New "PRAM" 30 Times Faster Than Flash

hairyfeet writes, "The EETimes describes the new Samsung memory, phase-change RAM, called PRAM. Samsung is dubbing it 'Perfect RAM' because it is thirty times faster than NOR flash, ten times more durable — and cheaper to produce, to boot." 512-Mbit modules should be available sometime in 2008. None of the initial coverage goes much beyond Samsung's press release. At the same time, Samsung also announced a 40-nm, 32-Gbit NAND flash device.

Comments

Advantages

[duerra]

It sounds like great marketing and all using the name "PRAM" for "Perfect RAM". However, can anybody tell me what, if any, advantages that this design has over MRAM? I'm all for a replacement of flash, given all of its disadvantages, but I would like to avoid a format war if one format between these two is clearly superior to the other.

Re:Advantages

[jimstapleton]

I didn't see it in this, does PRAM not have the read/write cycle limitation of flash? I hope so...

Re:Advantages

[P3NIS_CLEAVER]

it mentioned '10 times more durable'? Maybe this means that it will go bad at a rate 1/10th of regular flash.

Re:Advantages

[jimstapleton]

I was thinking more along the lines of physical durability with that one. Could be that though. If that's the case, I'd really want 1000x more durable, or more.

Re:Advantages

[kippers]

Atleast you would be able to add content to either formats and transfer at will. Unlike the whole Blueray thing...

Re:Advantages

[Firehed]

This PRAM won't get fucked up if it gets near a magnet? That's a rather important thing for portable data, since just losing a floppy's worth was enough of a pain not so many years ago, and we'll be talking at least three orders of magnitude more storage.

That, and MRAM was supposed to come out and have replaced DRAM entirely something like two years ago. So at this point, I'd effectively dub it vaporware. Of course this PRAM might end up in the same state, but it's not overdue yet.

In either case, I wouldn't be concerned about a format war. If they're used for portable storage, it's all just USB. And if it's for internal stuff (as I think MRAM is supposed to be, unlike PRAM by the sounds of it), expect there to be a standard in place - AMD and Intel don't want to have to force another memory change too quickly, seeing that they've barely moved all of their new chips to DDR-2 use (well, Intel has for a while, AMD less so).

Re:Advantages

[Forge]

To expose your USB drive to a magnet and not louse data what you need isn't a better memory technology inside. Fact is as long as you are rearranging electrons to store data your data is vulnerable to magnets.

What you need to achieve that is better shielding. I can't remember the brand but I once left a 32 MB thumb drive stuck to the magnet of a 15" Pile, speaker. None of the 20 MB of data on it was damaged.

This stick had a Metal shell. Perhaps someone who knows more than I do about memory technology and Physics could give a detailed explanation of my observed results.

Re:Advantages

[P3NIS_CLEAVER]

the wafer are thicker in the new memory but I have never heard about packaged semis breaking. Perhaps this will make them easier to manufacture.

Re:Advantages

[noidentity]

I'm all for a replacement of flash, given all of its disadvantages, but I would like to avoid a format war if one format between these two is clearly superior to the other.

A format war occurs when there are two or more incompatible interfaces for accessing data on media. Since these chips would be the guts inside a thumb drive or other medium using a standard interface, there would be no format war, just a new choice of implementation technology that could be adopted as slowly or quickly as desired.

Re:Advantages

[networkBoy]

Maybe, but thermal durability is not there.
Phase Change Memory (PCM) is thermally unstable (double edged sword). Durring reflow the temp is high enough to blank the device, so no factory programming. Devices will have to be in-circuit programmed.

Good side: increminating evidence? just place in toaster or other suitable device and set on high. problem solved.
-nB

Re:Advantages

[networkBoy]

In the case of flash, that speaker was not nearly powerful enough.
Flash uses a floating gate, and the electron tunneling effect to program (charging the gate). This requires roughly 9-12 volts, so there is an internal charge pump. To disturb those electrons on the floating gate, while not actively programming, would take well over 10K Gauss, likely in the 100K Gauss range. Don't have a powerful enough magnet to prove anything with, but the 5K Gauss one didn't do squat to a part.
-nB

Re:Advantages

[Firehed]

My point was that if you replace that flash memory with magnetic storage, it's much more prone to becoming fubar, a la floppies. You can't easily damage a hard drive with a household magnet (the strongest magnets I own came from inside a hard drive, and a pretty old one at that), but something likely to be as exposed as a USB stick might not be so hard-wearing.

Re:Advantages

[Jeremi]

I would like to avoid a format war if one format between these two is clearly superior to the other

I don't see why a format war is inevitable... different hard drives also use different storage technologies, but as long as they all provide the same interface (e.g. SCSI or IDE) nobody needs to know or care what sort of magic is going on inside.

Re:Advantages

[joto]

A format war occurs when there are two or more incompatible interfaces for accessing data on media. Since these chips would be the guts inside a thumb drive or other medium using a standard interface, there would be no format war, just a new choice of implementation technology that could be adopted as slowly or quickly as desired.

Yeah, exactly like you can use both RAMBUS and SDRAM in most computers. And they aren't even that different.

And nobody considers the differenses between IDE, SATA, or SCSI (and all of the slightly different SCSI connectors), fibre channel, USB, firewire, or whatever the fuck it is important. Because nobody is interested in just attaching a disk, they want to understand the politics of interface standards.

