may be used as write-back so the first 1GB writes 'for free' at maximum wire-speed. Would also be very beneficial for random write purposes like databases. Two 4Gb (=1GiB) DRAM low-voltage chips would be perfect for this task.
 
There is only 32MB of DRAM on the G2s. You said 1 GB for new SSDs. There is no way to do that without increasing the power usage a lot, even going to DDR3.
 
Well as i said i think two 4Gb DRAM chips would be feasible for a future-gen 6Gbps SATA SSD controller. That would be quite low-power especially the latest production process and DDR3 is more energy efficient than DDR2 and DDR1, etc.

Right now, the 32MB is enough for the write mapping tables, which it is used for. The 1GB would make sense as write-back but the controller must be designed to use something like that efficiently and without losing any writes upon a power failure.
 
And as I said, it may be technically feasible, but it is not going to happen for consumer SSDs. Adding that much DRAM and super-capacitors will increase the cost and the power consumption too much. Once enterprise SSDs become more common, maybe you will see something like that. But I bet it will cost more than twice that of the equivalent consumer SSD.
 
Again; if 16 DRAM chips consume 0.8W; one would consume only about 0.05W and two will consume 0.1W; that's not outside the margins where SSDs operate; so i don't see how a slight power consumption increase would make this feature so hard to implement.

I agree it likely would be targeted towards businesses rather than consumers. But i don't care; i would buy it if available. :)
 
I think you are far too optimistic. 1 GB of DRAM will consume about 400 mW operating, and that does not include the memory controller.
 
The memory controller is inside the controller-chip itself. 1GB of DDR3-DRAM can be done with just two 4Gb (gigabit) chips.

So i'm not talking about a 1GB memory module with 8 chips on it; i'm talking about 2 chips making a total of 1GB memory, directly coupled to the SSD controller.

I think the power consumption would be even below the 0.1W i anticipated; as the modules do not have to focus on speed but on low-voltage instead; its the storage capacity that counts.
 
Fine, and i'm sorry. But i wasn't the one to start with personal attacks. I'd love to adhere to the technical reasoning without getting such comments. I hope we can just continue in a more civilized manner and stop being offensive?


160GB is the biggest X25-M G2 you can buy now, 600GB will be the biggest X25-M G3 you will be able to buy around christmas this year, at least according to several news sites including Anandtech:
http://www.anandtech.com/show/2928

Ultimately, all pricing are dependent on production volume. Produce one thing of something and the cost is 100 million. Produce 100 million of something and the cost is 1000. Produce 100 billion and the cost is 100. Not that these values are accurate in any way, just to make the point here.

If Intel/Micron increases NAND production as they are planning, that leads to more products being sold, either via retail but also via OEM channels. Think netbooks with only SSDs produced in mass volume. Once a 'turning point' is made, the NAND market can grow very rapidly. In that scenario, it would not only be Intel/Micron producing at large scale, but other foundries would be keen on investing in this growing market with great revenues; the demand would be higher than the supply for years to come. This would warrant huge investments into high-tech fabs, even more than today.

Essentially, the HDD market is being (partially) absorbed into NAND market. The benefit here is that NAND products will mature rapidly and come down in cost.

At least that's the theory - with NAND prices not dropping the only real price cut we got was when going from 55nm to 34nm; the Intel G2 was significantly cheaper. A same price-cut will happen with 25nm products. Also, it could be possible that Intel would choose an agressive pricing plan, relately even lower than when extrapolating the price cut to 25nm. The advantage there would be for Intel to seize a much higher market share because they are the first to really ramp up their NAND production. With the higher market share they would have an edge a few years later when more foundries begin ramping up 2xnm 2-bit MLC.

Of course, this is all speculation. But i understand you're not against that all all. If some of it is true, then pricing could drop over time more than we've seen thus far.


Probably useless as even the 5-channel X25-V is already at the limits of SATA/300. Once their 6Gbps controller is ready to ship with the 25nm NAND it would make a 600GB SSD much more feasible.

