Everspin Launches Non-Volatile MRAM That's 500 Times Faster Than NAND 119
MrSeb writes "Alternative memory standards have been kicking around for decades as researchers have struggled to find the hypothetical holy grail — a non-volatile, low-latency, low-cost product that could scale from hard drives to conventional RAM. NAND flash has become the high-speed, non-volatile darling of the storage industry, but if you follow the evolution of the standard, you'll know that NAND is far from perfect. The total number of read/write cycles and data duration if the drive isn't kept powered are both significant problems as process shrinks continue scaling downward. Thus far, this holy grail remains elusive, but a practical MRAM (Magnetoresistive Random Access Memory) solution took a step towards fruition this week. Everspin has announced that it's shipping the first 64Mb ST-MRAM in a DDR3-compatible module. These modules transfer data at DDR3-1600 clock rates, but access latencies are much lower than flash RAM, promising an overall 500x performance increase over conventional NAND."
Re: (Score:2)
This sounds suspiciously like core memory.
Soon the term "core dump" may no longer be an anachronism.
The wheel turns...
Re: (Score:1)
Q1: Do the DIMMs in your PC run linux?
Q2: Are you DIMM?
Re:So NOT Vaporware? (Score:4, Interesting)
Still completely impractical. It may improve with time, but I wouldn't hold my breath. They basically have to improve fast enough to catch up and then surpass Flash memory, which is difficult at best with the enormous lead Flash memory currently has.
Re: (Score:2)
"still years away from practical consumer applications"
I read this as "still years away from BS Rambus patents from the mid 90's to expire"
Re:So NOT Vaporware? (Score:4, Interesting)
Re:So NOT Vaporware? (Score:4, Interesting)
Yes, but many existing silicon technologies are running up to lots of hurdles right now at current feature sizes, so the single atom problem isn't close to being a concern. Some of the newer technologies not only allow much smaller feature sizes than the current 20nm, but will also allow stacking of components, rather than having a single layer of components as we do now in chips. Not only that, but some are non-volatile, yet fast enough to replace DRAM, so they would have a greater market being able to provide both ram and storage solutions. Hopefully that greater market combined with increased densities will lend itself to greater production and economy of scale. Who really knows though, there may be limits to how much they can produce, and thus the greater market will instead cause them to be priced at a premium. I'm just gonna wait and see, and be hopeful.
Re:So NOT Vaporware? (Score:5, Interesting)
Some of the newer technologies not only allow much smaller feature sizes than the current 20nm, but will also allow stacking of components
Sure we can stack them, but can we cool them? Even the Ivy Bridge chips that lowered power consumption a lot compared to Sandy Bridge increased the watt per mm^2 die size due to the die shrink and now it's up to 77/160 = 0,48 W/mm^2. That is a lot of power you have to dissapate to keep a sane operating temperature. Having a flat chip - ignoring the 3D transistors, which are practically flat for this purpose - connected to a huge heat sink is a pretty effective way of doing that. If you stack the chips many of them won't be on the outside. First they have to transfer all the heat through the other layers, then to the heat sink. Supercomputers have worked on it for decades and they haven't really found a working solution.
Re: (Score:2)
Some of the newer technologies not only allow much smaller feature sizes than the current 20nm, but will also allow stacking of components
Sure we can stack them, but can we cool them?
Is there anything carbon nanotubes [llnl.gov] can't do? Besides, of course, have an acronym that you don't mistake for a naughty word.
Re: (Score:2)
I'm speaking off the top of my head, but I think the memristors supposedly used significantly less power(and hence generate less heat). But they were indeed much much denser, so as you point out, the heat/area might not have been lower. I had never considered that before.
Re: (Score:1)
PRAM? (Score:3)
Re: (Score:1)
HP has discontinued this research in favor of MRAM.
Re: (Score:3)
What part of hitting the process-reduction-scale wall haven't you understood? Right now there's not much boundless confidence that the scale of NAND can be reduced much more without rendering it useless. The already onerous guaranteed obsolescence of the medium is getting worse, not better, with every die shrink. Without some miraculous discovery NAND is rumored to be close to bottoming out, so it seems to me that an upstart technology can make some big leaps in the time it will take NAND designs to make
Re: (Score:2)
Re:So NOT Vaporware? (Score:5, Informative)
They claim shipping, so... yeah, a product. However, not a retail product, from the sound of it. Nobody makes a populate-your-own SSD or such.