Errata + Info + Opinion

[AKAImBatman]

1. Someone's getting their MBits and GBits mixed up. Samsung has announced 32-gigabit (4GB) flash chips, not 32 megabit.

2. NAND has traditionally been shunned in many uses because it can usually only be accessed as a block device, and not a standard ROM device. Which makes it unsuitable for many embedded applications. Thus this chip is probably targetted at the thumb drive market.

3. This is exciting stuff! According to the article, PRAM is supposed to have processing speeds similar to RAM, and does not require erasure or sectoring. The only downside is that they don't give any hard figures on what "fast processing speed" means. Depending on what that actually means, we could start seeing machines that are able to instantly hibernate like EROS, but without the added step of writing to disk.

4. The 512 MBit (64MB) device may sound small, but I imagine that more than one chip will be chained together to create a larger storage device. Samsung will probably also work to produce larger chips once they have all the early production issues worked out.

5. The CIO article is already slow, so I'll add one tidbit they had. According to CIO, Samsung is considering PRAM to be a good fit for replaceing Flash memory in mobile phones. Considering the lower price, this could be a good fit. The only question is, does it use more or less power during read/write cycles?

Re:Errata + Info + Opinion

[Manchot]

Also, what exactly does "three-dimensional transistor structure" mean? All major transistor structures are three-dimensional. In fact, I don't know how you could even make a useful device with photolithography that isn't three-dimensional.

Re:Errata + Info + Opinion

[AKAImBatman]

Also, what exactly does "three-dimensional transistor structure" mean?

Usually it means that the circuitry is layered or "stacked" in a 3D matrix rather than the traditional "flat" 2D matrix. This means that you can cram more parts per square centimeter because your circuits have depth as well as width and length.

Re:Errata + Info + Opinion

[mabhatter654]

no, current chips are 2.5D. basically they're just a series of circut "maps" laid on top of each other. Not much is different from circuit boards except they're a lot smaller. Look at it this way, no matter how many "maps" you pile on top of each other, you'll never have the simplicity of a globe, or a stack of blocks. Some of the new stuff does actual processing in the 3rd dimension.. in other words components and processing are stacked thru the layers, touching directly, which something that most chips still don't do.

Re:Errata + Info + Opinion

[jfengel]

One problem with stacking is heat; the transistors give out heat and if there's another transistor on top of it also giving out heat there's no place for that heat to go, eventually melting/burning the chip.

In processors and ordinary refreshing RAM, the transistors are continually being exercised. I wonder if part of the solution in this case is that stable memory exercises the transistors less often, giving more time for heat dissipation. You'd risk overheating if you continually read the same block, though. Perhaps tie it to a cache; stuff you've read more than once stays in a flat easily-cooled RAM cache until the primary store has cooled.

Re:Errata + Info + Opinion

[kdawson]

Gbit erratum corrected, thanks for spotting.

Re:Errata + Info + Opinion

[AKAImBatman]

corrected, thanks for spotting.

No problem. :)

Welcome to the list of Slashdot editors, BTW. Don't let the anti-editor crowd get you down. ;)

Re:Errata + Info + Opinion

[jZnat]

What's with the wrong use of data capacity measurements? n-bit is only useful for network speeds (historical I'm assuming since 1 baud = 1 bps), encryption, and very small data sizes (e.g. CPU register sizes; places where every bit counts). The article should use n-byte measurements since it's measuring storage...

Re:Errata + Info + Opinion

[grammar fascist]

no, current chips are 2.5D. basically they're just a series of circut "maps" laid on top of each other.

Which of course introduces limitations such as:

1. You can't put one ledge on top of another, which limits specific types of gameplay

2. Non-map opjects are sprites, which don't look so great, and dead bodies tend to rotate on the floor when you're not looking (which is a little unsettling)

On the plus side, you can generate maps using only a single 2D or stacks of 2D blueprints.

Re:Errata + Info + Opinion

[StressedEd]

..what "fast processing speed" means

I assume it means time taken for a cycle of read/write/erasure. If it's based on Chalcogenide phase change materials (typically found in optical media) I suspect the timescale for writing and erasing will be ~10s of nS.

Re:Errata + Info + Opinion

[AKAImBatman]

What's with the wrong use of data capacity measurements?

Memory parts are always referred to in bits rather than bytes, since the actual storage is usually at the bit level. Bytes are only used once the device has been packaged, and only accepts requests on a byte, word, or dword basis. So it's not really incorrect. You're just unfamiliar with the terminology. Do a bit of circuit design and you'll hear a lot more numbers given in bits rather than bytes. :)

Re:Errata + Info + Opinion

[CODiNE]

The only question is, does it use more or less power during read/write cycles?

Well considering it's PERFECT RAM... the power consumption had BETTER be less.

Re:Errata + Info + Opinion

[cyborg_zx]

My 2.5D has voxels you insensitive clod!

Re:Errata + Info + Opinion

[orasio]

That's why Quake rocks, and Doom II doesn't.

Re:Errata + Info + Opinion

[jZnat]

Oh, that makes sense. I'm into computer science much more than electrical engineering.