If they had released the 320GB G2, probably it would not be sold alot and thus not profitable as another production line would increase operating cost and thus it might not be profitable at all. With the 80GB G2 having chips on one side, 160GB G2 on both sides, the 320GB would need either multiple PCB or double capacity NAND. So i think Intel made the choice to skip the 320GB for now due to their own wallet.


I'm not so interested in comparing prices - the topic was 'next generation SSDs to have shorter lifespan'. In that context, you could say the next generation SSDs to have both shorter lifespan and higher lifespan than the current generation, depending on the size of disks you choose.

But we can see all capacities double:
Intel X25-V 40GB goes to X25-V 80GB
Intel X25-M 80GB goes to X25-M 160GB
Intel X25-M 160GB goes to X25-M 300GB

I would think after intoduction price, the pricing would settle below their older 34nm cousin. So a 300GB 25nm Intel would be cheaper than the 160GB 34nm. Not straight away though, it took its time before G2 actually became available and pricing dropped due to more actual competition instead of competition on paper due to low stocks.

If this is at least partly true, you would get the 600GB for a very reasonable price. Would be the first real Gaming SSD i think.


True - i wasn't taking that into account with my calculations. Though, this does not have to scale with the size of the drive on a 1:1 basis. For example, if you have a 600GB SSD as system disk (excellent for Gamers) you would probably not try hard to disable the pagefile, hibernation file, system restore, indexing and whatever not to save space on the 'tiny' system drive.

As games would not write randomly alot (mostly sequential writes during install) it also wouldn't wear a drive as much as the random writes issued by the Operating System.

In other words, i think the new generation will work excellent for the purpose they were designed for. If you need more write endurance, then you reserve more space. Though not official, rumors exist that the new X25-E "extreme" generation will have up to 50% capacity reserved for internal use, but now uses 25nm MLC instead of SLC process.

If that is not enough, you can simply RAID0 or several disks to get more write endurance. This may be useful in database servers which see a high number of random writes. With each disk added, you increase the total write endurance.


Yes, you can extrapolate. But that carries the danger of being naive (how ironic; isn't that?) - progress is more than linear increase of 'raw power' like more 'mhz' or more 'rpm' or more 'cells' - it can also mean progress by doing things a little differently. That can mean a little trick to make NAND more write endurable, some intelligence in the firmware that causes lower write amplification or even a new fundamental way to actually store bits - such as with Phase-Change memory. The consumer probably wouldn't notice; its just an SSD with Serial ATA like before. I can still see memory chips - how they work is actually very different.

A good counterargument would be that this is actually cheating; i'm not comparing the same product but actually one technology with another piece of technology. But hey that's exactly what has happened when we look at history. We don't have single-core CPUs anymore, we don't scale linearly with MHz anymore, actually MHz have become much less important.


Point well made. For the moment, i think you're right though i feel it's also panic for nothing - some posts in this thread could be read as "we can't have reliable/usable SSDs anymore in the future" - which is what i meant with "FUD". Things don't look that bad and honestly some technology adjustment will come and fix our write endurance problem forever. However, it may also introduce new weaknesses inherent to that technology. Looks like this can be a long thread. ;-)

OK, fair enough. Let me take back the naive comment and I did not intend it as a personal attack. I'm sorry for that.

Just want to reiterate why its not completely "fair" IMO to compare the 160gb intel G2 with the 600GB Intel G3....

Intel said the G2 would be available in 320GB max, just like they are saying the G3 will be available in 600GB max, So I think its better to compare the 160GB G2 to the 300GB G3 or the 320GB G2 to the 600GB G3. BThe intel 600GB G3 will probably be around $1100 or maybe $1000 if it is released. let's say $900, that would be a pretty good price.

Do the calcs and the $900 600GB G3 won't really have any more total write endurance then the $400 160GB G2. Just a point, it probably will be good, and more then enough for 99.7% of userbase.ut, if your comparing on a cost basis, then it doesn't matter as the 600GB G3 will be double the cost of the 300GB G3, leaving all those results the same.



