Slashdot Mirror
32 GB Flash Storage Drive Announced
Audrius writes to tell us TG Daily is reporting that Samsung has just announced a new 32 GB Flash storage device. The aim of this new solid state disk (SSD) drive is to completely replace the traditional hard drives in many laptops on the market. Some of the advantages offered are the 1.8" form factor, read speeds more than twice that of a normal hard drive, and the promise of 95% less power use.
This will only work if they can get the prices of flash down.
$50.00~70.00 per gb is still nothing in comparison to $0.40~$0.80 you can get on hard drives.
If you RTFA you would see the target price is $750 and $1000 ... $6400 is the price of current flash hard drives in that size range.
Its just a matter of time for flash.
I believe he was probably talking about the upper range of commonly available laptop form-factor drives.
During heavy disk read activity, the HD is only uses 15% of all the power. (source) The real key to decreasing laptop power consumption is dimming the screen, which can reduce power consumption percentage from 26% down to 7%.
This technology has already been put to use in a commercial environment, and has given outstanding performance from what I've seen. The game EVE Online http://www.eve-online.com/ has already done this with their clustered servers and greatly reduced the lag. Keep in mind that this is a game where there is only a single universe (No shards or other servers) and they quite often push over 20,000 simultaneously logged in accounts at a time.
When placed in the right environment, this technology just screams. A good example would be for huge database operations that have hundreds if not thousands of concurrent accesses. The databases that maintain the pay information for the US Military come to mind easily.
Someone had posted this on another flash drive story here but it basically went that if you reserved 10% or so of the drive simply to keep rotating blocks it would last as long as a hard disk, more or less.
~S
The author is talking about 1.8" hard drives like what is used in the iPod. I don't know about you but I have seen Apple selling any 400gb iPods yet...
It estimated to cost$700 - $1000. While this may seem like a lot, for something new, this isn't. I remember reading how much a hardrive would have cost for an old IIGS that had maybe 8 disks worth of storage space I think. And although expensive, $700 isn't expensive enough to be out of the reach for consumers. Just expensive enough to be out of the reach for most sane typical consumers.
In undeveloped countries, the consumer controls the market. In capitalist America, the market controls you.
Price point on the 32gb drive is expected to be $750-$1000. The $6400 product is a currently available military grade drive. It'll take a wee bit more abuse and temperature range then the 'cheapest bidder' built one that will hit the commecial market.
-Rick
"Most people in the U.S. wouldn't know they live in a tyrannical state if it walked up and grabbed their junk." - MyFirs
Most flash can handle something like 100,000 erase cycles. And most flash file systems have wear-leveling algorithms to ensure you're not hitting the same sectors over and over. Even with standard usage they should be good for several years at the very least.
The current number-of-writes for flash is somewhere in the 100,000-1,000,000 write cycle range. That's a lot of writes. Also, keep in mind that all flash chip controllers include logic that performs "write-leveling". This means that a specific chunk of data will 'jump' from one area of the memory to another in order to prevent one area from being worn out. Add to that the fact that flash chips contain some extra capacity to compensate for bad blocks.
With a careful configuration of Windows (no page file, no IE cache, no temporary files, use a RAM disk), this is certainly viable. In the absense of music/movie collections and monster games, even the 32GB size isn't that restrictive.
Help find a cure for cancer. Join the [H]orde
Flash memory that works has a much longer MTBF than hard drives, but each cell fails at approximately 10000 writes. HDDs fail randomly, Flash fails predictably, so this can be a good thing. Just make sure your filesystem rarely does or needs defragging, and does not log every read.
The device you link to has only 2.5" and 3.5" form factors available. This device fits in a 1.8" form factor. Nice try, though. I can see why you post as an AC.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
This is certainly a valid worry. As I understand it, however, modern flash memories have more or less dealt with this problem, because:
(1) The number of rewrites is now quite large (hundreds of thousands?)
(2) The writing-to-disk software/hardware implements "load balancing." If you rewrite the same file 1000 times, it won't use the same exact block on the flash disk for each of those writes. Instead it will move from block to block with various writes/deletes/modify actions. This, coupled with some "slack" (the actual disk size is a little bit bigger than the "useable" disk size) allows for the wear to be distributed over the whole device.
(3) The system uses conventional error-correction and flagging of bad blocks.
As another poster pointed out, magnetic hard disks also have a limited number of rewrite cycles. But in practical terms we usually don't reach this limit. For critical applications I imagine you'd use a RAID of flash disks just like a RAID of magnetic drives.