Re:Errata + Info + Opinion

[rkanodia]

1 baud != 1 bps. Baud is the signaling rate in events per second; a signal event can represent more than 1 bit of data. If you have a system that signals 10 times per second, with each signal encoding 4 bits of data, you have a system that is 10 baud, and 40 bps. Using baud in place of data rate is a common misusage. See the wikipedia entry on Baud for more information.

Re:Errata + Info + Opinion

[networkBoy]

Reading produces almost no heat in the array (the stacked part) the perifery, where the heat is generated is flat, so still no issues.
Writing would have to be limited, as heat is used to cause the phase change much like a CD-RW laser heats the media to effect the change in it. In fact IIRC the compound is the same CD-RW and PCMemory.
-nB

Re:Errata + Info + Opinion

[WuphonsReach]

Samsung has announced 32-gigabit (4GB) flash chips, not 32 megabit.

Considering that I can get a 4GB SD flash card for $60, I'm guessing that those 32 gigabit chips have been out for a while? Or do those tiny SD cards have 2 chips in them?

Re:Errata + Info + Opinion

[AKAImBatman]

Or do those tiny SD cards have 2 chips in them?

They most likely have more than one chip in them. Look at a stick of RAM sometime and count how many chip packages are mounted on it. You'll usually find more than one chip.

Probably the craziest example of this was the SuperPaint Framebuffer which was spread across 16 circut cards, each populated with TONS of memory chips.

If I'm doing my computations right, he needed 1215 chips to complete his project! I hope he got a bulk discount. :P

Re:Errata + Info + Opinion

[WuphonsReach]

I guess there's room in a SD card for multiple chips. It's just that the SD Card is the size of a postage stamp with the upper 1/4 consumed by the contact points. I guess as long as the 16Gbit chips are less then around 0.8 cm^2 then there's enough room inside the packaging for a pair of them.

Re:Errata + Info + Opinion

[AKAImBatman]

The actual silicon is tiny. For example, Samsungs 1GB NAND chip is a mere 146 square millimeters, or approx. 12mm x 12mm. The real space consumption on a board is always in the chip packaging, which is often sized to surface mount or slot into a board. So the 32mm x 24mm packaging of SD cards should pose little challenge to placing more than one chip, assuming the traces are etched directly into the SD card. Considering their size, I see no other way they'd do it.

Re:Errata + Info + Opinion

[mabhatter654]

you realize they are talking about actual silicon chips here and not games right?

Good

[MindStalker]

Cause I'm sick of how long it takes to load those damn Flash ads that hang over my browser window.

Re:Good

[ElephanTS]

Heh heh - good point. And don't forget to reset the PRAM hold down ++P+R and wait for the double bong.

(personally I'd be happy with one bong but some people are greedy that way)

Re:Good

[Aladrin]

You do realize that now you could have 30x as many of them before your computer slows to a crawl. I somehow think that you should be unhappy about this.

(Yes, I do realize he is joking. So am I.)

New tech? We still get stuck

[Trashhalo]

cheaper to produce, to boot.
Yea but since it is "better than flash" they can charge more for it and just get much larger profits. I hate hate hate how companies throw on "new technology" surcharges even when it costs less than the old tech to produce.

dont mind me I am just bitter today.

Re:New tech? We still get stuck

[mabhatter654]

It's business... they are selling something that does more for only a little more money. That makes them a tidy profit!!! That's what they develop new chips for is PROFIT. In another year or two, the prices will drop to be lower, or we will have devices so large we couldn't have afforded them at all... i.e. price of a 4GB SD card isn't going down any time soon due to complexity of manufactureing.. but the new chips may make 4GB as affordable as 1GB sooner.

Re:New tech? We still get stuck

[Trashhalo]

What I was more thinking about is how the record company passes high costs of production to the consumer but keeps the profits from low cost of production. For instance when the cd technology came about they said that prices would be high for awhile because of how much it costed to produce cds at the time. So there is a example of the consumer paying more for cost of production. Yet when itunes and friends came about and the cost of production went to almost nil (bandwidth) did we see a decrease in the price of a album? Nope. You can get a CD at the store for 12$ and yet a album on itunes comes out to about that also. Where is the money going from the reduced cost of pressing the music? Into the RIAA's pockets.

You can also seen this done with hybrid engines costing more yet not costing more to make. Though I dont know enough about that to make a solid arguement. Feel free to shred what I just said.

But I thought my original post was more offtopic than troll. but okay.

Re:New tech? We still get stuck

[Jonny_eh]

You can attribute the higher price to the R&D costs.
Why should a company invest in R&D? To get a return on investment!! (and to stay competitive)

Re:New tech? We still get stuck

[Znork]

CDs and music in general are subject to monopoly rights, and thus have no actual competition. Prices on them will rise to absorb any economic elbow room available, and even worse, the cost structure will adjust to the revenue, essentially negating the entire purpose of a free market economy.

Memory on the other hand is interchangable with other memory, thus subject to a much more fierce competition, which drives prices down.

Re:New tech? We still get stuck

[thefirelane]

That's what they develop new chips for is PROFIT

Its funny how people (like the grandparent) usually don't get this. They expect companies to develop new tech and loose money doing it.

Interestingly, I've never met someone who says this and pays to go to work... that would be sticking to your guns.