I think intel makes a good profit on the G2 drives, however. Why?
  • The 160GB drives were selling like hotcakes for $550 a pop early on.
  • now $380 isn't uncommon and retailers still have to profit, which puts the "Intel price" at mabye $330 or so.
  • On my lappy, the upgrade factory option (CTO) was $550 for two 160gb intel G2's add in the always there Bing Cash Back and it's $247 per 160GB SSD! Knowing this the real "Intel cost" could be as low as $210-$220 per 160GB G2. Which means they might see ~$150 for 160GB or ~$600 for the 600GB G3... meaning we might start finding the drives on ebay and cheap CTO options as low as $700-800 for the 600GB model after a few months, which would be awesome. Make them all in 1.8", too, Intel. That is all I will buy.

$~375 for the 320GB / 300GB model is sounding quite good! See you in Summer of 2011 on that one :D

If this is at least partly true, you would get the 600GB for a very reasonable price. Would be the first real Gaming SSD i think.

Hey, I resent that. My 80GB G2 serves me quite well for gaming in my desktop. Just because a single install of Star Wars Force Unleased eats away at 42% of the total useful storage of the drive doesn't mean I can't do other stuff with it. :D (screw you lucas arts for really dumping this port on PC users. Although it really is a great game once you install the latest update patch. Where the hell would Lucasarts be in the first place without PC gaming?)
Now I just got my laptop which came with 320GB of intel SSD goodness and I don't even know what to do with myself on this machine - I've become so accustomed to limiting myself. :)
 
I am just waiting for

www.anandtech.com

to write an unbais article on these problems on G3. Because you'll lose data as the data sits in the hard drive, I don't care how fast it is, no1 can use it to store data.

I'm willing to buy 1 as the boot drive, because my boot drive is nothing but applications. So if I lose the application, I just re-install it from my original CD. Any forecast on prices for a 300GB?

Here's a solution if you are afraid of losing data:

Root Drive = G3 at 300Gb or higher

data drives = more than 1 x G2 at 160GB ea.
 
Just want to reiterate why its not completely "fair" IMO to compare the 160gb intel G2 with the 600GB Intel G3....

Intel said the G2 would be available in 320GB max, just like they are saying the G3 will be available in 600GB max,

Where is this info come from? Why would intel rel. G2 @ 320GB when they are releasing their G3 soon?
 
More capacity than what? The limiting factor for most people today when it comes to SSD isn't the amount of apace one can fit in a 2.5-drive, but price.

but the high price is also linked to the limited space. 99% of people would never buy a 64gb or even a 128gb os drive unless all you so is surf the net and email, which would pretty much moot the point of needing an ssd.
 
According to Dailytech, Intel will build SSD based on 25nm NAND flash memory. That means SSD will have even shorter life span. Who will purchase these drives?
 
According to Dailytech, Intel will build SSD based on 25nm NAND flash memory. That means SSD will have even shorter life span. Who will purchase these drives?

I don't know. I love my 50nm X25-E. I might buy a 34nm X25-M G2 160GB for another PC shortly but will stay away from gen3 initially. I have no doubt the SSD controller engineers are intelligent and will do the best job possible with trim, write amplification and wear leveling - but they can't beat the laws of physics. Stagnant data will decay much faster under a 25nm process.

Does anyone know how that works today? If I have an SSD and write to it rarely (meaning I do my best to keep writes to a minimum), then it's not really doing a lot of wear-leveling. I assume it's not just moving data around in the background (or is it?) because that would increase wear for no reason (how could it since the drive wouldn't know when I was shutting down the PC?) So what happens to the static data?? What if it sits there for a year without being refreshed? My worry is under a 25nm process, it would just go away.
 
If that would be true Intel could stop working on 25nm altogether and just continue selling their 54/34nm products.

Of course, the reality would be somewhat different, with much higher capacity SSDs (600GB) and lowered price point, actually means you may have more erase blocks and thus a larger total write cycle count. Also we didn't get any numbers of max write iterations on the 25nm IMFT NAND process.

I think it is legitimate to discuss the different trade-offs of new innovations. But to imply future SSDs are useless would be the same as those who said 512K 'ought to be enough' or that CPUs can't be scaled beyond 1GHz.

For example, the controller could maintain integrity by re-writing to weak cells before the actual data has degraded beyond the point of recovery.
 