More importantly perhaps, MRAM supposedly doesn't suffer from the page problem that NAND requires. Individual bits are accessible for reading and writing conveniently, unlike NAND, which requires writing by page. In addition, MRAM is supposedly much more robust than NAND, surviving many more write cycles. It hasn't existed long enough to know this for sure, but in theory, these two advantages means an SSD controller for an MRAM SSD could be vastly simpler than the ones required for NAND. No need for wear-leveling or page rewrite logic. This should both reduce the expense of SSDs and increase their real world performance and reliability.
However, while the article summary blathers about "from hard drives to main memory", this is not a competitor to modern DDR SDRAM. Assuming the quoted 500X faster than NAND is accurate, MRAM latency should be on the order of 100 nanoseconds for a random read. (NAND read latency is on the order of 50 microseconds.) DDR SDRAM random read latency is on the order of 22 nanoseconds.
Having said that, it is comparable with SDRAM from a decade ago, which probably translates directly to modern mobile devices. Low power suspend mode using MRAM instead of SDRAM could conceivably lower mobile device power consumption and improve battery life. If manufacturers get really silly, in theory a mobile device could be built that doesn't distinguish between its main memory and its mass storage. The two functions would be served by the same solid state circuitry. Obviously accommodating such a hardware design would give the kernel guys fits, but it could simplify things in the software a great deal, and incidentally net an interesting performance gain that's visible to users. Notably, the process of launching a program consists of nothing more than creating a stack and a heap for it somewhere--the program's code can stay right where it is. This also results in the somewhat bizarre (to modern ears) situation where suspend mode consists solely of persisting the CPU's state. Memory state is already persistent, always. As a final side effect, once scaled to SSD capacities, a device operating as described above effectively has an absolutely absurd amount of main memory, in theory, equivalent to the entire remaining capacity of the mass storage device.
MRAM has been around in labs for 20 years now, so the possibility of this being a real, viable, product-ready device is reasonably high. MRAM doesn't suffer from Fusion Power Syndrome.
Re: (Score:2)
From Outer Space?
Re: (Score:1)
Obviously accommodating such a hardware design would give the kernel guys fits
Already been there and done that. There's an entire "Memory Technology" section in menuconfig to cover all of this, plus the exec-in-place (XIP) stuff for executing programs from memory-addressable storage.
Re: (Score:3)
NOR Flash is more resilient than ST-MRAM in two ways; it requires more than just a write enable line to write or overwrite data and NOR Flash is insensitive to magnetic fields.
Re: (Score:2)
Re: (Score:2)
They claim shipping, so... yeah, a product. However, not a retail product, from the sound of it. Nobody makes a populate-your-own SSD or such.
But they do. You can buy a populate-your-own SSD that takes DIMMs. They have a battery backup so that they can refresh themselves. You can use them for cache volumes in contexts where you need a "real" disk. I don't know if you can get anything that takes DDR3 though, and you'd probably need to diddle the firmware to get the right timings for their memories.
Faster than you can guess... (Score:1)
However, while the article summary blathers about "from hard drives to main memory", this is not a competitor to modern DDR SDRAM. Assuming the quoted 500X faster than NAND is accurate, MRAM latency should be on the order of 100 nanoseconds for a random read. (NAND read latency is on the order of 50 microseconds.) DDR SDRAM random read latency is on the order of 22 nanoseconds.
Why not just visit website of the people who actually make the thing rather than guessing... it's 35ns that's pretty damn close for a just to market technology (how long has SDRAM been around?
source: http://www.everspin.com/products.php?hjk=16&a1f3=0 [everspin.com]
Re: (Score:1)
Re: (Score:1)
Re: (Score:1)
If manufacturers get really silly, in theory a mobile device could be built that doesn't distinguish between its main memory and its mass storage. The two functions would be served by the same solid state circuitry. Obviously accommodating such a hardware design would give the kernel guys fits, but it could simplify things in the software a great deal, and incidentally net an interesting performance gain that's visible to users.
Why would it gives kernel guy fits ? How would it be different to have a swap partition ? Anybody could say me what would be the problem running with no memory and only swap ?
