Windows Has a Future In RAM: AgigaTech Samples DDR3+Flash DIMM 139
An anonymous reader writes "AgigaTech appears to be the first company to produce a non-volatile SDRAM DIMM — an SDRAM memory module that retains its contents even without power supply. The modules combine DDR2/3 SDRAM with NAND Flash as well as a data transfer controller and an ultracapacitor-based power source to support a data transfer from the SDRAM to Flash and vice versa. If this memory makes it into production, this is something that I instantly will want and will stand in line for."
Eh? (Score:5, Insightful)
What's windows got to do with it?
No, wait.... (Score:3, Informative)
"The most obvious application is the vision of keeping, for example, Windows completely stored in a DIMM." - is that it? Is that one sentence the reason for the headline?
Re: (Score:1)
Microsoft announced the launch of Windows 8 and Windows Phone 8 two days ago and they need the column-inches.
Expect another six weeks of Slashdot product placement from Burson Marsteller and Microsoft.
Re: (Score:1)
And this product placement is supposed to positively influence /. readers to pay money for Win8?
Re: (Score:2)
You'd only want parts of windows I'd think since a large part of the OS is helper apps and dialogs and such that most people don't ever use (but people use a different subset so they still are in the box because a large group really really wants it). Better core OS + frequent apps IMHO.
Re: (Score:2)
Comment removed (Score:5, Interesting)
Re: (Score:2)
Hybrid sleep? battery? Having trouble finding this on my desktop computer.
Do you mean the CMOS batery? if so, wrong battery.
Re: (Score:2)
I've been using S3 standby on a desktop with a UPS for the last 9 years or so with windows XP.
Fantastic, basically i never reboot XP, walk away it goes to sleep after 15 minutes, and wakes on a key press in under 1 second.
Power draw in S3 is ~6W including monitors after I purchased a 80+ power supply I found on some website talking abut standby power usage (most of them are only 80+ efficient in the intended power range which is generally > 50 watts).
Re: (Score:2)
6 W? That's a lot, just to keep a few DIMMs and some helper electronics going...
Yes, but I think most of it is the power supplies being 80% inefficient at that low of current draw. If someone were to hard power the monitors off when the PC goes to sleep it would probably save a watt or two there too. Its actually more than my (ARM based) NAS running full out, but I'm pretty happy though.
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
It will benefit Windows and OSX a lot more since suspending there actually works.
If you ever successfully unsuspend a *BSD or Linux box, you should post a bragging video on youtube and never, ever update anything.
Zero watt 'suspend' instead of 'hibernate' (Score:2, Informative)
Instant on and zero-watt suspend instead of having to hibernate. This would be faster even than booting from SSD. The summary is implying that windows (or other OS) would reside installed on RAM instead of to the hard-drive, so there would be no load time.
Re: (Score:3, Insightful)
The summary is implying that windows (or other OS)
Then why not say something non-ambiguous like "keep entire operating system in memory"? This is Slashdot not NewbDot. They don't have to "imply" anything; just say it.
Re: (Score:1)
Well, we've been able to do it for years with other OSes. I guess they are finally able to make a DIMM big enough to hold Windows in it.
Re: (Score:1)
Err no, there would be a load time because it's exactly "hibernate to SSD". It's just that they control the RAM-to-SSD bandwidth so they might in theory make the load time faster than a software solution, but they give no number.
Re: (Score:2, Insightful)
Err no, there would be a load time because it's exactly "hibernate to SSD". It's just that they control the RAM-to-SSD bandwidth so they might in theory make the load time faster than a software solution, but they give no number.
You are assuming that there will be separate RAM and NAND chips on this module. If the flash memory is interleaved with the RAM cells in the IC's then the data will never have to hit the bus and never have to be serialized. If this is the case the NAND-cells only have to be connect to the RAM-cells in a way that they are "nudging" the cells in the correct direction during power on and the memory will be fully loaded before the motherboard even releases the reset-signal to the CPU.
