Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
ST-MRAM the end of hard drives?
thinktank001
Member UncommonPosts: 2,144
in Hardware
http://semiaccurate.com/2012/11/16/everspin-makes-st-mram-a-reality/
Aside from mass storage or a back up, does this mean the end of hard drives in the near future?
Comments
I stated to disregard that part. My interest in the topic was more along the lines of it possibly replacing hard drives as the primary storage for the system.
If its not broke dont try to fix it.
For primary storage, it doesn't have any real advantage over solid state drives, but has a huge disadvantage in cost. The only way it has any chance of replacing hard drives is if it not only works as well as hoped, but keeps scaling with something analogous to Moore's Law for several years longer than NAND flash.
What's more plausible is that it will replace system memory. Imagine being able to turn off your computer entirely, not merely hibernate but completely off and not draining the battery at all, and then resuming exactly where you were within a fraction of a second.
Take note that this is very low niche and thus low quantity thing. Additionally they have to use 90 nm to produce ST-MRAM, newer lower node fabs are not avabile for that kind of low-amount production volume and you know how huge diffrence in cost that make.
NAND is good and I support buying it, but it is dead end technology cause oflimited amount of cycles with lower nodes. With 10 nm and TLC it will be incredibly low. Software / firmware get better but not that much. Something will have to replace NAND propably. Not now. Not in 5 years. But in 10 years? 15? Yes.
According to the article, if everything goes well, they're hoping to match DDR3 pricing--not today, but after they catch up to DDR3 in cutting-edge process nodes. That's several times the price per GB of NAND flash. While NAND flash will stop scaling eventually, so will SDRAM, ST-MRAM, and everything else. ST-MRAM would need to last about three full node die shrinks longer than NAND flash before it has a chance to become viable as a hard drive replacement.
Yeah it might not be ST-MRAM unless something will drastically change in this technology cost-related.
I am not saying that NAND cannot scale. It does and it still can, but cost in amount of writes you can do to single cell is insane and it will start taking it's tool soon and limit this technology alot.
Most other storage technologies don't have this problem. ST-MRAM like "normal" DDR RAM theoretically can be written infinite number of times.
There are also other companies working on other variants of MRAM tech. I think there is some French company. Anyway interesting to keep an eye of.
My bad I think this is where I was confusing people. My point was that it is a non-volatile solution for system memory that with a large enough size could completely replace the need for a primary hard drive.
I'm more optimistic about NAND flash scaling than you are. Given a choice between 128 GB at 3000 P/E cycles, 256 GB at 1500 P/E cycles, and 512 GB at 750 P/E cycles, which would you pick? The reduced P/E cycle numbers on the larger SSDs look bad, but they get you exactly the same total write capacity. That's basically what die shrinks are going to give us going forward: you don't get that much more total write capacity, but you do get a lot more total storage capacity for the same price.
Furthermore, more total storage capacity will enable more total write capacity, especially if people tend not to get their SSD anywhere near full. More free capacity to play with means you can do more things with wear leveling to reduce write amplification. A given number of P/E cycles at a write amplification of 2x gives you double the total write capacity of the same number of P/E cycles at a write amplification of 4x.
Maybe something else will replace NAND flash. If some new tech that is really awesome comes, it could replace NAND flash even if NAND keeps on scaling flawlessly. But apart from that, NAND would have to stall for several years before anything else has a good chance at replacing it.
Problem is with SSD failure rate, total amount of write cycles does not decrease failure rates unless user is very conservative about using it's SSD space, Lower write amount per cell does decrease it though. Of course manufacturers are aware of that and that's why they invest into better frimware that try to distibute data over cells better and make cell usage more evenly and by investing into file systems specifically for SSD's like Samsung. That can make things better but it won't change physics.
It is not that high, but already I know multiple people that had their SSD's die on them and that were 25 or bigger nm MLC drives of various companies. ( OCZ, Intel, Crucial, Kingstone and something else don't remember). Because of NAND tech it frequently happen without any warning. Of course if it happens of warranty then you get it replaced but that's not the point.
Of course HDD's also do fail and they are prone to damage / bad sectors if they are moved during work which SSD's are immune to, but thing is that as schrink is going down failure rate of SSD"s will increase and we'll see that imho with low process nodes TLC's soon.
Besides SSD's are and even as they increase in capacity they will still remain relatively expensive, compared to HDD's.
Of course HDD's are performance dead end technology and SSD's are insane increase in everyday use comfort due to responsiveness, but if you pay more you expect more and not only in preformance but in reliability.
If 50$ HDD get bad sectors and you have to make backup copy and replace it it is not felt as bad if 150$ SSD die without warning on you.
I agree though that this matter is a bit subjective and how well NAND technology will scale down is remain to be seen.
Interesting link:
http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923-9.html
Replacement for NAND technology. Well it won't happen anytime soon,I agree that for it to happen NAND tech will have to stall and there would have to be other technology showing enough promise to invest serious bucks in it.
Not things that can happen fast. Long years.
Anyway NAND tech used as mass storage SSD's is great thing. Nice tech, but imho without long future.
Color me a bit skeptical.
From my understanding, the only inherent "more reliability" in SSD's comes mainly in portables - where mechanical shock is as big a killer of mechanical hard drives as anything, and SSD's are near indestructible by G-Force.
The actual technology I would say about ~as reliable~ as a traditional hard drive. The Tom's Hardware article that you link doesn't confirm or deny that - it just says that we don't have as much data, because SSDs haven't been around as long.
As far as "many people" with failed drives - oh well. Anecdotal evidence. Without failure rates from an official study, or RMA rates from manufacturers, you don't have much other than a personal observation on a small data set. I could just as well make the claim that SSDs are infintely reliable because of the dozens I've either installed in systems, or know of that are installed in systems, I've seen 0 fail, including those that are approaching 4 years old now. But that would be an absurd claim to make.
As far as I'm concerned, SSD's are more expensive, but that cost buys you orders of magnitude better performance - not anything to do with reliability. There are other ways to insure reliability (the easiest example to start with being - get enterprise level drives rather than consumer drives, and proceeding on through using other technologies and strategies aside from just drives to ensure data reliability, such as redundancy and backups).