Re: (Score:2)
If the only swap you have is a spinning disk, I suspect Linux will not work reliably. There are timeouts in many many places that would be breached repeatedly if all execution required spinning disk I/O.
But see one of the other replies about how the Linux kernel is already outfitted with Execute In Place and everything needed to accommodate the architecture I described, as long as the storage device is fast enough. It has been pointed out elsewhere in the comments that a unified single storage architectur
Re: (Score:3)
Universal law of YouCan'tHaveItAll (Score:4, Informative)
non-volatile, low-latency, low-cost
AGoodThing, AnotherGoodThing, YetAnotherGoodThing, pick any two.
Re: (Score:2)
I'm hoping the density of memristors, being speculated at an order of magnitude better than flash, will lend themselves to low cost. They already have the other two. This is why technologies like this are speculated at becoming a "universal" memory. It's all hope and dreams at this point. I'm sure lots of people at one point and time would have not believed for a moment that flash could ever become a consumer harddrive.
Re: (Score:2)
As possible speculative circumstantial evidence of this, HP is delaying memeristors due to concern that it would cannibalize their partner's flash business. I think that's a bad move anyhow. They should take the opportunity to be ahead of the game for awhile with an exclusive product, and charge a premium for it. The extra premium would offset losses of flash sales(which really would not be much because there's a good chance the Joe Shmoe was going to by a competitor's flash anyway).
Re: (Score:2)
Or you know, just get ahead of the game with the superior product, undercut the market and take over.
Everyone wants solid-state mass storage on a large scale, and the market flocks to the lowest cost-per-gigabyte.
Re: (Score:3)
I guess non-volatile could be found as a negative. eg. It can be a security problem waiting to happen.
Re: (Score:2)
Re: (Score:1)
Must...save..porn...Oh, that was fast!
Re: (Score:3)
Pick any two.
The breasts?
Re: (Score:3)
Re: (Score:3)
Oh noes 800mw. Where ever will be get such levels of power?
The cost is a much bigger issue.
Re:power (Score:5, Interesting)
It uses a lot more power while you are using it. Because it runs 500x faster you have to use it for a lot less time though, and it doesn't need power to retain state.
It uses 5x more power for that 500x performance. Of course, people will think up new ways to use that kind of performance.
Re: (Score:1)
Not to mention that its fast enough you could use it for primary memory on some systems. Eliminate a whole component out of the system - that saves cost and power. I'm guessing we'll see this in embedded stuff first.
Re: (Score:2)
That's how the Palm models originally worked, though with battery-backed ram instead of flash.
Power, Price, and Density (Score:2, Informative)
Speed = 500x
Price = 50x
Density = 1/64x
Power = 5x
So what you gain in speed, you lose in density, power, and price. Still, if it makes it 500x faster to boot a device, then you could imagine this being great for the embedded market as a boot-up device where the OS resides. The only problem is the 5x power consumption requirement. Maybe the power consumption should be compared to SDRAM, and this might be a good replacement -- imagine not having to wait for the OS in a mobile device to have to write to flash wh
Re: (Score:2)
Re: (Score:1)
what if you could boot in nanoseconds?
I don't know much about modern computer engineering but hell, if you can wake from sleep that fast, you might as well sleep every time there is no schedulable thread.
...Weird to think about your computer sleeping between each key press, or between each network packet. You'd probably want to keep the display powered a little longer though ;)
Re: (Score:2)
Re:power (Score:5, Informative)
For those curious, it performs 500x faster than NAND, costs roughly 50x more than NAND, and uses 5x more power than NAND. All-in-all, not too bad, considering it's new technology and is actually shipping, but it definitely has limited applications at the moment. Assuming they can get the cost down a bit or come up with a few more ideas to reduce power consumption (it's actually worse in older MRAM), it could be something interesting in the near future. I'm guessing MRAM will be showing up more and more often in the next few years, since it seems like it's finally cracked the wall between "cool in the lab" and "semi-practical" after years of being stuck.
Re:power (Score:5, Insightful)
5x more power than NAND
This seems confusing to me, because arguably it's going to use significantly less power than NAND. If I have something to write and it takes NAND 10s at 10w to write it, that's 100J of energy. MRAM would take .02s at 50w, that's 1J of energy. Unless I'm missing something? Seems like they could have quoted that to be both more accurate and show their product in a better light.