As long as something is implemented into the firmware to move things around every once in a while while the drive is idle so no data sits on the same block for a year I don't really see much of a problem other than eating write cycles not that one every four to six months matters much.
 
As long as something is implemented into the firmware to move things around every once in a while while the drive is idle so no data sits on the same block for a year I don't really see much of a problem other than eating write cycles not that one every four to six months matters much.

Exactly. That seems like such a simple and elegant solution that could very well work, unless I am missing something. Of course, part of the concern was with data longevity on an unpowered drive, but I for one don't plan on loading an SSD up with data and leaving it on a shelf for a few months or years...at current prices, anyway. I'll admit that if we get to the point where 2TB SSDs only cost $50 then it might be tempting to throw a bunch of data on there for archival purposes since the drives are so small and durable, but if the data is only going to last for a couple of months in that state then it could be a problem. Hopefully they'll find some way to address this, or maybe by that point we'll have a storage medium that fits that purpose better than an SSD would. Tape isn't for everyone, and if you use that method with a hard drive you have to pray that it spins back up in a year when you want to access your data. I think storing data on a huge SSD would be appealing, were it not for the degradation. It'll be interesting to see where this goes as the technology evolves.
 
PCM-based SSDs could be used for archival purposes. The phase-change process behind it is quite stable (just keep it cool enough *g*) and the data could last almost indefinitely.

PCM is currently replacing NOR Flash memory. NAND will be its next target, so who knows :)
 
Cramming more NAND flash into a 3.5" form factor isn't going to do that much to drive the price down though. You generally need to increase the density of the silicon to do that. So those with a $/GB obsession will not be satisfied.

The fact that manufacturers are thinking about new form factors is a good sign though, as this makes it easier to put 2 drives in small laptops, e.g. 30-60GB boot drive buried inside the case somewhere on a SODIMM or mini-PCIe card, and a big mechanical drive in the regular 2.5" bay.
 
I just wish manufacturers would focus more on longevity than space (more SLC products).

I agree. I love my SLC but it was expensive. If they just made more of them the price would go down. They are inherently more reliable than MLC.
 
I guess we have to face the reality that NAND is simply not the solution we all hoped for as it won't overtake standard hard disks.

I think SSDs will totally replace all hard disks for the laptop market, they have about the same life expectancy, don't heat at all, consume very low power, everything you need in a laptop. Plus they are very fast for gaming, so they'll probably eat a fair share of the desktop market.

As for desktops, it's hard to say. People with low storage needs will be perfectly happy with a single 500GB 2.5" SSD drive for both OS and data. A mixed bag of OS SSD and data HDD is just too complex and costly for PC manufacturers. Hybrid SSD/HDD or full SSD 3.5" drives might very well become the norm.

In fact, the decrease in size from 5.25" to 3.5" was to reduce platter vibrations and increase precision and density. SSDs don't care about this, so you could theoretically design 5.25" SSD drives. Not really necessary with the 4-in-1 adapters, I could fit even today 12 x 2.5" SSDs in my 3 x 5.25" free external bays. Except I'd need SATA 48Gb/s or something.
 
I guess we have to face the reality that NAND is simply not the solution we all hoped for as it won't overtake standard hard disks.

I can see memristors take the storage market by storm. They're absolutely tiny compared to a bloated Flash cell, even at 20 nm, thus it should be no issue to get high densities with them. They also don't suffer from significant degradation while writing, are bit-addressable (they're like a special kind of resistor, basically) and will even rival or beat DRAM.

With PCM slaughtering NOR Flash ATM, and PCM and memristors ramping up, it may mean that Flash will be a distant memory in a few years time. I can't wait :)
 
I agree. I love my SLC but it was expensive. If they just made more of them the price would go down. They are inherently more reliable than MLC.

The price of SLC SSDs has little to nothing to do with the number produced and almost everything to due with both the inherent and market costs of SLC flash. SLC flash is simply significantly more expensive than MLC and as such it has less general volume as most devices are costs conscious that use flash (CompactFlash, SD, usb, etc). Even if the volumes of SLC SSDs were equal with MLC SSDs the cost differential between the two would be largely the same.
 
