Slashdot Mirror
512GB Solid State Disks on the Way
Viper95 writes "Samsung has announced that it has developed the world's first 64Gb(8GB) NAND flash memory chip using a 30nm production process, which opens the door for companies to produce memory cards with upto 128GB capacity"
Capabilities aren't very important if they aren't affordable. So maybe some government contractors can afford those things now, I don't think it would be that interesting to the consumer until SSDs get to a tenth of the cost.
It's not a dupe. The previous article said that 64 Gb chips could be combined into a 128 GB device. Now they can combine 64 Gb chips into a 512 GB device. A huge advance!
You could use the same logic to conclude that 512 terabyte solid-state media is on the way.
......when I think that porn, or some equivalent thereof, has been responsible for all human progress throughout history.
... should be enough for all dupes.
...when I saw the, today typical in many submissions, "Source: http://news.bbc.co.uk/1/hi/technology/7057717.stm"
Why go to the BBC directly when we might read something more obscure first.
It's no so easy to use the 1,000,000=1mb with this system. Unless they do it anyway.
Gone!
Why won't this meme die already?
"All you have to do is be fragile and grateful. So stay the underdog." Chuck Palahniuk, Choke
Does anybody know how well flash SSDs perform in RAID arrays? 15kRPM SAS drives are horrendously expensive so if I could plug a couple small flash drives into my RAID card (RAID 0) I'd be a happy camper. Can't find benchmarks anywhere and flash drives have horrible write speeds which means they have terrible OLTP performance.
What if Digg added local news and a Slashdot inspired comment karma system? ---
http://houndwire.com
There are ~31.5 million seconds in a year. If you assume that the write speed is 1 GB/s and that you were writing constantly, you would generate ~62 thousand writes to each bit. Roll the write speed back to a still unlikely ~100 MB/s(still writing constantly) and you generate about 6 thousand writes to each bit in a year.
Throw in the fact that the controllers for these chips spread writes around and you can be certain that the endurance is not a problem.
Nerd rage is the funniest rage.
It's already so high as to be meaningless, it will outlast mechanical failure of a traditional hard drive for example.
This has been discussed before. Modern flash drives use wear leveling to avoid writing to well-worn blocks and to move unchanging files from unworn blocks so they can be used more. Yes, it adds complexity and yes it slightly delays the write process. But it's invisible to the CPU and OS and takes far less time that it would to move the heads of the standard mechanical HD. An SSD is free to organize blocks in any order in the address space because there is virtually no penalty for fragmentation.
I think you will find that even in very heavy use applications (e.g. working with HD video or using the SSD for virtual memory) that the lifespan of these drives be longer than a decade (and longer than mechanical HDs). Moreover, they will fail gracefully as blocks become tags as worn.
Two wrongs don't make a right, but three lefts do.
FusionIO.com demo'd a 640Gb flash-based storage product with 160 parallel memory channels. Up to 4 of them can be put into a single server to give 400,000 IOPs. A 10k drive does about 100 IOPs, a really good RAID card with a rack full of drives peaks at 5,000 IOPs, a freakin huge $1M+ SAN does maybe 100,000 IOPs. The cost is estimated to be $30/GB, which is dirt cheap in the enterprise storage market.
These devices can already do block relocates.. The MTBF on these drives is on the order of 2 million hours. WAY better than winchester drives and so far out there that I kinda wish people would stop bringing this up.
simple... people keep modding it interesting/informative/etc. instead of Troll. And if they're not trolling, there needs to be a new mod: -1 Clueless/Oblivious.
And how many "reads" does someone around here have to soak in before they get it? We've talked about the write limitation every time this comes up at length. SSF is at least an order of magnitude more reliable than the spinning platters.
In a recent article on write endurance published in STORAGEsearch.com, editor Zsolt Kerekes provided theoretical computations on the longevity of solid state flash drives deployed in enterprise server applications. His test solid state drive had the following specifications: total capacity of 64GB, sustained write speed of 80MBps and a write endurance rating of 2 million cycles. By assuming that data is written in big blocks and there is perfect implementation of wear leveling techniques, Kerekes estimates disk endurance at 1.6 billion seconds, which translates to 50.74 years.
Debunking Misconceptions in SSD Longevity
So when does the 512GB iPhone come out?
SSD, doesn't that stand for Single Sided Disks, as in floppies... ; may as well...
anyways, if we had 1000 terabyte solid drives for $10 then you'd hear wining for the yet to be released Googleplex drive for $5...
Like damn, anyone using up their new 100 gig drives faster than the next size is out for less money?
