Flash SSDs to decrease in durability, increase popularity

E

Elledan

[H]ard|DCer of the Month - April 2010
Joined
Oct 18, 2001
Messages
15,913
Just read this interesting series of articles in Nikkei Electronics, the part which really caught my eye was this part:
SSD prices may continue to drop in the future, and thanks to high-density packaging and other technologies from the semiconductor industry, will continue to shrink in size, thickness, etc. The future indeed seems rosy for equipment manufacturers.

There is, however, a problem that still has to be overcome before SSDs can achieve that future growth. Evolution toward lower cost will be made possible by finer manufacturing technology, multi-level architecture and other advances, but these changes will also degrade NAND Flash memory quality.

NAND Flash memory quality is also beginning to drop. Chips manufactured using 90nm-generation technology in 2004-05, for example, were assured for about 100,000 rewrites and data retention of about a decade. As multi-level architecture and smaller geometry are introduced, quality is showing a sharp decline. The 30nm 2-bit/cell chips expected to enter volume production in 2009-10 may well end up with a rewrite assurance of no more than 3,000 cycles, and a data retention time of about a year. The first 3-bit/cell chips are hitting the market now, with only a few hundred rewrites.
Click to expand...
This quote is from the 3rd page, the series starts at: http://techon.nikkeibp.co.jp/article/HONSHI/20090528/170917/

True, they point out later in the article that at the 'average' write rate of 20 GB/day an SSD would last 35 years even with just 3k write cycles per cell, but it's still a big step down from 10k-100k cycles. For those writing larger data sets to their SSDs on a daily basis, this could possibly become an issue.

And then there's the issue of data-retention. While 10 years isn't a lot compared to HDDs which can retain their data for decades, 1 year hardly makes it qualify as non-volatile storage for many usages, like long-term backups. Don't expect to turn on that PC after a year of storage and expect it to still contain all data on its SSD, or to see those photos on that flash card you didn't use for a while. This is an issue which most definitely needs solving.

Still, SSDs seem to have a bright future as caches in servers, especially where rapid, random data retrieval is crucial, such as in database operations.

Discuss amongst yourselves :)
 
Retention of a year?! WTF!

This means that your OS (or files that are only read, hell, the MBR) will self-destruct within a year. I would be pissed if it was 5 years, 1 year is just not acceptable.
 
Hypernova said:
Retention of a year?! WTF!

This means that your OS (or files that are only read, hell, the MBR) will self-destruct within a year. I would be pissed if it was 5 years, 1 year is just not acceptable.
Click to expand...

Do you leave your computer off for a year normally?

I don't see this as an issue for 99.9% of people. All this means is that if you are doing archival, you need another storage medium, not flash.
 
Lazn_Work said:
Do you leave your computer off for a year normally?

I don't see this as an issue for 99.9% of people. All this means is that if you are doing archival, you need another storage medium, not flash.
Click to expand...

It'd reduce Flash media to short-term storage, yes. However, people have come to treat Flash media (USB sticks, SSDs) as the same kind of long-term storage as for example FDs and HDDs. I can imagine this coming back to hurt them because they weren't aware of it. Heck, I didn't know that the current data retention of Flash is 10 years right now. I always assumed that it'd be somewhat similar to magnetic media (decades).

Anyway, I could imagine for example a college student leaving his PC at home and using another PC at the dorm, or going overseas for a year or so, only to return to a blank SSD a year later. Plus, this data retention of Flash is only going to get worse as feature sizes get smaller on the chips, reducing data retention to what? 6 months? 3 months? At what point will we just stop calling Flash media 'non-volatile'?
 
I always figured this would happen. Cling on to your SLC drives while you can. This is the same thing as the LCD panel wars where quality kept declining in order to costs, and we went from S-IPS to VA to TN technologies.
 
astrallite said:
I always figured this would happen. Cling on to your SLC drives while you can. This is the same thing as the LCD panel wars where quality kept declining in order to costs, and we went from S-IPS to VA to TN technologies.
Click to expand...