The price of SLC SSDs has little to nothing to do with the number produced and almost everything to due with both the inherent and market costs of SLC flash. SLC flash is simply significantly more expensive than MLC and as such it has less general volume as most devices are costs conscious that use flash (CompactFlash, SD, usb, etc). Even if the volumes of SLC SSDs were equal with MLC SSDs the cost differential between the two would be largely the same.

I disagree with that statement.
It is true that SLC will remain more expensive than MLC.
However, SLC is particularly more expensive today because of its low production.
 
That said, SSDs are plenty usable in their current form and provide lots of benefits.

Very true.

spectrumbx said:
I just wish manufacturers would focus more on longevity than space (more SLC products).

That would be nice, but I think it's fairly safe to say that many (not all) consumers are concerned with having more space over having a drive that will last longer. Lots of people are fine with buying an 80GB SSD today because they know they will probably replace it with a larger, faster drive by the time it dies anyway.

By that same token, many people realize that USB drives have a finite lifespan but I think people would be interested in being able to cram 8GB worth of documents/music/etc. onto a flash drive than having a 4GB drive with space limitations that would last longer.

There is no universal though, and ideally we'd have both space and longevity. This is where I'm hoping SSD manufacturers will get crafty and figure out how to extend the lifespan of large-capacity drives. It sounds like a battle against science/physics, but like someone else said, many people thought we'd have hit the platter density limit by now...yet engineers have had to get creative since the demand for more storage exists and they're figuring out ways to make 2.5"/3.5" HDDs larger since they're basically forced to. NAND isn't a replacement for spinning platters yet, so they can't just call it a day. They have to make it work. I just don't see 4TB hard drives coming out and then us being told "Sorry guys, that's all we can do, better stock up on them or hope the prices on 2TB SSDs comes down quickly because this is all we can do for a conventional hard drive."

I do think the wall will be hit eventually, but I also think they will try to come up with new developments to avoid hitting that wall for as long as they can. When there is no reason to anymore (i.e. it becomes a waste of money because the ROI will be minimal), they'll move on and focus on the next technology.
 
Last edited:
I can see memristors take the storage market by storm. They're absolutely tiny compared to a bloated Flash cell, even at 20 nm, thus it should be no issue to get high densities with them. They also don't suffer from significant degradation while writing, are bit-addressable (they're like a special kind of resistor, basically) and will even rival or beat DRAM.

With PCM slaughtering NOR Flash ATM, and PCM and memristors ramping up, it may mean that Flash will be a distant memory in a few years time. I can't wait :)

I doubt we will see memrister drives in consumer sizes (32GB+) and price for another 5 years.
 
I doubt we will see memrister drives in consumer sizes (32GB+) and price for another 5 years.

Really?

They can also be fashioned into non-volatile solid-state memory, which would allow greater data density than hard drives with access times potentially similar to DRAM, replacing both components.[37] HP prototyped a crossbar latch memory using the devices that can fit 100 gigabits in a square centimeter, and has designed a highly scalable 3D design (consisting of up to 1000 layers or 1 petabit per cm3).[7]
http://en.wikipedia.org/wiki/Memristor

The nice thing about memristors is that they're so bloody simple, two terminals, simple structure, easy to mass-produce with little chance of flaws, at production processes of far below 10 nm. Switching speeds of around 1 ns, making Flash look like a bloody snail, bit-addressable so very fast read/write speeds, again making Flash look ridiculous.

I think you may be surprised by how quickly memristors will find their way into the consumer market :)

HP says 2013: http://www.technologyreview.com/computing/25018/?a=f
 
I didn't say the data will vanish instantly, but that it's the point where data retention can no longer be guaranteed. Slowly bits will start to go poof after this point.



I don't know why there aren't any articles on this (as far as I know, but I haven't searched for them). The loss of data with Flash memory is a fact, however, as this type of memory is based on the retention of electrons (electron trap). Since no kind of retention is perfect the electrons will slowly work their way out of the trap. Once the number of electrons drop under the read limit, the data will be gone. The speed with which this limit will be reached depends on the number of electrons stored initially and the number of electrons needed for a successful read. With newer, smaller process sizes both numbers become smaller, this is why data retention with old SLC (lots of electrons in a big, relatively electron-tight trap) is about 10 years, whereas 32 nm MLC Flash, where far fewer electrons are used in a far more leaky trap since the structures are smaller and thus leakage is higher, data retention is much shorter, much like current leakage is a problem with CPUs and similar as they're made smaller.

