Kioxia, formerly Toshiba Memory Holdings, has demonstrated "hepta-level-cell" NAND memory:
Kioxia Researchers Demo Hepta-Level Cell NAND Flash, Nearly Doubling the Capacity of QLC
NAND researchers at Kioxia have successfully demonstrated a working concept of a new storage architecture called Hepta-level cell NAND flash. This new type of NAND can house up to 7 bits per cell, giving it nearly twice the storage capacity of QLC NAND flash. If Kioxia can stabilize this storage architecture at room temperature, it might become the ultimate successor to spinning hard drives in consumer and enterprise applications.
To create hepta-level NAND flash, Kioxia is using a new design called new silicon process technology to increase cell density, in conjunction with cryogenic cooling. New silicon process technology replaces current poly-silicon materials with a single-crystal silicon that is used in a channel inside a memory cell transistor. This apparently reduces the amount of read noise coming from the NAND flash by up to two-thirds. In other words, new silicon process technology produces clearer read signals for reading data off of the NAND flash, enough so to increase the bits cell capacity to 7.
7 / 4 = 1.75. Nearly double!
Kioxia has already demonstrated 6-bits-per-cell NAND. Multiple companies are mulling a transition to 5-bits-per-cell (PLC) NAND. Western Digital's technology and strategy chief Siva Sivaram doesn't expect PLC NAND to take off until 2025 or later.
(Score: 3, Funny) by turgid on Thursday April 06, @04:11PM (4 children)
Connect a serial cable and send it some ASCII!
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 3, Funny) by DannyB on Thursday April 06, @08:15PM (3 children)
I'm going to ASCII stupid question and get a stupid ANSI, but couldn't you send it some EBCDIC?
How often should I have my memory checked? I used to know but...
(Score: 2) by turgid on Friday April 07, @09:40AM (2 children)
Encrypted storage?
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 3, Insightful) by DannyB on Friday April 07, @02:06PM (1 child)
I believe that EBCDIC encryption predates ROT13 encryption, and is much more secure.
I don't know it qualifies as encrypted storage. Sending EBCDIC out the cereal port to another device would seem to be encrypted communication.
How often should I have my memory checked? I used to know but...
(Score: 0) by Anonymous Coward on Friday April 07, @03:08PM
I've ported code, but never cereal. Does dumping chocolate syrup on plain puffs mean I've ported them into cocoa puffs?
(Score: 5, Interesting) by Anonymous Coward on Thursday April 06, @04:12PM (18 children)
I was wondering what this "7 bits per cell" meant and looked at the Tom's Hardware link supplied--
This is starting to look more and more like "analog" to me, with a lot of careful processing needed to get the desired error-free "digital" output. To do 7 bits means one part in 128 which is easy to do with just about any modern analog-digtal (AD) converter...for one signal at a time. But holding the voltage in a tiny cell and then AD'ing at the same high cell density sounds really hard.
(Score: 5, Informative) by hopdevil on Thursday April 06, @04:25PM (16 children)
It is analog, and really always has been. Things done at the silicon level, or even tubes, are fundamentally concerned with current and voltage signals. It just appears to be digital when you look at the results. I was really hoping memristors would have become more prominent
(Score: 3, Interesting) by RS3 on Thursday April 06, @04:32PM (15 children)
My fear is long-term reliability of the data. How long can a cell hold a voltage level? The tighter we make the thresholds, the more we risk some kind of voltage loss which will change that cell's data. Will E/M fields affect the cell's voltage?
(Score: 4, Interesting) by hopdevil on Thursday April 06, @05:32PM (14 children)
Even existing technology can only hold values reliably for a relatively brief period without a refresh. Think on orders of 1 year.. Not the typical duration you'd expect from a spinning disk, for example. There are options that would be built onto the controller like error correction and redundancies.
EM totally affects them, they would also be very sensitive to temperature fluctuations too, the article talks about using cryogenic cooling. I'd call this normal and expected problems that need to be overcome, not something that would a user would need to worry about.