Re: (Score:2)
This will not be the case, semiconductor processes can be optimised for one or the other.
You get a more dense RAM or flash by not interleaving.
And as area is proportional to cost, it would end up more expensive.
Re: (Score:3)
You obviously haven't used 8 on an SSD system. It's like using a faster, better iPad.
Re: (Score:1)
Re: (Score:2)
Beat me to it. Yeah don't get what windows has to do with this. If anything its very aggressive memory management (Ie cache everything just in case) is a drawback because you'd need more of this stuff to keep your whole resident memory footprint.
Re: (Score:3)
Re: (Score:2)
Re: (Score:2)
They could have the same density. This is one tech that sounds like it could make use of some of that empty space that generally lies immediately above the ram sticks. For blade servers this doesn't apply of course but anyone deploying blade servers isn't going to want this stuff until its a much more mature tech anyways.
Re:Eh? (Score:5, Informative)
There was a false idea floating around back in the days of the SIMMS-> DIMMS transfer that SDRAM stood for Static Dynamic Ram and that it switched to DRAM as the abbreviation for DIMMS. This was actually propagated in marketing for awhile.
I've also seen people claim that SIMM stood for Static Inline Memory Module, what was actually the case was that SIMMS were most often a form of PRAM which is just battery backed DRAM, which added to the confusion.
Static Dynamic RAM was actually sort-of a term for awhile but Synchronous Dynamic Ram needed the abbreviation far more as it extremely quickly became the standard.
Therefore you have a situation where often SRAM and SDRAM can mean the same thing to some people.
My cousin who is a Computer Science grad who got his degree around that time actually learned from a Prof who also had incorrect information and argued with me for a long time on the matter until I managed to care enough to dig out the relevant technical documents and show him why he was wrong.
Re: (Score:1)
Re: (Score:3)
Who needs Windows when Windows can be broken?
Not broken, but feature enhanced to improve air flow.
Re: (Score:2)
And jagged edges for security protection!
Old News (Score:3, Informative)
Viking has been doing this for awhile. This is their second incarnation.
http://www.vikingtechnology.com/arxcis-nv
Uh yeah, right (Score:1)
If this really does work it'll be tremendously useful - even if 20 years later than predicted (Yes, flash-backed ram modules were being touted as the next big thing that long ago)
UPS (Score:3)
Re: (Score:2)
DC's don't jump on this sort of stuff till i'ts much more common.
Server soften require A LOT of RAM, and no consumer is going to pay two extra zeros for a server with this non-volatile RAM when they could get one from a competitor with volatile RAM and SSD.
Re: (Score:2)
Re: (Score:2)
oh so when your computer crashes you have no boot menu and the os boots immediately to a crashed state sounds like fun
Re: (Score:1)
How about holding a button when turning on the computer?
Less interesting than the writer thinks. (Score:5, Insightful)
This will require essentially the same software infrastructure as normal suspend to RAM.
The system still has to go through the steps:
Check to see if any critical tasks are running - if so, pause suspend, and ask user.
Same with any communications tasks that may be interrupted.
Stop tasks.
Save state from all hardware to RAM.
Suspend to RAM.
Just capturing an image of the running system does not result in a system that will resume.
It's not a case of put one of these magical DIMMs in, and you're fine for power cuts.
Is it possibly interesting - sure.
But in real life, it may have very little advantage over a seperate flash device, for main memory.
Now, as a super-fast SSD - truly awesome.
Also - WTF - this should never be patentable.
This is not an invention worthy of patent.
It does nothing novel that is not implicit in the problem statement.
'I want a non-volatile RAM'.
Re: (Score:2)
This will require essentially the same software infrastructure as normal suspend to RAM.
The system still has to go through the steps:
Check to see if any critical tasks are running - if so, pause suspend, and ask user.
Same with any communications tasks that may be interrupted.
Stop tasks.
Save state from all hardware to RAM.
Suspend to RAM.
Just capturing an image of the running system does not result in a system that will resume.