Re:New tech? We still get stuck

[rbanffy]

New semiconductor technologies consume a lot of cash to develop, test and put into production so then, maybe, they can cover development costs. Most of them never reach this last stage and with luck, become patents so their owners can sue whoever solves the problems they didn't.

Re:New tech? We still get stuck

[Firethorn]

You can also seen this done with hybrid engines costing more yet not costing more to make. Though I dont know enough about that to make a solid arguement. Feel free to shred what I just said.

Yet a hybrid DOES cost more than a non-hybrid of the same power, especially when you consider the whole fuel train. Going to a hybrid power solution adds complexity, additional materials, etc... Right now that cost is $3-5k for automobiles.

As for this tech, even if it's ultimately cheaper to produce, you're still going to have retooling costs for the factories, so the price will be initially higher, but then, like what was said, this technology is ultimately transparent to the end user. So in order to be able to command a price premium it'd have to be able to provide something current chips don't have: IE faster speed/higher capacities. So your nice new 32 GByte USB key will use this tech at a price premium, then it'd trickle down until it pushes the older, less efficient technology into the waste bin.

Re:New tech? We still get stuck

[Trashhalo]

>>They expect companies to develop new tech and loose money doing it.

Oh I get the parents logic but I don't agree with it. Lets say there are four companies producing flash chips. Your company is one of them. For sake of my arguement lets just say the market is split between all four companies. Your company then develops a alternative to flash that is better on all benchmarks. You have patented everything so you are the only company able to produce this new technology. Due to the fact that it is so much better than classic flash usage sky rockets until it kills off the older technology. Instead of the market being split between four players now the market is stuck on one.

Now you are saying that the fact that your company has a stranglehold on the market is not adequate compensation for research. *shrug* I disagree which was the point of my post.

Re:New tech? We still get stuck

[mabhatter654]

you almost get it... Yes, you have the new technology with a patent, so in a case like this though, it is CHEAPER for you to make the new stuff! You are still in a market with the three other companies that have to charge $x to make a profit. Instead of under cutting them, because you couldn't make that many anyway, you market yours at a small premium.. because it's better... split up the market, and make more profit than the other guys... everybody wins! Customers don't loose because the other makers will keep the new guy from gouging too badly.. after all, you can't get the product to market quickly unless it's compatible with the connectors already out there... so it's still a commodity.. the time you enjoy "premium" status is short.. in the tech industry measured in months, not years!

Ob. Python

[TexasDex]

I have to push the PRAM-a-lot!

A pram that is 30 times faster?

[bmo]

Since a pram is generally used at walking speed, doesn't anyone think of the baby's safety when it's rolling 30 times faster, i.e., 90 miles/hr or 145kph?

Think of the children!

--
BMO

Re:A pram that is 30 times faster?

[vistic]

Babies have important meetings to get to.

The wall street babies get Hummer prams.

Re:A pram that is 30 times faster?

[tverbeek]

What amazes me is that this pram is going to be faster than the Flash (the fastest man alive)!

Re:A pram that is 30 times faster?

[Peldor]

Think of the hang time!

The XGames participants just keep getting younger and younger.

Re:A pram that is 30 times faster?

[cschmidt]

"I like to push the pram-a-lot!"

This could be confusing for Apple.

[jZnat]

PRAM is Parameter RAM in Apple Macs, and that's a vital part of the computer (well, logic board). Maybe Apple will start using PRAM for its PRAM?

Re:This could be confusing for Apple.

[Pahroza]

And zapping PRAM just became more fun. Toss it in the microwave instead of having to hold down 15 (exaggeration) keys at the same time in some sick attempt at becoming Digital Twister Champion.

"perfect" ?

[Stavr0]

With all our luck, it's probably going to turn out to be destructive to the ozone layer or cause cancer in lab rats. Whenever someone discovers something amazingly useful, it always turns out to have negative side effects. I believe that's called the Law of Unintended Consequences.

Sorry to rain on the parade.

Re:"perfect" ?

[Morphine007]

Whenever someone discovers something amazingly useful, it always turns out to have negative side effects. I believe that's called the Law of Unintended Consequences.

There's a belief that women are starting to get bigger breasts as a result of the various hormones that are pumped into chickens to give them bigger breasts.... all this to say: the Law of Unintended Consequences does not only refer to negative side effects.

My sources for this belief are strictly anecdotal and generally as a result of seeing 12 year olds with C-cup sized chests... the parents of these chil'uns would probably agree with you about the side effects being negative though ;-)

Re:"perfect" ?

[jafiwam]

My assumption is that if there is a difference (who says there is? with all the sizing changes in women's clothing and style differences it could be just that) was due to the fact that at 12 years old, the average Westerner kid has eaten 3 times as many calories than other kids.

More calories and more fat and more protein equals more stuff built earlier.

Re:"perfect" ?

[Petaris]

Is that a subset of Murphy's Law? :P

Re:"perfect" ?

[JeanPaulBob]

Whenever someone discovers something amazingly useful, it always turns out to have negative side effects. I believe that's called the Law of Unintended Consequences.