Re: (Score:2)
Re: (Score:3)
Bravo, you both understand and fail to understand something at the same time!
Oh my gosh, he's in a quantum-super-imposed state of consciousness!
I haven't seen that one before.
Damn it, now you've observed him and collapsed his waveform. I bet you enjoy killing boxed cats in your spare time, too.
Re: (Score:2)
Power is energy over time.
I.e. 1 W is 1 joule per second.
The reason this technology requires more power, is that the state changes and reads require more power to be activated than for NAND.
They use less energy, because in order to read or write the same amount of data, they need significantly less time than for NAND.
Since the scale for speed is so much bigger than it is for power, the end result is a reduced energy need - but you still need more power than for NAND.
Re: (Score:2)
i don't think you get what he means:
it uses more power when reading/writing
but if your computer writes/reads the same data over a period of time, it'll use less power over that period of time (because as you said, it'll use less energy to perform those actions, over this time period, that's power)
It's maximum power draw is indeed 5X higher, but its average power draw under real workloads will be a lot lower?
Re: (Score:2)
No.
100 W for 10 seconds is the same amount of power as 100 W for 10 minutes, but it's not the same amount of energy.
It's a bit like speed (meters/second) vs distance (meters). If you travel at 10 m/s for 10 seconds and 10 m/s for 10 minutes, it's the same speed (10 m/s) but not the same distance.
Re: (Score:2)
This is /. sir. Using math to solve problems is strictly prohibited. +5 banned for life.
Re: (Score:2)
This is for your sig only.
I wonder how many of those people are illegal immigrants.
I don't know about you but for people who think illegals should be deported should also not count them for values like those.
Even money just giving the poor people wouldn't need to go to the illegals since they would be sent away.
Re: (Score:2)
Those numbers are only people who responded to the census.
Re: (Score:1)
Re: (Score:1)
Re: (Score:1)
You do know that spinning platters also have a limited lifetime right? Not as low as NAND but it does have a finite lifetime.
Re:It is very volatile (Score:4, Informative)
Re: (Score:2)
The post was talking about NAND, not MRAM, when it mentioned limited read/write cycles. Current Flash memory technology (i.e. NAND) has an average of about 10,000 read/write cycles before it starts to fail. MRAM does significantly better in this regard.
Re: (Score:1)
And this number of read/write cycles for NAND flash is continuing to fall. Most consumer NAND flash SSDs are now rated for more like 3,000 cycles.
Re:It is very volatile (Score:5, Informative)
It's actually very durable. "In contrast, MRAM requires only slightly more power to write than read, and no change in the voltage, eliminating the need for a charge pump. This leads to much faster operation, lower power consumption, and an indefinitely long "lifetime"." (https://en.wikipedia.org/wiki/Magnetoresistive_random-access_memory)
Re: (Score:3)
What does non volatile mean if it is not durable?
I can't believe I'm having to answer this question on Slashdot, but "non-volatile" in this context means that you can cut power to it and it'll retain the information. The RAM in your computer is an example of volatile storage, since it requires a constant charge in order to preserve the data. NAND, MRAM, tapes, DVDs, HDDs, and a load of other technologies are non-volatile, since they can safely go without power while retaining the data contained within them.
What you're talking about is something else entir
Re: (Score:2)
Unless you're recording video most of the day non stop, or have a DB processing a huge number of transactions(which should be on a server, not your desktop), then a quality SSD should outlast your harddrive. If SSD's had been the forerunners, followed by HDD's, people would be talking about all the things that go wrong with an HDD at 3-5 years lifetime. Except they would probably be just oblivious of the all the failure scenarios that an HDD can suffer. At least with an SSD there is a predictable wearing
Cache for SSDs? (Score:5, Interesting)
It has much higher performance flash and persistence but at a big cost in size, power and money. I think this sounds like good case for using it as write cache for SSDs that you don't need to flush. Imagine for example a log file that's very volatile, a line gets written every few seconds. Or that document or spreadsheet or email you're working on that Office auto-saves all the time or game autosaves for that matter. With this you could commit it to MRAM and it'd be written "for real" even in case of power failure with no supercap to flush to NAND without wasting write cycles on it. They say a 50:1 cost compared to NAND so on a 256 GB SSD a 512 MB cache should add ~10% to the cost.