It's not a case of put one of these magical DIMMs in, and you're fine for power cuts.
Is it possibly interesting - sure.
But in real life, it may have very little advantage over a seperate flash device, for main memory.
Now, as a super-fast SSD - truly awesome.
Also - WTF - this should never be patentable.
This is not an invention worthy of patent.
It does nothing novel that is not implicit in the problem statement.
'I want a non-volatile RAM'.
RAM Disk.
Re: (Score:2)
No, you aren't. The CPU state, Video Card, Sound Card, etc states will all be wiped out. The only use I can see is to salvage the word document you spent the last two hours not saving
Re: (Score:2, Insightful)
Having to save was fine back in the days when it took 20 seconds to save to disk. These days, every change should be able to be saved and logged to a change log on a second by second basis. Saving should be banished and everything should be able to be rolled forward and back. People keep saying computers are more than fast enough, so put that extra power to actual useful things.
Re:Less interesting than the writer thinks. (Score:4, Insightful)
Saving to disk is an explicit action of "This is at a state I want it to be in". If it is persistent, for instance, and my kids/cat/whatever edit it beyond repair, I don't want that existing instead of my work. You could argue about rolling back, based on your logging suggestion, but you just made a simple paradigm into an over-engineered tedium. Also, think about having to play back that log every time you opened it, multitudes of keystrokes and menu commands could be needed before it is ready.
Re: (Score:2)
Or you could do it like OS X does it: Auto-save, and implicit saves make snapshots that you can then roll back to. No having to wade through tons of keystrokes, no ridiculous overhead.
It's not a perfect system (for instance, snapshots aren't persistent across systems if you copy/move the file; they're stored on the local machine. I actually tried to get it to sync over Dropbox via symlinking, but ran into permissions problems. Maybe someone else will have/has had more luck.), but it works pretty well. I
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
Re: (Score:1)
I sort of agree with you. The user wants a touchstone of "I saved this in a known good state at this time". It's nice to be able to capture all the intermediate changes as a log that can be rolled back, if required, but this should be the exception, not the rule. This level of detail should be hidden at or below the filesystem level, and only used when disaster happens.
Re: (Score:2)
So now I have to go through a 600 line changelog to find the version from 10 minutes ago that I was pleased with.
There would need
Re: (Score:2)
On a Linux system maybe. Other systems will build a decent UI around it so the end user can do revolutionary and unpredictable things like specify "I want to see what this document looked like ten minutes ago".
Indeed. Fortunately the idea of a "rel
Re: (Score:2)
Having to save was fine back in the days when it took 20 seconds to save to disk. These days, every change should be able to be saved and logged to a change log on a second by second basis. Saving should be banished and everything should be able to be rolled forward and back. People keep saying computers are more than fast enough, so put that extra power to actual useful things.
Unless you're working with big files. The DRAM on my computer can easily handle 2GB files. At SATA3 data transmission rates, it would take more than 3 seconds to transmit over the bus, but the devices in typical computers can't keep up with SATA3 on a sustained basis. If you have 8GB RAM, you need 8GB to cache it and it takes more than 12 seconds on a typical system. Not to mention that the caching operation can't take place without interrupting whatever process you're doing because it requires the bus.
Re: (Score:2)
Since most "critical" tasks such as writing to a disk or communicating over the network already require handling of drop outs (SATA is hot swappable, most communication is USB based which can cut out) it should be as simple as retrying those tasks when the power comes back on. In theory this should be able to handle power drop outs. You might have to alter the OS to have effectively a journal with respect to the CPU/other hardware, but that's not a terr
Re: (Score:2)
The patents likely refer to solving various problems in integrating an ultracap with a DIMM.
That's probably down to refreshing the memory. Dynamic RAM is much more power-hungry than static RAM. Static RAM is just a bunch of flip-flops that latch either into 1 or 0 - make them with tiny CMOS gates and they can hold their state as long as power is applied and draw hardly any current, but you need a lot of gates so the die is large, density is low and the part is expensive.