Yeah, like when they figured out how to replace vacuum tubes with transistors--only to discover that after 60 years, all transistors mutate into rampaging insectoid machines that think of nothing but murder all day.

tweak the p-ram and reglaze the subroutine

[Bloodwine]

"So l toasted the dated directory, tweaked the P-RAM ... and reglazed your subroutine."
- Pauly Shore

Pauly Shore

[RudeDude]

Your PC got harshed, right, 'cause your
system heaps at the wrong parameter.
So l toasted the dated directory,
tweaked the P-RAM...
and reglazed your subroutine.
  - Crawl, in "Son In Law" (Pauly Shore)

Re:Pauly Shore

[Quiet_Desperation]

In some parts of the world, quoting Pauly Shore can get you shot.

Re:Pauly Shore

[Aladrin]

I don't know what's funnier... That they really said that, or that you remembered it.

Re:Pauly Shore

[roman_mir]

and then we will hear about it on /.
twice.

Re:Pauly Shore

[foniksonik]

Shootin' the WEEEEEEEEEAZELLLLLLLL..... chkcchckhckhckchck

Re:Pauly Shore

[mgblst]

Yeah, we are really sorry about that. We are trying to get it pushed to every country.

Pessemist

[Cybert4]

I doubt it will do either of those. I'm more worried about lifetime and corrupt bits.

Overall system performance

[grumpyman]

What I really want is not more super-duper fast RAM but real cheap and abundance of FLASH - to replace (at least partially) hard-drive. Not only performance but reliability will be better for a system overall.

Re:Overall system performance

[EvilMoose]

Um. NAND has a maximum amount of writes available before it's broken.

Re:Overall system performance

[Overzeetop]

But there is a limit beyond which it is effectivly unlimited. How many write cycles do you get, and if you turn off your swapfile what kind of life could be expected? If it's greater than 5 years for 90+% capacity, the point may be moot. (yes, I've turned off my cache. If XP can't fit what I'm doing today in 2GB, I trying to do too much at once.)

How awesome

[inviolet]

Just when we start to worry that we've hit a wall in our attempts to grow, somebody goes and cuts a completely unexpected doorway in it.

Same thing happened with telecommunications vis-a-vis wireless.

I'm looking forward to something similar happening with domestic power generation and distribution.

Proof is in the pudding... when it arrives.

[KokorHekkus]

Some day this might very well be the next big thing. But I wouldn't take the manufacturers word as the gospel for when it's going to arrive. And it really has to be cheaper to replace flash memories en masse. Flash is good enough for phones, memory cards etc. The new tech doesn't solve any obvious fundamental flaw (100 000 read/write cycles is still enough for most applications)

Compare with SED displays. We all want them (if they're not much more expensive than our current ones). But it's been slipping a bit. For now it looks it will arrive in 2008 in the mass markets according to Canon chairman Mitarai (Canon and Toshiba is working together on this). And I don't think he's being conservative. (The reason given for the delay is that "we have not yet established the manufacturing technology for mass-producing SEDs at low cost," Source: http://www.eetimes.com/news/latest/showArticle.jht ml?articleID=192700864.

Which Flash?

[JoshDM]

Jay Garrick I can see, and maybe Barry Allen and Bart Allen, but Wally West was boo-koo speedy.

What's next, "Super-Duper" RAM?

[Jabrwock]

PRAM is Parameter RAM in Apple Macs, and that's a vital part of the computer (well, logic board). Maybe Apple will start using PRAM for its PRAM?

Yeah I was kinda wondering about that...

Besides, isn't "Perfect" RAM a little arrogant? What do they call the next stuff, "Super-Perfect", "Ultra-Perfect", "Mega-Perfect", "Super-Duper"?

Re:What's next, "Super-Duper" RAM?

[Jabrwock]

What do you want them to call it? PCRAM? You don't think that will confuse anyone?

Why not, that's what it's already being called, isn't it?

How is this different from "other" Phase Change RAM?

Re:What's next, "Super-Duper" RAM?

[karnal]

No, it'll be Super-Perfect RAM 2.0 E-X Plus Alpha.

Re:What's next, "Super-Duper" RAM?

[Jabrwock]

Because when you say it, it sounds like you are referring to everyday RAM for your PC.

True, but I don't see an issue though, since Joe Average isn't likely to be upgrading his "PCRAM". Someone who knows enough about computers to know about the existence of actual PCRAM won't likely be confused.

But I guess since one group already calls it that, everyone else is powerless to call it anything other than that, right?

Not really powerless, just silly. Why invent new names when it already has one that doesn't conflict with other industry-used terms?

I suppose if people feel wierd calling it PCRAM, then they can use Phase RAM (just like they still call the other kind "Flash RAM")

This is Slashdot

[Anonymous Coward]

Everyone here should have a general idea, what Flash-Memory is. PRAM, on the other hand...

Posting anonymously, as otherwise it would fall under the term of "Karma whoring": "Often these will be needless information (such as a link to a Wikipedia article relevant to the subject being discussed)..."

Sorry, for being so blatantly direct...

RAMming Speed

[Doc Ruby]

More immediately, this announcement means that Flash prices should start dropping even more. Maybe not to $0.25:GB like desktop HDs today, but maybe $2:GB (instead of today's $16:GB), for mobiles. When they release PRAM devices in 2008, they'll sell them for much more money than Flash because of the speed, but start dropping prices faster because of the cheaper manufacturing. By the time PRAM costs $2:GB, they'll be dumping Flash for $0.25:GB. HDs will probably cost $10:TB, if they're used outside datacenters at all, maybe sometime around 2010-2.