QLC isn't quite ready for mainstream, but it is looking very promising for it's density and it is getting closer to reality.
(Score: 3, Insightful) by DannyB on Thursday April 06, @08:16PM
Can't we have a mod called: Frightening ?
How often should I have my memory checked? I used to know but...
(Score: 2) by RS3 on Thursday April 06, @08:56PM (12 children)
OMG! I'm with DannyB on this- frightening! Does anyone make a refresh utility? I remember the good old days of hard disks where you actually "low-level" formatted them- re-magnetized everything. I used SpinRite all the time- it would "lift" (copy) a track, re-low-level format, test, test, copy back the data, test some more with ECC off. Can't low-level format any more. Why are we going backwards? Why is cheap density more important than data integrity?
(Score: 4, Informative) by Rich on Thursday April 06, @11:09PM (9 children)
I suspect the "refresh utility" is already built in and runs when idle. A while ago, there was a case of Samsung SSDs massively losing performance if they hadn't been on for a while, where some priority scheduling must have messed up. Or rather given the refresh too much priority. They "fixed" that with a firmware update. I don't know about the intricate details of SSD storage, but I suspect they store known reference levels near the data and use that to guess the data levels on readout and to decide when to "refresh", which would rather be a relocation of the deteriorating data with a fresh new write and then erasing the old location for further use.
There is little information of neglected storage media losing its data, but the consequences of the above mean that you would have to plug in "modern" high capacity solid state media, like once a year, for a day to let the refresh do its job.
(Score: 2) by RS3 on Thursday April 06, @11:43PM (8 children)
Thanks. How about USB FLASH? Just powering them on for a while should do it?
If that's true, what happens if you de-power it (including SSD) while it's doing some refreshing? Corruption / data loss?
If that's true, is there some way to know it's churning, and when it's okay to turn it off?
Some of mine have an activity light, like good old floppy drives. I always wait for that to go dark, then do the "safely eject" thing. Some of my USB drives have an LED that's always on. Not helpful. Many have no indicator.
I think I'm going to dig out my old full-height 5.25" ESDI drives... I'll have the biggest heaviest laptop on the planet. :)
(Score: 3, Informative) by Rich on Friday April 07, @12:02AM (7 children)
I have no idea and I'm just guessing based on what I've read so far. The manufacturers aren't going to share their dirty secrets as long as sales continue and they avoid a shitstorm.
As for power loss, don't worry. If they relocate, they'd write the bulk data first and then remap by writing a few bits where they likely have enough energy from the bypass capacitors for a microsecond or two.
I'm curious myself how long a USB stick (both top brand/top performance, or cheapo supermarket variety) or an SSD can hold their data in powered-off state. In addition to what I wrote, there most be some error correction going on and they use statistics to, "guarantee", say, one permanent error in 10^17 bytes, or one bad data retrieval in 10^18 bytes written or so. We also know that trapped electron storage doesn't like high temperatures, so your stick at constant 18 degrees in the basement could last 10 years, while the same silicon in a hot car would lose data in a matter of months.
(Score: 4, Interesting) by deimtee on Friday April 07, @01:36AM (6 children)
I have an old car stereo that can take USB sticks and play music on them. A random collection of sticks have been sitting with it for about five years without being used. In the interests of "Science!" I just went and got them.
Results:
Lexar 2GB Mini - Data is ok. (music still plays)
Lexar 512MB Jumpdrive - Data is ok.
No-name 1G - Data is ok
No-name 1G - Data is ok
SanDisk Cruzer 128MB - Data is Ok
Co. Promo 2GB - Data is Ok.
These were not in optimal storage, they were actually in the old car which is parked out in the open. (climate range is occasional frosts in winter, summer up to 40C)
No problem is insoluble, but at Ksp = 2.943×10−25 Mercury Sulphide comes close.
(Score: 0) by Anonymous Coward on Friday April 07, @08:14AM
Of course if the SSD is made by some random Chinese manufacturer it may not even last 6 months unpowered.