RTFA- their write speed is reasonable (at about half that of current hard drives, supposedly, though see below for questions about this) and on a 32GB drive with a reasonable usage pattern- well, how often do you reformat an entire drive? With over a million writes on modern flash memory, it's going to take you a while to use up all the writes this drive has.
And now for that questionable bit, from the article: While the SSD's capacity of 32 GB cannot compete with traditional hard drives that currently offers up to 80 GB space,
I don't know abut you, but I've seen hard drives in this price range offering up to 500GB and one USB/Ethernet external that offers 1TB at less than 2x the price. Which throws the write speed into question- if 80GB drives are considered their max.
SJW: a person who perceives an injustice, and while correcting it, commits a greater injustice.
That's probably Taiwanese dollars, that's where the byline is.
0.0308676(Taiwan/US) * 6400(Taiwan) = 197.55264(US)
That's still $6.20 US/GB so still not very desirable, but if they can EoS down, and get the battery life trade off it may be worth it.
"There are no facts, only interpretations." --Friedrich Nietzsche.
Some filesystems (ie Reiser4) move or consolidate files (aka "defrag") in the background , and don't know what kind of block device they're on. You'd want to tell it not to do bother doing that then. Except the kernel/ATA interface still reads and writes by the block, but a block in some filesystems (Reiser4) may contain parts of several files, so you'd want to eventually consolidate files so you don't have to read/write a whole lot of blocks to access a single file which might be smaller than a block.
A worst case scenario would be a filesystem similar to Reiser4 with consolidation turned off, and lots of files growing by small amounts frequently.
Yes, but, don't 'flash drives' suffer from a more limited number of times you can read/write to them than a regular HD?
The typical number tossed around for NAND erase cycles is 100,000. You can read as often as you like, but to write data, you have to erase a block of data first, 132KB on the devices that I design with.
Of course, those are the data sheet numbers - that is what the manufacturer guarantees. Reality is usually quite a bit better. And it wouldn't surprise me if Samsung and others had some much higher performance flash memory in the pipeline.
-h-
My old Piper Cherokee isn't pressurised at all. I can get an expensive navigation system, but if I want to use a less-expensive laptop, I can't take it above aboute 10-12,000 feet without the risk of crashing the heads.
Actually, this won't work. The wear levelling doesn't know if a block is 'full' or not, so it will just switch the contents of a pair of blocks. Your frequently-written file will move all over the flash chip(s), and so will your static files.
I am TheRaven on Soylent News
The nice thing about Flash is that after a cell has failed, it just becomes read-only. You can get around this quite easily in the OS by just marking the failed block as bad in your inode list. Over time, your flash drive will shrink in capacity. When it gets too small, you just copy it over to a new one and repeat the process.
I am TheRaven on Soylent News
Samsung announced these drives last year and said that they should be available by 8-05. I have been looking regularly to see when they are going to hit the market for close to a year now. I wish they would go ahead and release them or plan announcements a little better.
r essRelease.asp?seq=20050523_0000123980
If they could get the price down to near $10/GB I would buy one. I don't store mass amounts of audio or video on my laptop so a 16 or 32 GB would be fine for me. The decrease in power consumption and the increase in battery life could make the investment well worth the cost. I can currently get about 5.5 hours of battery life out of my laptop, if this could be increased to 7 or more from the reduced power consumption I would be all for it.
http://www.samsung.com/PressCenter/PressRelease/P
We are talking about different layers, now I understand what you're saying.
The flash device has a control processor on board that manages the charge pump operation (for writing and erasing) and keeps the status of the array (bad blocks). When the application layer asks to store data the flash control processor goes and stores the data on the next erased block with the least number of erase cycles against it (erase is what causes damage to the flash cell). The application layer has no idea where in the array the data is physically located, nor does it care. The FCP tracks all of this and it was this layer of the device I was refering to. There is not any reason to move data from one block to another, just to free up a block with less erase cycles, as you have no idea whether that data is persistant or not.
Most flash devices have three arrays: Main array space, microcode array space, and processor(configuration) array space. There may be redundant blocks available, but of these arrays only the main array is useable for storing data.
YMMV as I work on NOR devices and as such are more closly alligned with memory devices rather than block devices.
-nB
whois gawk date unzip strip find touch finger mount join nice man top fsck grep eject more yes exit umount sleep dump
Damn, that answer really made me laugh. It was the last answer I ever expect. Is it really true. ;)
Thanks for the laugh.