To back up very large drives today, it near cheaper in time/labor and costs to just use hot swap drives, where the back up is the removed drive, plugged in and run for 15 minutes a few times a year, if even that. Or a rotation system as was done with tape.
Notebook drives currently cost as little as about $50:80GB, or $6.50:GB, which is a good size for a mobile device, and almost the largest available.
Flash is as little as $64:8GB (USB), $8:GB. Removing the redundant USB connectors and packaging, putting it in a single drive the size of a notebook drive, would give an 80GB Flash drive for somewhere closer to $50 than to $80.
FWIW, a 4GB microdrive is $30, or $7.50:GB.
These numbers show that a Flash drive competing directly with a disc drive is already right around the corner. By the time 2010 comes around, what will mainly be different is the upper capacity around 1TB, with probably Flash cheaper than discs.
--
make install -not war
So solid state disks are all about NAND flash memory, right? I thought that SSDs would be all about MRAM, and that MRAM SSDs would be viable by the late 2000's. What's up with that?
You just got troll'd!
Don't flash chips have a much shorter lifespan than regular hard drives and relatively low number of reads and writes? Or is that just with older flash tech?
Does everything in the damn world have to be blamed on Vista? Next on Slashdigg: The fires in California - Vista responsible?
I thought SSDs are slower than regular disk drives when making lots of small writes (like when you're installing software, for instance)
http://en.wikipedia.org/wiki/Solid_state_drive#Disadvantages
What the hell are you talking about? It's the media and other content you need storage for. You can run any operating system on 16 GB if you wish, but whining about how big Vista is makes you look stupid.
Today's operating systems (OSX, Vista, etc) are not big because the software is bloated with meaninglessness, but because there is not a living soul out there who is considering XP, Vista or OSX but cannot get it because their hard drives are too small. Is it not obvious that developers want to make full use of the current generation of hardware?
I'm sure Microsoft could strip down Vista to something the size of 300 MB or so if only they wanted to remove drivers, icons and other graphics, sounds, media players, web browsers, etc. On the other hand, that would kind of kill the whole purpose of the operating system.
Full Tilt
I doubt Microsoft would allow Dell etc to pare their Vista install down, or change its component structure to move portions to a separate drive, just so it could boot on a flash drive.
Unless you count the Asus EEEPC, which (depending on the model) has either 2, 4 or 8 Gig drives that come with Linux. They don't run Vista, but they do come with instructions & drivers for installing XP.
I'm sure Dell & such would follow suit much sooner if M$ would let them load XP instead of Vista, but Dell isn't afraid of Linux and will even be introducing solid state 32G laptops (I'm assuming running some sort of Windows) soon.
Notebook drives cost around $0.63 per GB. A 80GB flash drive would cost around $64 * 10 = $640.
Hi,
I already boot/run my main Internet-facing server (Ubuntu) from a 4GB memory SSD card to minimise power consumption, and I have more than 50% space free, ie it wasn't that hard to do.
http://www.earth.org.uk/low-power-laptop.html
I'm not being that clever about it: using efs3 rather than any wear-leveling SSD-friendly fs, and simply minimising spurious write activity, eg by turning down verbosity on logs. And laptop-mode helps a lot of course.
Now that machine does also have a 160GB HDD for infrequently-accessed bulk data (so the HDD is spun down most of the time and a power-conserving sleep mode), and it would be good to get that data onto SSD too. But a blend, as in many memory/storage systems, gives a good chunk of maximum performance and power savings for reasonable cost.
Rgds
Damon
http://m.earth.org.uk/
A UAC prompt was holding up the fire alert system for San Diego County
"Hello 911? I just tried to toast some bread, and the toaster grew an arm and stabbed me in the face!"
There, I fixed if for you.
Faster! Faster! Faster would be better!
Well, umm. Vista takes up more processor time, runs the computer hotter.
Computer running hotter means more power used.
Power generation contributes to global warming.
Global warming contributes to increased forest fires.
Therefore, it follows:
Vista is responsible for the fires in California.
What could possibly be more logical?
Faster! Faster! Faster would be better!
I don't follow your math.
8GiB (64gbit) capacity, 1GiB/s write speed, 8 seconds to write every bit on the chip.
31,536,000 / 8 = 3,942,000
So, you would completely rewrite the chip almost 4 million times per year. Scale it back to 100MB/s writing constantly and you'd generate almost 400 thousand writes per year.
All they need to do is get the cost and MTBF in the right place and the Terrabyte memory will appear.
I think we've pushed this "anyone can grow up to be president" thing too far.