(Score: 0) by Anonymous Coward on Friday April 07, @01:28PM (2 children)
> No-name 1G - Data is ok
> SanDisk Cruzer 128MB - Data is Ok
So what's the difference between "ok" and "Ok"?
Ok almost looks like 0k (zero kbytes = dead).
(Score: 2) by deimtee on Saturday April 08, @12:06AM (1 child)
No difference. I just wasn't very careful with the caps key.
The music on all the sticks played. It was all MP3s, I don't know how many bit errors it takes to make an MP3 not work.
Funny thing is I remember buying that one, I needed it for a class. It's over 20 years old and cost me AU$98
No problem is insoluble, but at Ksp = 2.943×10−25 Mercury Sulphide comes close.
(Score: 0) by Anonymous Coward on Saturday April 08, @12:33AM
https://images.highspeedbackbone.net/skuimages/large/S153-2024.jpg [highspeedbackbone.net]
Good times.
I remember when 1 GB was a hot Black Friday item. Now Micro Center gives out 32/128 GB like free candy.
(Score: 3, Informative) by RS3 on Friday April 07, @03:24PM (1 child)
My simplistic hunch is the older ones were lower bit-density, and less subject to data errors due to charge loss.
The more data per cell, therefore the more voltage-levels per cell, therefore the finer they must define the voltage level thresholds (smaller voltage difference between bits), the smaller the margin of error, so the greater the chance of error.
I think.
But maybe and hopefully they've found ways of improving reliability?
(Score: 3, Informative) by deimtee on Saturday April 08, @12:17AM
I think they are all old enough to be single bit per cell. Definitely the 128MB Cruzer is.
No problem is insoluble, but at Ksp = 2.943×10−25 Mercury Sulphide comes close.
(Score: 3, Funny) by DannyB on Friday April 07, @02:10PM (1 child)
Because speed is substitut fo accurancy.
How often should I have my memory checked? I used to know but...
(Score: 3, Funny) by RS3 on Friday April 07, @03:11PM
> Because speed is substitut fo accurancy.
You wouldn't say that if you heard me play guitar! But I shred on...
(Score: 4, Informative) by Anonymous Coward on Thursday April 06, @04:27PM
(Pssst! I'll let you in on a little secret: all electronics is analog! No really, it is! Someone decided what voltage level constitutes a one or a zero, and the circuits, in analog (linear) fashion look at a signal's voltage level, and in analog circuitry, "decide" if they're seeing a "1" or a "0", and pass on a signal that's above or below some predefined voltage thresholds. It's all about analog voltage levels. These multi-level memory cells are simply an expansion of the "1" or "0" concept.)
(Score: 3, Funny) by Snotnose on Thursday April 06, @04:28PM (6 children)
They call the new design "new silicon process technology". Damn, wonder why nobody thought of that one before.
I know! I'll call this "new post". I'm a genius. Genius, I tell you!
I just passed a drug test. My dealer has some explaining to do.
(Score: 4, Funny) by Tork on Thursday April 06, @04:45PM (2 children)
Making fun of the quality of AI-generated blog posts is just poor sportsmanship.
Slashdolt Logic: "25 year old jokes about sharks and lasers are +5, Funny." 💩
(Score: 0) by Anonymous Coward on Thursday April 06, @09:06PM
(Pssst: he's an AI too. Just sit back and watch the show. I'll bring popcorn.)
(Score: 3, Informative) by DannyB on Friday April 07, @02:17PM
Let's ask Mr. Owl, he knows everything . . .
Q. Is it bad sportsmanship to make fun of AI generated blog posts?
How often should I have my memory checked? I used to know but...
(Score: 3, Funny) by turgid on Thursday April 06, @05:26PM (2 children)
I would like to invest a very large sum of money in your brand new venture! You sound like a clever fellow.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 1, Funny) by Anonymous Coward on Thursday April 06, @06:20PM (1 child)
> You sound like a clever fellow.
Have you confused clever with congested? (see login name).
(Score: 2) by DannyB on Friday April 07, @02:19PM
confused?
constipated?
How often should I have my memory checked? I used to know but...