If you only need to push the most stale writes to NAND you could download a 50MB installer, install it using 100MB writes then delete the installer and it'd never need to touch the NAND at all - it's marked free again before it's ever written to disk once. Oh yes and you'd also get better burst IOPS as a bonus. If it really can't be worn out like RAM that is going to be huge, even if it just comes on top of the technology we already have and doesn't replace anything. After all, most of my SSD is the same from day to day - the "active set" that gets written to is much smaller.
Re: (Score:1)
Yes, they have said as much in their press release: http://finance.yahoo.com/news/everspin-debuts-first-spin-torque-140000887.html [yahoo.com]
Re: (Score:2)
Are we going to have hybrid HDD's with an SSD and then MRAM and then the usual volatile cache?
Actually MRAM, if fast enough to replace traditional caches on RAID cards and Hard drives, could(barring other complications) eliminate many of the problems where you lose power or some other failure occurs and causes incomplete writes that hadn't been flushed from cache. It would hold the cache and could complete the operation when the system recovers.
Re: (Score:2)
This would make it's low density less of an issue perhaps.
Re: (Score:3)
I'd count on it. People's appetite for storage is not going away, and SSD pricing isn't coming down fast enough. Due to the way most data access patterns work, making ALL of your storage super fast (be it cpu registers, cpu cache, RAM, etc) is far more expensive than it needs to be to get 90% of the performance.
I see this as being used for NAND based SSD write cache, and the SSD being used for cache of spinning disks in the interim, and MRAM eventually replacing NAND entirely if they can get the densit
Re: (Score:2)
A big capacitor would be cheaper. Just use ordinary DRAM as a cache, flush it to NAND when the power goes off using the juice left in your capacitor. Some RAID cards already support it, and I seem to recall someone makes DDR2 based RAM drives that have batteries that allow them to save their contents to CF card after the machine turns off.
Re: (Score:3)
So, it sounds like these could be immediately deployed for an ext* external journal or ZFS zil/log device.
Pretty sweet, now somebody just give me a barebones PCIe card with DDR3 slots and a linux block device driver.
I'm confused, or ill-informed (Score:3)
Everspin has announced that it's shipping the first 64Mb ST-MRAM in a DDR3-compatible module. These modules transfer data at DDR3-1600 clock rates, but access latencies are much lower than flash RAM, promising an overall 500x performance increase over conventional NAND.
Wait, so, is this to replace RAM (the mention of DDR3) or to replace drive storage?
These modules transfer data at DDR3-1600 clock rates, but access latencies are much lower than flash RAM
Isn't that comparing apples (DDR3) and oranges (flash RAM)?
Re:I'm confused, or ill-informed (Score:5, Informative)
It can actually replace both, which is pretty interesting and might change how our current computing model is built.
There are already applications and systems in place to model the data storage like this, for example memory-mapped file I/O, where you basically tell the operating system that "please let me pretend that this huge file on the hard drive is already in RAM", and let the RAM be some sort of huge cache. The same model would apply to storage here, except we would get rid of the whole RAM layer between storage and CPU.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
To me that seems like the Holy Grail of storage, a universal high capacity, high speed storage medium that is disk AND RAM in the conventional model at the same time.
Even if it was slower than current RAM (up to a point, at least), eliminating paging as an entire concept and eliminating disk/ram/cpu/ram/disk cycle ought to make it overall faster.
And it would be really cool if it was as modular as storage is now, allowing you to tack on an additional "disk drive" whenever you wanted, although it would requir
Re: (Score:2)
It's a technology that fits between them.
It's faster than Flash, but not as high-capacity as DRAM (and it's probably a bit slower than DRAM as well). Just like Flash-based SSDs fit between DRAM and hard drives.
Right now, it uses the interface of DRAM. They could probably have shoved it in a PCIe or even a SATA interface, but it was more logical to use the fastest possible.
Overall, I'm not sure where it's headed. If they can't get density up, it's doomed to niche uses, like embedded hardware. Maybe replace b
Re:I'm confused, or ill-informed (Score:5, Informative)
Everspin has announced that it's shipping the first 64Mb ST-MRAM in a DDR3-compatible module. These modules transfer data at DDR3-1600 clock rates, but access latencies are much lower than flash RAM, promising an overall 500x performance increase over conventional NAND.