Let's imagine a daemon that updates a file every 5 seconds. That's 518400 writes a month. Most flash based storage devices are rated between 100k to 1 million writes.
Your example usage and my example usage are the two corner cases, the endurance problem and it's effect depends on what the drive is used for. Still, I just wanted to illustrate that your scenario is an optimistic one.
It takes a man to suffer ignorance and smile
Be yourself no matter what they say
I'll stop bringing it up when my usb flash drives and sd cards stop failing after just a few weeks of intensive use.
Ugh. And I though that they had seen the light and decided to go in base 10 and count the actual bits.
I lost my sig.
Time to buy some stock in solid state manufacturers, perhaps... I can only foresee one evolutionary change in data storage for common home use, really. The technology is still young, but already showing lots of promise.
Beware: In C++, your friends can see your privates!
I was using the 512 GB capacity of the drive.
Nerd rage is the funniest rage.
That's where the wear leveling comes in, those 518400 writes don't go to the same spot(and if it works fairly well, the cycle count for the device is basically limited by the write speed).
Nerd rage is the funniest rage.
I was a bit vague and this is why I stated it's a corner case. It is entirely plausible that someone uses a reasonably full hard drive and updates some files a lot. Consider the scenario of simple atime updates, databases, logs, package management, emails...
This is what I ment by a corner case. I was actually quite generous since I didn't presuppose utilities that check whether a file has changed every second. The bottom line is, I can see the drive failing under relatively uncommon but normal operating conditions.
It takes a man to suffer ignorance and smile
Be yourself no matter what they say
Are there any limits on the number of rewrites with these type of solid state drives, vs like a typical SD card? Surely, they must be a bit better than the average SD card.
The basic pages and blocks of flash are themselves not powers of two! Most 512-byte page NAND devices have some number (~16) bytes of extra area in each page for bad block management, spare bits, and ECC. It's really arbitrary.
Indeed, most flash cards are odd-sized when you count the sectors, just like disk drives, and for much the same reason -- the ECC logic, bad block reserve, and logical sector tracking take some amount of space. Don't take my word for it, check for yourself! Grab a handful of 4GB cards and see if any two brands have exactly the same number of sectors.
Back when cards were smaller, 12MB, 80MB (Lexar) and 96MB (i2GO) CompactFlash cards were not uncommon.
Dude... I know techies has a binary fetish, but get this:
People don't fucking care about the manufacturing process or memory adressing details. Non-techies always count in powers of ten, and I and you will do nothing but making ourselves look like retards if we try to argue that 512 + 512 = 1k.
Does power stations redefine a kilowatt to 978W? Does butchers sell kilos of meat by the 1012th gram? Nope. Would I allow them to redefine these terms based their maufacturing process? Nope.
This is probably harder to grap for someone living in the US. But keep in mind that in other parts of the world, every single person used base 10 for everything (except time). Kilo = 1000. Always. No exceptions. For all you care, that's what God wrote on the stone tablets he handed over to Moses. You learned that in elementary school and have been using the fact daily, for everything from driving to buying meat.
Thus: 64Mb = 64,000,000.
Obviously. What else could it possibly be? We are talking about a number that is on the packaging shown to end-users, correct? Not predicting the future here, this is a guarantee: Base 10.
Oh, and I killed the Byte for you. Bits all the way. These international unicode-days, the arachaic byte won't last much longer than base2. (Not to mention that marketing can 8-up the numbers just by switching case of the 'B'. Bring a camera when you tell them!)
I lost my sig.
Okay, how about a terabyte in a form factor small enough for a thunb drive, that costs one-tenth the price of traditional flash memory, and is a staggering 1000 times more energy efficient.
Researchers Develop Technology to Make Terabyte Thumb Drives Possible
Makes a mere 512GB flash chip look a bit sad, doesn't it?
Any sect, cult, or religion will legislate its creed into law if it acquires the political power to do so.
Not to ruin your joke or anything, but Vista runs about 3 watts cooler than Ubuntu on my laptop.
I hereby place the above post in the public domain.
On the headline you will see something like..."THE GIGANTIC STACK OF CHIPS THAT CAN STORE THE ULTIMATE QUESTION!!(cannot process tho, don't sue me for this)"
Chances are you'll be more likely to be hit by a car than your SSD running out of write-cycles.
The average consumer does not care about the base. The average consumer just feels ripped off when their 64MB flash card show up as 61MB on their computer. The average consumer expects a 64MB flash card to actually be 64MB. Since the computer uses base2 to report the size of the flash card, it behooves manufacturers to use base2 as well - since that is what the consumer is expecting.
You are wrong, the whole purpose of an operating system is for one thing and one thing only, that is to run programs.