It seems that SLC is becoming mostly used for enterprise market. Hopefully we'll still see manufacturers making also SLC drives. It would be bad if they'd ditch those almost completely so that there wouldn't be any competition driving prices.
 
Looks like I'll be buying that SLC intel drive when I get payed next week.
 
You overreacting. Do you even know how many write operations your computer typically performs in a month?

If you can't answer that don't knock the rewrite issue because of ignorance and fear.
 
mutantmagnet said:
You overreacting. Do you even know how many write operations your computer typically performs in a month?

If you can't answer that don't knock the rewrite issue because of ignorance and fear.
Click to expand...

A 128 GB SSD has got 262,144 blocks of 512 KB, assuming a write rate of 4 KB to each block each 2 seconds (around 20 GB/day), the drive's 786,432,000 available write cycles (3,000/cell) across its cells would be used up in 200 seconds, or just over 3 minutes. An SLC drive with 100k cycles would last ~111 minutes. In a more realistic write scenario this would still mean that the 3k cycle drive would last around 33 times less long than the 100k cycle drive. A 35-year lifespan of an SLC drive could be reduced to just a year or so with the MLC, 3k cycle drive.

I think that's pretty darn significant.
 
http://bits.blogs.nytimes.com/2009/05/22/counting-down-to-the-end-of-moores-law/

also

http://www.supertalent.com/datasheets/SLC_vs_MLC whitepaper.pdf

to get an idea of what MLC means and what the challenges really are. In the nytimes article, he seems to indicate sandisk has a 4-bit MLC while the supertalent whitepaper references a 2-bit MLC

The more bits per cell, the fewer electrons per bit, also as cell size diminishes with smaller processes (30nm, 22nm etc). To me, fewer electrons per bit must mean greater susceptibility to data corruption/loss.

However, I think the SSDs made today and in the near future have plenty of write cycles available for the average person. I mean, I've never kept a Hard Disk longer than 5 or 6 years before it died or was obsolete.
 
I find it interesting how SanDisk's Harari said how he expects to see the capacity of flash chips only double twice more before they run out of electrons to count. I'd indeed agree with him that it's a brick wall.

I wonder how MLCs fit into this... at least I won't be investing in Flash media any time soon :)
 
extide said:
Sad day for you, you are missing out...!!
Click to expand...

Well, I could, but I'd need to pull out of it shortly before it's going to crash, if those predictions work out and no magical solution is found.
 
The question I have is about the 1 year retention of that flash chip they mention. Does it mean 1 year no matter what, 1 year from the time the data is written to that block or 1 year from anytime that block is accessed? If I had to guess it would be 1 year from the time the data is written, which means infrequently accessed data could become corrupt after 1 year even if you use the drive daily. To me, this is unacceptable for any storage device, daily use or otherwise.
 
Astral Abyss said:
The question I have is about the 1 year retention of that flash chip they mention. Does it mean 1 year no matter what, 1 year from the time the data is written to that block or 1 year from anytime that block is accessed? If I had to guess it would be 1 year from the time the data is written, which means infrequently accessed data could become corrupt after 1 year even if you use the drive daily. To me, this is unacceptable for any storage device, daily use or otherwise.
Click to expand...

Data retention with Flash depends on the presence of sufficient electrons in a cell. As the progress the past years has been towards using fewer electrons per cell, this has affected write cycles (each cycle damages the cell somewhat, making it 'leaky') and data retention (the electrons trapped in a cell will eventually get out of it, it's just a matter of when). It's similar to how small feature sizes with CPUs has resulted in more leakage with the transistors, forcing manufacturers to use new techniques to compensate for this.

As far as I am aware, Flash media isn't refreshed outside write cycles, so after data has been written to a block, the data retention timer will start ticking. After the rated time (10 year, 1 year, or a few months for those 4 bit/cell chips), enough electrons will have escaped the cell to render the cell blank ('0'). The reading process of Flash media does not refresh the cell, as it only 'counts' the electrons in the cell, it doesn't read it out destructively and writes it back after reading like with DRAM.
 
You must log in or register to reply here.
Back
Top