Wait, so, is this to replace RAM (the mention of DDR3) or to replace drive storage?
MRAM might be a potential candidate to replace current solid state storage (NAND-flash) which is a candiate to replace drive storage. In a system with small amounts of DRAM, MRAM might be used to replace the DRAM as well. Unfortunatly, because of its current high price and low density, it is currently not very good substitute for either one except in perhaps a very small embedded system.
These modules transfer data at DDR3-1600 clock rates, but access latencies are much lower than flash RAM
Isn't that comparing apples (DDR3) and oranges (flash RAM)?
Instead of implementing the slow standard flash memory electrical interface on MRAMs, they (everspin) elected to support the same fast electrical interface that DRAMs use (DDR3). They can do this because just like DRAM, writing data on MRAMs is about as quick as reading data (which isn't the case with NAND-flash). By choosing the standarized DDR3 interface, chips that might want to use these MRAMs won't have to be specially designed to do so (which wouldn't be the case if they came up with a non-standard interface). It will apparantly just look like a small capacity DRAM chip that doesn't forget when you take the power away (I'm guessing the MRAMs probably also ignore any refresh requests that come across the interface).
The reason that current flash memory electrical interfaces are slow, is that flash memories have pretty slow access times and are read/written in large blocks. This led to an efficient interface that multiplexes the address and data on the same pins. DRAM is however more randomly accessed in smaller blocks and has separate address and data pins. This allows a higher duty cycle of data transfer on the data pins for smaller transactions: you don't have to constantly turn the bus around between sending commands and reading data, and you can pipeline new addresses on the address bus at the same time data from older commands are transfered on the data bus.
By targetting the DRAM interface, it appears that Everspin is positioning their chip as a DRAM+Flash replacement for systems that don't require much total storage. They need to target the DRAM interface for this because you can't really do random access efficiently on the flash interfaces (but you can do block transactions on a random access interface). In fact in many embedded systems, the first action of the bootstrap code is to copy parts of the NAND into DRAM (for fast access). With MRAM, you could just bypass this step.
Re: (Score:3)
If they can get the price down significantly, I could see this implemented in things like settop boxes and home based networking equipment.
I could also see it being used intelligently on a bootable network card. (Card has say, a 128mb mram capacity. That's enough for a very robust preboot environment, as well as an actual internal boot image, an MD5 hash cache, and all the software it needs to make sure network and local images match. Basically, the card boots the internal preboot env, then MD5s its local i
Price? (Score:1)
Re: (Score:3)
50 per GB
http://www.extremetech.com/wp-content/uploads/2012/11/Flash-vs-MRAM-performance.jpg [extremetech.com]
Gauging from that comparison image i'd assume it is 50$ per GB as they are comparing it to 1$ per GB NAND which is upper consumer market price right now.
RTFA? :)
Re: (Score:2)
Yes, the quote I saw was "The company said the first chips were about 50 times the cost of flash memory by size".
As I had guessed previously, they are touting it as an SSD cache as one potential application. So for an extra $25 maybe you get 512mb cache on your SSD, and hopefully protects you from lost write buffers during a power loss since it's non-volatile. That last bit is speculation though, it could be more complicated.
Different writing technology (Score:4, Insightful)
holy grail..... NOT (Score:1)
500 times faster and 1000 times smaller and this should have a title even mentioning NAND because why?
It would be like going "New SRAM just produced is much faster than DRAM!!!!" without bothering to mention those minor issues of size, cost, power, etc that make SRAM != DRAM.
PCM? (Score:2)
What happened to PCM otherwise known as Phase Change Memory or PRAM [wikipedia.org]? From what I have read it can be written to like normal memory, address by address unlike flash that is block by block (very important for write speeds). It also boasts faster read and write times as well as one hundred million plus write cycles per memory location vs. flash is what, five thousand. PCM has a memory retention of 300+ years which makes it attractive for long term archival. It is also in production and shipping but density and
Re: (Score:2)
Not Found
The requested URL
VoxCAST Server at imgs.xkcd.com Port 80
Re: (Score:2)
It's solid state flash so it doesn't need refresh nor does it have destructive reads. It's just faster than current SSD's but slower than current gen RAM (1ms vs 100ns vs ~10ns)
100ns RAM was common circa the 486