A thumb drive using [programmable metallization cell memory technology] could store a terabyte of information
http://www.wired.com/gadgets/miscellaneous/news/2007/10/ion_memory
There is no sig.
The price premium for laptops with even small SSD's is astonishing. Almost $1000. As much as I love the idea of an SSD laptop and ever bigger storage for phones and PDA's the price has to become realistic before anyone will buy these in volume.
Well, not quite. When we're talking bits, not gates, the ECC/bad-block/controller logic doesn't count; a byte of flash is retrieved by reading eight data lines of output after punching in "n" address lines of input that uniquely address the byte. Look at the pinouts on a EPROM (2708 through 27080, and I think there was even a 27160?) or a 90s-era flash ROM (similar numbers, usually "28xxx" instead of 27xxx")
2708 - 8 kilobits, 1024 bytes when read as 8 bits x 1024 addresses
27512 - 512 kilobits, 64 kilobytes when read as 8 bits x 2^16 (16 address lines) for 65536 addresses
27080 - 8 megabits, 1 megabyte when read as 8 bits x 2^20 address lines.
> Back when cards were smaller, 12MB, 80MB (Lexar) and 96MB (i2GO) CompactFlash cards were not uncommon.
Which are all the sums of multiples of powers of two. Suppose you have space on the board for two chip packages. A 12MB card is an 8MB chip and a 4MB chip (or a 16MB chip with a bad spot, so we leave a couple of address lines disconnected to disable the 1/4 of the 16MB that's defective). An 80MB card is a 64MB chip and a 16MB chip. A 96MB card is a 64MB chip and a 32MB chip.
> the ECC logic, bad block reserve, and logical sector tracking take some amount of space. Don't take my word for it, check for yourself! Grab a handful of 4GB cards and see if any two brands have exactly the same number of sectors.
True -- but the OP wasn't talking about 512-byte sectors. He was pointing out the reason that you can't (efficiently!) address blocks of RAM that are arrayed in rectangular grids in anything other than powers of two.
no no, you have to make it even closer to the truth, yet, completely sensationalized! That's the Slashdigg way!
"Microsoft employees caught starting forest fires."
"We told them to upgrade, but they didnt listen. now that their computers have been destroyed they have no choice!" said Steve "The embalmer" Ballmer, picking up fragments of a broken chair.
On my new laptop, I installed Vista on a 20GB partition and I moved all of \Program Files and \Users to a separate 160gb partition, while linking them to C:\Users and C:\Program Files using NTFS junctions. The remaining 20gb is used for dual-booting Ubuntu/Sabayon/Whatever-Linux-Flavor-Of-The-Month To me, having Windows applications separate from the OS is a necessity and I will never go back to the boneheaded default of having all of my data stored as one big blob on the C drive.
/J, and then reboot into Windows and hope you didn't screw up.
Like most people, I have nowhere near enough *extra* storage capacity to actually back up 200gb of data. As such, I simply back up the entire contents of the 20 GB C:\ partition containing the core of the OS (using DriveImageXML, and not Windows' craptastic "backup" utility). This way, when my Windows installation gets hosed again, I simply restore the C drive, create the junctions to my apps and profiles on the G:\ drive, and continue as if nothing had happened.
Creating the junctions is troublesome. I believe the problem stems mainly from the fact that you cannot mess with these folders at all while the OS is up and running. Even if you shut down all your apps, Windows still acts as if these files are in use. Junctioning must be done from the another OS, from BartPE, or from the Vista recovery console (the 2K/XP recovery console is crippled in that you cannot alter files outside of \WINDOWS). You must create the new partition and assign it a drive letter, boot the CD or alternative OS, go to the console, change the drive letter of your target partition to the same one Windows uses (using the SUBST command), move all of the files to that partition (making sure to preserve their permissions), create the junctions on the C: drive using MKLINK
This is a feature that has lacked MS's blessing for many years. It has been possible to make junctions since Win2k, but it's always a big pain in the butt and AFAIK no one has created a simple tool to do it. Furthermore, it is totally unsupported by MS and other software developers. Ever tried to install an App to D:\Progams? Just watch as your C:\ drive is *still* filled with crap. WTF?! You pretty much have to *trick* the OS into thinking it is still using the C drive for everything using an ugly hack. To me, this is a huge design flaw in Windows and makes Linux that much more attractive as an OS.
This is all very well but you are totally wrong. Go download a datasheet of a popular FLASH part. Guess what? The capacity is an exact power of 2.