In short, the fact that Flash memory has a limited data retention is very well established at this point. What the exact limit is per Flash chip differs, of course, but the fact remains that it isn't going up while process sizes are going down.

I want to follow on the above, is there a fix on the above? The last time I heard, Intel SSD G3 has to re-read and re-write the data every ?? days to retain the electron integrity. Otherwise, the data will be lost due to lack of use, and as such, a loss of electron integrity
 
I want to follow on the above, is there a fix on the above? The last time I heard, Intel SSD G3 has to re-read and re-write the data every ?? days to retain the electron integrity. Otherwise, the data will be lost due to lack of use, and as such, a loss of electron integrity

Flash manufacturers are trying new methods to make those electron trap structures better, so it's not a linear degradation. It's becoming pretty clear at this point that 1X nm flash memory is going to have major issues, though. I'd be interested to learn how often the data on that SSD you mention has to be refreshed.
 
So would I. Anandtech has an article on the degradation of data, but there is no follow up article, I couldn't help but wonder if intel pay him off and ask him to stop writing those articles

and I found it impossible to believe that no1 talks about this as so many of us buying and using SSD. How can we use a technology if we don't know it is safe to use? That the data can disappear?
 
Last edited:
I've been using my Intel 320 160GB since launch over a year ago and have never had any issues. Crystal disk info reports the drive as still being good as new.

I've heard about data disappearing before we started seeing 22nm nand in actual drives. I've never heard of it being an issue in actual practice. I guess that its possible if you were to let your pc sit for a couple of years but who is going spend a couple hundred bucks to buy a ssd to sit for a couple of years?
 
One of the questions is, how long will stagnant data stay valid...even when the SSD is powered up? Other than Windows updates, I imagine 90% of the files in my Windows directory haven't moved an inch in years, especially since you aren't supposed to defrag an SSD. Does it get refreshed constantly? If so, wouldn't that kill some of the write endurance? These are questions which have not been answered. Only time will tell.
 
One of the questions is, how long will stagnant data stay valid...even when the SSD is powered up? Other than Windows updates, I imagine 90% of the files in my Windows directory haven't moved an inch in years, especially since you aren't supposed to defrag an SSD. Does it get refreshed constantly? If so, wouldn't that kill some of the write endurance? These are questions which have not been answered. Only time will tell.

I don't know for sure, but as long as you're doing some writes to the drive, static data should get moved around and refreshed periodically as part of the normal wear leveling algorithms. Otherwise, if you had a drive that was 90% full of static data, the remaining 10% would get written over and over, and would receive a hugely disproportionate amount of wear.

Now if you had a drive that was always powered on but you absolutely never did any writes to it, I don't know how that would be handled. I would hope the better firmwares had some kind of timer that would keep track of the age of the last write to a block and would move that data periodically on its own initiative.
 
I've been using my Intel 320 160GB since launch over a year ago and have never had any issues. Crystal disk info reports the drive as still being good as new.

I've heard about data disappearing before we started seeing 22nm nand in actual drives. I've never heard of it being an issue in actual practice. I guess that its possible if you were to let your pc sit for a couple of years but who is going spend a couple hundred bucks to buy a ssd to sit for a couple of years?

when we said data disappearing, we are talking about files not access for a while. For e.g. you could have some family photo simply sits in a sub-dir on your drive. So unless you go thru ALL your data files, you have no way of knowing for sure
 
when we said data disappearing, we are talking about files not access for a while. For e.g. you could have some family photo simply sits in a sub-dir on your drive. So unless you go thru ALL your data files, you have no way of knowing for sure

The drives firmware should take care of moving the data around when needed.

Chkdsk should tell you if any files are corrupted.

Look at how old this thread is. 22nm nand has been around for a while now and I haven't heard of this being an issue for anyone in real world usage.
 
Back
Top