Dynamic RAM contains thousands - or these days, more like millions - of tiny capacitors that need to be kept charged. Over tim
Re: (Score:3)
TFA tells us the technology will not be targeted towards PC's, but for RAID controllers.
I guess it could help against data loss in critical systems, combined with disk caching it can offer nice responsive write times.
Re: (Score:2)
Meh. That's been done with conventional RAM, flash and ultracapacitors for years, for example in Fujitsu's DX80. Even PCI RAID controllers do that now, for example LSI's CacheVault option for the newer MegaRAIDs.
Re: (Score:2)
It's an awesome SSD speedup device.
While the basic concept may not be patentable, within the present patent system I'm sure there's plenty of details of implementation that can be patented, since nobody seems to be able to make the "obvious to one skilled in the art" clause fly with a jury.
Re: (Score:2)
Also - WTF - this should never be patentable. This is not an invention worthy of patent. It does nothing novel that is not implicit in the problem statement. 'I want a non-volatile RAM'.
So what? The PROBLEM is obvious. You shouldn't be able to patent the idea of a FLASH-backed DRAM module, but if you do something clever to make it work, that cleverness can be patented, so anybody else who makes a flash-backed DRAM module will have to do it another way.
Great idea, (Score:1)
this will do wonders for security. Oh wait.
Re: (Score:1)
I don't see it (Score:4, Interesting)
Okay, I've broken the Slashdot rule and read the article.
Can anyone tell me why this is so much better than traditional RAM with a SATA attached SSD? Or using hibernate to disk with an SSD? Is SATA so slow and laggy that there's a big benefit to attaching flash chips to our RAM slots?
Retaining data in RAM without power is cool as a technical feat, but my SSD doesn't take long to fill my RAM chips.
Re:I don't see it (Score:5, Funny)
Okay, I've broken the Slashdot rule and read the article.
Can anyone tell me why ...
No, because we didn't break the damned rule! Now, do you see why we have it?
Re: (Score:2)
One of the advantages laptops have over desktop computers is effectively a built-in, relatively lightweight UPS. When someone kicks out the cord on my laptop, I don't even notice, but on my desktop, that would be very annoying. If some upgraded RAM/Flash + operating system support would allow hibernation on a desktop when the power was cut, that would be very handy. Some thought would have to be given to how this should interact with filesystems, since the hard drive would instantly lose power as well.
Re: (Score:2)
If some upgraded RAM/Flash + operating system support would allow hibernation on a desktop when the power was cut, that would be very handy.
You don't need upgraded RAM.
Just a capacitor that can power teh sytem for 10 seconds to allow a graceful suspend. Cost trivial.
Spend a few dollars and you could get something to last a few minutes, have it set off a beeping countdown which you could interrupt by plugging it back in.
But I guess that the people who know they need this aren't enough to justify the feature, and so they are forced to spend a hundred times as much to get a full on UPS.
I've got dodgy home electric wiring, damp rises up and shorts
Re: (Score:2)
so they are forced to spend a hundred times as much to get a full on UPS.
You mean tens of dollars. Which I will choose buying a UPS over fiddling around with capacitors. Not worth my time.
Re: (Score:2)
FWIW, my desktop doesn't wake up from suspend properly - the networking stuff goes weird. My work laptop doesn't have such problems. However my laptop takes quite a long time to hibernate and restore from hibernate (I tend to run a lot of things... ).
Another thing to think about regarding hibernation vs suspend: if your HDD transfers at 100MB/sec and you've used up 16GB of RAM, it'll take at le
Re: (Score:2)
I don't think windows saves/restores the entire RAM range on hibernate/resume. AFAIK, only the kernel/non pagable and dirty pages are flushed, the rest is simply marked paged out. Then when you resume it reverse the whole thing and starts to fault in pages.
So the average user probably isn't initially moving more than a hundred or two megs of stuff in a multi gigabyte system before the system starts to respond. Especially if most of the ram was disk cache (or other garbage that won't be saved/restored).