I'm not just making this up. NAND is naturally base-2 capacity sized. Yes, there is sparing, but pages are normally 2048 byte (or larger these days) with a few extra bytes per 512 for ECC. The non-ECC areas are still power-of-2 based, and the chip area itself is square and ends up being another power-of-2 pages. End result, a power-of-2. I've been working on this stuff for about 6 years now - I'm not just coming up with it randomly.
In eighteen months, we're assured.
Disk capacity is reported to my mother in powers of 2. This simply does not make sense.
Technical details should not trump users. This makes us look like geeks with a binary fetish instead of professionals.
I lost my sig.
So where does the 512 come from?
America, Home of the Brave.
That's not a concrete example. My Windows XP was over 5 GB, Vista is slightly over 6 GB on my machine. You'd think that so many years of development between XP and Vista would make it OK for Vista to require another GB or two.
The only reason you're bitching is because you really want your entire computer to run perfectly fine on a 32 GB flash drive. Let's face it, 32 GB is ridiculously small for a hard drive even if you're running a stripped down dist of Linux. Regardless of how many GB Vista requires, you will still want to fill it with music, videos and such. My archive is well over 200 GB and that excludes all my valuable PSD files and such.
Thankfully, you will be silenced in a couple of years when flash drives have started filling up the missing gaps.
Full Tilt
Minesweeper is _not_ necessary to operate a machine, even if some people will disagree.
scnr:
Just imagine the installation would give you a dialog where you could choose if you would want to install a "basic system", or to include a "desktop environment" or a "database server" etc. etc.
oh wait, then it would be linux.
You're not accounting for space taken up on the drive that can't be used for wear leveling.
Consider if the drive is nearly full then anything you write to it is going to have a limited subset of the total space available for leveling.
Unless the controller is willing and able to move data around. If I can think of it...
Nerd rage is the funniest rage.
With all the advances in memory density, I find myself wondering why they still sell movies on DVDs, and not on a chip-based memory device of some sort.
Monitor bandwidth usage on IIS6 in real-time: http://www.waetech.com/services/iisbm/
We find you in violation of Slashdot's EULA. Specifically you have.
[x] Written something reasonable and / or based in factual knowledge.
[x] Written something supportive of Microsoft
[x] Written something supportive of Vista
[x] Dissed a Lunix distro
[ ] Written something bad about Apple.
Wow, 4 out of five. You need to repent your ways and fast.
Faster! Faster! Faster would be better!
Perhaps you should get a flash media device that actually uses wear leveling. You're probably being cheap and purchasing cheap $5-$15 USB media devices, and then are curious why the quality isn't there.
OK, say a TiVo records video at an average of 4500 Kib/s and audio at 224 Kib/s. (Just a couple figures from a website about converting DVD video to TiVo.)
4724 Kib/s * 31,556,952 seconds/year = 149,275,041,248 Kib/year.
Assuming a 128 GB flash drive is actually 128 GiB (as you would assume for a storage medium that is always sold in capacities that are powers of 2 before the units)
128 GiB == 134,217,728 KiB == 1,073,741,824 Kib
and assuming perfect distribution of the rewrites and no recording had any retention rules, then over the course of one year each bit will be rewritten:
149,275,041,248 Kib/year / 1,073,741,824 Kib == 139.8 rewrites per year
Assuming NAND figures of 1,000,000 rewrites before a failure:
1,000,000 writes/bitfailure / 139.8 writes/year = 7153 years/bitfailure
that's 7,153 years of continuous recording before failure of a bit.
That sounds extremely good, but realize that that is an ideal utilization case that doesn't occur in reality. Writes get distributed, but data that isn't erased stays put, putting more wear on the rest of the space.
So let's assume a worst case is if all but 30 minutes of storage is KUID on the TiVo, leaving only enough space for the 30-minute buffer to rerecord over and over again. This would be the same as a TiVo with only enough space for 30 minutes:
4724 kb/s * 30 min. * 60 s/min. == 8,503,200 Kib
149,275,041,248 Kib/year / 8,503,200 Kib = 1,755.5 rewrites/year
1,000,000 writes/bitfailure / 1,755.5 writes/year = 570 years/bitfailure
which is still very good. And that's just a 1 GiB stick under a TiVo's load.
So I'd consider installing two 128 GiB SSDs with ATA interface in a Series1 TiVo to max out its capacity without a patched kernel. And with the right card-edge device (comparing to SiliconDust's CacheCard ability to use extra RAM as a database cache), it might be possible to not even require an ATA drive at all. Maybe even lose the fan.
Oh, say does that Star-Spangled Banner entwine / The myrtle of Venus with Bacchus's vine?