Re:RAMming Speed

[swordfishBob]

if they're used outside datacenters at all, maybe sometime around 2010-2.

HDs will be used for a while yet. Someone will find a reason. In the early 1990's when "FeRRAM" was announced (well, when I saw someone looking for funding on it) I remember postulating with a colleague about the possibilities: Operating Systems that can suspend instantly, applications that don't have a separate "install" and "run" - you just load them and they're there - probably more like "mounting" the app, laptops that run a long time on battery. What's happened since then? Palm pretty much did that with Static RAM as the chips got cheaper. Desktop PCs went from 16MHz, 2M RAM and 40M HD to at least 1000x on all three counts. Yet I'm still mostly using email, web, word proc and spreadsheet, which worked ok back then too..

Re: New "PRAM" 30 Times Faster

[mybecq]

New PRAM 30 Times Faster

Go BABY, Go!

pram

[CrazyJim1]

pram powns 'puters

goodbye hard disks?

[petes_PoV]

I got a new phone recently that came with a 1GB microSD card the size of a fingernail. It made me think back about 15 years to when 1GB hard disks were new. SCSI interface, 3.5 inch, high power consumption, hot, noisy and cost a thousand quid,

c.f. 18 for the phone memory.

Maybe if (a big if, I know) this technology does actually deliver, we may see the same kind of tech. leap in the next 15 years.

By then the biggest problem would be losing - literally - all your data.

That's progress.

Re:goodbye hard disks?

[DeadboltX]

"It made me think back about 15 years to when 1GB hard disks were new"

They were actually out quite a bit earlier than that
Wikipedia's article on Hard drives has some interesting timeline information
1982 - Hitachi 1.2 GB H-8598 consisted of 10 14-inch platters and two read-write heads

Not exactly a laptop sized hard drive though..

Re:goodbye hard disks?

[birder]

It doesn't seem that long ago to me that paying $200 for a 4GB stick of 72pin RAM was a good deal or that hard disks were $1/MB and 250MB was the max size.

Whenever I think of these, I have no problems justifying buying tech stuff today =p

And no doubt in 15 years capacities will be bigger, hardware smaller, and prices roughly the same as we pay now.

Re:goodbye hard disks?

[Ant P.]

Only 15 years?
Someone gave me a bunch of obsolete hardware a few years ago. Among it were three 1.5GB hard disks from 1980. 5¼" full height, SCSI, and heavier than a brick. And they had heatsinks.

In this case, format war can only help you.

[Vellmont]

The format war would take place with manufacturers and memory producers. Consumers will likely never install pram or mram into their computers just like they don't install flash now. The only time you might buy pram would be a USB flash device, and that already has the standard interface of USB. So really there's no reason consumers would ever have to choose one technology pathway or another. Manufacturers would have to choose, but they're much better prepared to switch technologies and avoid having dead-end technology (they really do this all the time when it makes economic sense).

Re:In this case, format war can only help you.

[rwven]

Consumers don't install flash into their computers now because Flash isn't all that fast... It's (much) faster to have DDR SDRAM than it is to have Flash...and for the purposes that general computer ram is used for, it works great. While it would be nice to have a very fast non-volitile solution, there just isn't anything currently being made in volume at a price worth paying that can compete with the likes of DDR, etc.

MRAM may be the solution to that problem...which is EXACTLY why it will probably make it big once the processes to produce it at cheaper than they are now. It's faster than DDR (much faster from what they say), non volitile, and it has a life-time comperable to any great SDRAM that you can find out there.

By the admission that PRAM does have a limited life time, although longer than Flash, they pretty much are telling consumers that it's not something they would use to replace their hard drive or ddr... Anything that is going to start dying after it gets used a (relatively fixed) certain number of times is not something people are generally going to be interested in. When a person is at a store and a salesperson tells them "this computer has a solid state hard drive which is really fast but will randomly start losing data and dying later in life" do you think many people are going to be interested?

Re:In this case, format war can only help you.

[Air-conditioned cowh]

When a person is at a store and a salesperson tells them "this computer has a solid state hard drive which is really fast but will randomly start losing data and dying later in life" do you think many people are going to be interested?

The operating system most people buy does this but it doesn't stop them!

Re:In this case, format war can only help you.

[joto]

Wow, by your logic nobody would have bought digital cameras, mp3-players, or USB flash ram. And the major harddrive vendors wouldn't be thinking about combining traditional HD technology with flash ram.

Does that tell you that you might be a bit out of tune with reality?

Re:In this case, format war can only help you.

[JunkmanUK]

I would argue against that slightly. There's a distinct difference in approach to digital cameras, mp3-players and flash RAM, mainly because these devices are all portable and people feel safer knowing that they have our files safely ensconsed on the PC at home. People don't consider these portable devices as being failsafe, primarily because they're easy to lose but also because they're so small people wrongly associate this with being fragile.

The fear of portable media devices being prone to data loss over a long period of time is an irrelevance to the user as they regard the whole portability issue as a general risk. If you were to tout this within the PC hardware and didn't mention the risk of data loss the users would be bringing lawsuits for 'emotional damage of losing pictures of litle Johnny's birthday', and if you DID mention the data loss issue then the average customer would not accept it - "My old PC didn't lose data after x months! and this is meant to be _new_ technology?!"