That
Re: (Score:2)
Restarting stuff takes more human interaction so I stick to restoring from hibernation and do something else.
Re: (Score:2)
Re: (Score:1)
This is interesting as memory for the SSD itself, so that writes only go to the flash when the device powers down. This could reduce the number of writes to the NAND chips considerably if the SSD is used heavily as a cache or swap device.
Re: (Score:2)
I think it's cool from a systems perspective in that I can make a system without any support for SATA interfaces or drives, smaller, cheaper, maybe faster maybe not, but a driveless system takes less silicon and less power.
Re: (Score:1)
New Memory Technologies - The Impact (Score:5, Interesting)
Since computers began we have had hierarchal memory systems. Cache is the most expensive, but the fastest. DRAM is much cheaper slower and denser, but also volatile. Flash is faster than rotating media, slower than DRAM, but non-vloatile. It also has the drawback of limited programming cycles. Magnetic media is very dense, non-vloatile and slow. It is also mechanicly delicate. There are new technologies being developed that are both fast, dense, and non-volatile. With a fast enough, cheap and non-volatile memroy system, you would not need cache, RAM or disk. You could use on unified memory system. This is where I think many syustems are going. Windows, Linux, or OSX have nothing to do with it. Though they will all be greatly impacted.
Re: (Score:1)
With a fast enough, cheap and non-volatile memroy system, you would not need cache, RAM or disk. .
Well, that's the rub isn't it? We have always had hierarchical memory systems for exactly the reasons you state and there is certainly nothing on the horizon to believe that that is going to change. Until someone comes up with something that is faster, cheaper, denser and non-volatile there will continue to be trade-offs between these. We still use tape for really dense and really slow applications, and despite the many predictions of the HDD, those don't look to be going away any time soon.
Re: (Score:3, Interesting)
At 4 GZ light can only propagate 7.49 cm in one clock cycle. Communication in a computer is slower than the speed of light so the cache has to be physically very close to the CPU and preferably on the same die. That means that a harddisk at the end of a cable is completely out of the question to replace cache even if the harddisk could satisfy IO requests in zero seconds - the time to get the signal to the harddisk and back through the cable would be a limiting factor. Now you could imagine the CPU and hard
Re: (Score:2)
Actually, you bring up an excellent point. I actually meant to refer to external memory systems. The speed of light is an absolute barrier which obviates the physical placement of a very high performance system. I can't imagine any replacement for an on-chip cache. But memristors, phase change memory and magnetoresistive external memory systems have made me hopeful. How exactly files would be stored in such a system and booting I think would create some interesting new retrieval and storage solutions.
Re: (Score:2)
Since computers began we have had hierarchal memory systems.
And always will. It's physically impossible for distant RAM to be as fast as CPU registers (the lowest addressable layer of the hierarchy). I don't imagine a future where fast random-access modules will have the capacity of slower stores that have the luxury of time to do their work. For instance, add "cloud storage" or "network fileserver" as yet another common layer of that hierarchy. It's many orders of magnitude slower than my laptop's memory, but my laptop doesn't have 100TB of RAM.
No, that hierarchy i
Re: (Score:2)
There are new technologies being developed that are both fast, dense, and non-volatile. With a fast enough, cheap and non-volatile memroy system, you would not need cache, RAM or disk.
No matter how dense memory technology you use you can put a 32K L1 cache much closer to where it's needed than a 16GB main memory. One is simply hundred thousands of times bigger. That means lower latency and higher performance, so cache is here to say. Here's is for example the latency figures for L1/L2/L3/memory for Ivy Bridge:
L1: 1 ns
L2: 3.0 ns
L3: 3.8 ns
Memory: 39.2 ns
Sure you could do it without the cache... if you want to wait 40 times as long. And it's not just to say that you will make a 1ns memory t
Re: (Score:2)
Uhhh, no. The cache hierarchy was added over time. The first few generations of computers did not have caches at all. Even the processors that powered a lot of early PC-era computers did not have caches, unless you count the registers. For example, 8086/8088 did not have cache, 6502 did not have cache, 6800 did not have cache, 68000 did not have cache, etc. Cache hierarchies were added later.