Sadly a large percentage of PC users are still stuck trusting the hardware to store their precious possessions with no questions asked, but that's a whole other thread...

Re:In this case, format war can only help you.

[Vellmont]

Anything that is going to start dying after it gets used a (relatively fixed) certain number of times is not something people are generally going to be interested in.

That's actually true of any technology. Entropy is a bitch, isn't it? It's even true of current information storage devices on a short lifespan. How many people have lost data from a HD crash? This technology will likely be used as a third storage teir, mostly for fast startup times, etc. Put in 5 gigs of the stuff and let it act like a disk cache + hibernate cache. The computer can shut off completely when not in use, and come back almost instantly.

The point is that consumers really won't to choose between mram and pram. Just like they haven't chosen between SDRAM, DDR, and DDR2. That's a minor technical point as it's all abstracted away. It'd be like worrying about the "format war" between fuel injection and carborators. A few people might care about that kind of thing, but for the vast majority it's irrelevant.

Yeah...

[GotenXiao]

...but does it run Linux?

When can we get portable battery-backed DDR?

Cheaper?

[Roadmaster]

Cheaper to produce, my ass. They'll be charging an arm and a leg for this type of memory.

Re:Cheaper?

[timeOday]

Not if they really want to displace flash they won't.

Pram?

[ArchAbaddon]

Isn't that something you push babies in?

'Perfect RAM'

Wow. I'll believe it when I see its "perfection" :p

Camelot!

[autophile]

I.... have... to-o-o-o.. push.... the.... PRAM a lo-o-o-o-o-t!

(quick montage of knights dancing and kicking chickens)

No, on second thought, let's not go to Camelot. It is a silly place.

Re:Camelot!

[cranesan]

Thanks! Someone had to post this...

How storage should be stated

[splatacaster]

Storage should be stated in megaBYTES and gigaBYTES not bits. No one uses bits anymore, we've all upgraded to bytes. Then there was that half standard of nibbles, what the heck was that all about.

Re:How storage should be stated

[vidarh]

When it comes to a chip manufacturer stating memory storage capacity, then megabits is the standard. Deal. The reason it's stated that way is because it indicates the number of addressable units.

2 years, vaporware red flag alert!

[mark-t]

FTA:

...availability beginning sometime in 2008...

You decide.

Re:2 years, vaporware red flag alert!

[cyfer2000]

The technology behind the PRAM has been developed 20 or 30 years. 2 more years is not too much.

I'm assuming this is different from PRAM?

[StarkRG]

As in Parameter RAM? As in "a small amount of NVRAM used on early Apple Macintosh machines to store configuration information"?

If we store porn on these

[syntap]

will we have to call them "pr0m" disks?

Re:If we store porn on these

[Skevin]

> will we have to call them "pr0m" disks?

Only if it's read-0nly memory. In my day, we used to have plenty of Read-Only porn... they were called magazines.

Solomon

Faster than Flash? Bah!

[ROMRIX]

They've never seen me on a crowded pier in a trench coat!

.

.

"PRAM?"

[Aqua OS X]

I speak for all Mac geeks when I say, ugh, "PRAM."

http://docs.info.apple.com/article.html?artnum=861 94

Couldn't they pick an acronym that's not already being used?

Where's the catch?

[edxwelch]

If all is true in this story we can wave goodbye to Spansion.
But the artical is a bit inconsistant on some details...
for instance they say it's 30 times faster than NOR (but they don't say if this is read or write speed)
and then they say it's more reliable than flash (but they don't say if it's more reliable than NOR or NAND)
Then they say "Samsung claimed the PRAM would become a competitive choice over NOR flash" ...what? 30-times faster, cheaper to make and more reliable, but it's only "more competitive"?

This leads me to think that there must be some strings attached that they are not telling us about

Monty Python

[pravuil]

Oddly enough, I'm reminded of the term "pramalot".

Damned straight....

[Overzeetop]

Why even talk about binary storage methods in bits? All of us endusers only know bytes, and it's a significant imposition to make us multiply or divide by eight to get anything useful. It's not like you're storing your information as individual bits of...oh...right. :rolleyes:

Re:Damned straight....

[splatacaster]

It's a joke, geniuses. And a byte is 8 bits not 10.

Re:Damned straight....

[oracleofbargth]

Only untill someone is crazy enough to design a 10bit cpu core. ....

I think I'll go and check opencores.org to make sure someone else hasn't done it already.

Re:Damned straight....

[codican]

Reporting memory size in bits is less prone to error because contrary to popular belief, a byte is not necessarily 8 bits.

-C

So there's a baby in there, huh?

[dangitman]

Interesting.

TLAs and FLEAs

[camperdave]

There are only so many three letter acronyms to go around. Same goes for fourth letter enhanced acronyms. We have to double up on them somewhere.

Re:TLAs and FLEAs

[Aqua OS X]

It could've been called "awesome RAM"

Wooooosh!

[Overzeetop]

the sound of...

Sure PRAM maybe faster than FLASH...

[__aaasvk1266]

But is it faster than Superman?

glad that finally paid off for you

[Draconnery]

You've just been waiting, haven't you?

Re:glad that finally paid off for you

[IpalindromeI]

I think this one is better.

Flash does not have read limits

[EmbeddedJanitor]

Flash does not have read cycle limitations. It has erasure/write limitations. However, in most typical usage scenarios this is a red herring or can be designed around quite simply with various wear levelling strategies.

MRAM, OTOH, does have read limitations too making it poor for execute in place applications. MRAM is also expensive (per byte) and not very dense, making it useless for most typical flash applicaitons.

At this stage NAND is the prefered flash for large arrays and file system applications because it is cheap, dense and very fast, but it cannot do execute in place. NOR is prefered for smaller applications, or where execute in place is desirable (boot code, and some other applications).

Will PRAM make a big difference? We'll see.

and they probably...

[zogger]

...still work, too. I know all my older HDDs are still quite functional.

PRAM vs. MRAM

[TerranFury]

Good question. Not sure of all the answers, but I'll help as much as I can.

First off, PRAM = Phase change RAM.

PRAM is a writable DVD on a chip. It works by heating and cooling material at different rates. By annealing the stuff appropriately, you can cause it to enter either an amorphous or a crystaline state. These two states differ in two ways: Reflectivity and conductivity. You can read and write either with an optical or an electrical system. On a DVD-R, you write a bit by heating with a laser; you read a bit by bouncing a laser off; if it's crystaline stuff you're hitting, it bounces off like a mirror; if it's amorphous, you get less back. On a PRAM chip, this stuff is lined up in an array of cells; you write by heating a cell with a pulse of current shaped to give the correct cooling profile, and you read, in essence, with an ohm-meter.

All that heating sounds like it sucks a lot of power. But then, MRAM requires pretty substantial currents to generate the magnetic fields required to flip the magnets as well, so power consumption on both PRAM and MRAM are, as far as I know, actually about the same.

PRAM does have the advantage of multi-level recording. In my earlier description, I implied that a cell is either entirely in the amorphous state or the crystaline state. In fact, some percentage of the cell is in one state and some percentage is in the other. It is possible, then, to with four levels store two bits in a single cell. I overheard people talking about a 16-level four-bit cell the other day. You can't do that with MRAM.

Nevertheless, they're both contenders, and as far as I know neither is clearly superior. Some companies are researching both.

Also, neither will ever be the "Perfect RAM." Neither will oust DRAM for density or SRAM for speed.

(Whosoever knows more than the above is invited to post a reply.)

Here's a useful measurement.

[Kamineko]

This new type of memory is at least as fast, if not faster than Sigourney Weaver, but only when SLIME is installed.

More information on phase-change memory

[pm]

Although someone else posted it - but have a score of 0 presently and are probably filtered by most people - the wikipedia entry for this type of memory is: http://en.wikipedia.org/wiki/Phase-change_memory

I did a literature search on this type of memory, and found a paper by Samsung on a 64-Mb PRAM published earlier this year.
"Enhanced write performance of a 64-mb phase-change random access memory"
Hyung-rok Oh; Beak-hyung Cho; Woo Yeong Cho; Sangbeom Kang; Byung-gil Choi; Hye-jin Kim; Ki-sung Kim; Du-eung Kim; Choong-keun Kwak; Hyun-geun Byun; Gi-tae Jeong; Hong-sik Jeong; Kinam Kim; Solid-State Circuits, IEEE Journal of, Volume 41, Issue 1, Jan. 2006 Page(s):122 - 126

My summary:
The 64-Mb PRAM was produced on a 0.12um CMOS technology, and had a voltage of 1.8V. The "0" resistance is 1.2kohms, "1" resistance is 100kohms (a nice delta which could mean that you could use the cell to store intermediate states ie. 4-states per cell, instead of just 2 (binary) ). Random access time is 68ns, and write time is 180ns, at 1.8V at room temperature.

This abstract at the Japanese Journal of Applied Physics seems to be on a similar 64-Mb chip from Samsung (although some details are different):
"Ge2Sb2Te5 Confined Structures and Integration of 64 Mb Phase-Change Random Access Memory", Fai Yeung*, Su-Jin Ahn, Young-Nam Hwang, Chang-Wook Jeong, Yoon-Jong Song, Su-Youn Lee, Se-Ho Lee, Kyung-Chang Ryoo, Jae-Hyun Park, Jae-Min Shin, Won-Cheol Jeong, Young-Tae Kim1, Gwan-Hyeob Koh, Gi-Tae Jeong, Hong-Sik Jeong and Kinam Kim,

http://jjap.ipap.jp/link?JJAP/44/2691/#search=%22p hase-change%20random%20access%20memory%20reliabili ty%22

Cutting and pasting the abstract:
Phase-change random access memory is considered a potential challenger for conventional memories, such as dynamic random access memory and flash memory due to its numerous advantages. Nevertheless, high reset current is the ultimate problem in developing high-density phase-change random access memory (PRAM). We focus on the adoption of Ge2Sb2Te5 confined structures to achieve lower reset currents. By changing from a normal to a GST confined structure, the reset current drops to as low as 0.8 mA. Eventually, our integrated 64 Mb PRAM based on 0.18 m CMOS technology offers a large sensing margin: Rreset 200 k and Rset 2 k, as well as reasonable reliability: an endurance of 1.0×109 cycles and a retention time of 2 years at 85C.

1E9 cycles is pretty impressive.