The cache hierarchy also continually changes, albeit at a slow pace. Current generation CPUs typically have a
you really want code to get stuck in ram and leak (Score:2)
you really want code to get stuck in ram and leak???
I like the reboot to start clean
this is good! (Score:4, Funny)
now a reboot of windows won't solve anything!
Re: (Score:2)
WTF (Score:2)
Freescale MRAM (Score:2)
The MRAM concept would be awesome if they ever got it onto a PC or server motherboard, though. It requires zero power to retain its data, since the
Nothing new under the sun ... (Score:5, Interesting)
I remember some 40 years ago using a PDP-7. When I got tired at about 4am I would note the accumulator and program counter and switch the machine off. Coming back later I restored these and continued the program - it having remained in the core memory that the machine had.
Re: (Score:2)
And if you trashed the boot loader you had to toggle it back in by hand :)
Good times.
Re: (Score:2)
The times when you were part of the core memory.
Are we forgetting something here? (Score:1)
Okay you morons, wake up! Did we already forget the lesson about encryption keys being recoverable in RAM by rebooting into a minimal kernel and cooling the chips? And now you want ram that is capable of permanently storing those keys?!
Come here so I can beat all of you senseless with a large cluebat...
The death of software support (Score:2)
What this means is that the support dept's favourite saying "switch it off and on again" won't work. If this takes off, they'll have to actually start diagnosing and fixing software faults.
On the upside, maybe then we'll get better quality software .... no, I didn't think so either.
Two use-cases (Score:2)
Two obvious use-cases: A mail spool directory and a database transaction log directory. Both can benefit enormously from RAM speeds yet must survive power failures. As a bonus, they don't usually use that much storage, so a few GB or tens of GB is probably fine.
Does Slashdot (Score:1)
Does Slashdot charge for obvious advertisements like this one?
Useless stunt (Score:2)
At resume time the operating system needs to reinitialize the hardware and check anyway if the memory contents is still valid. Not that much for an embedded system with components welded to the board - but then you won't have memory as modules.
This has existed for years (Score:2)
Even before SSD drives came into the mainstream (I had a 1MB SSD in my 80286 which had 8ms latency) this technique already existed:
- 5.25" RAM drive with Flash (CF) backup from Acard as well as the DDRDrive which is a plugin card
- HyperDrive RAM-based SSD which was a PATA solution
- Rack mount systems (NAS, DAS) which were basically RAM and batteries with a few either SSD's or even HDD's as backup solution. TI's RAMSAN for example.
- SAN based "accelerators" which sat on FibreChannel
- 2.5" and 3.5" SAS and FC
Re: (Score:2)
Yup, After a short web search I have a solid sightings for you from http://www.storagesearch.com/ [storagesearch.com]
1997 - in the SSD market
Bridgeworks designed a RAM SSD with hard drive backup. Sales Director - David Trossell told me - "It was a little ahead of its time and the company dropped it after poor sales."
And later, once Flash was big enough your nice little ramsan ...
Houston, Texas - July 22, 2008 - Texas Memory Systems today launched the world's fastest SSD - the RamSan-440,
The RamSan-440 is a 4U rac
Revolution, yeah (Score:2)
Re: (Score:2)
Yes! Finally the encryption keys for my disk will be saved when I turn off my laptop!
Finally I won't have to enter the passphrase every time I turn in on!
Oh, wait...
Encrypt Thy Ram (Score:3)
Horrible Idea (Score:1)
Too Small!!! (Score:2)
I read TFA. Saw a picture of the DIMM and it said in big bold letters, AGIGARAM...
A gig of ram? Is that all?
Too small, do not want. /s
this already exists (Score:1)
Re: (Score:3)
Clint
ahem...