Intel and Micron Unveil New Memory Class That's 1,000 Times Faster Than Current Technology

[Rocketman]

I was thinking of buying a new computer next year (was waiting for Skylake and Thunderbolt III), but this throws a monkey wrench in things. How long until we see actual consumer products? How much better will they be over the next few years than conventional tech? How much will they cost by comparison and how long will it take for the price to come down.

The articles I have read state this is largely for mobile devices but can also be applied to laptops and desktops. I also see a 2+ year time horizon so in an Apple world more likely 3+. Mac OS 10.10.3 already supports the associated bus standard this hardware requires for interface. It shipped with MacBook Pros last year. So that means Apple has both OS and hardware with which to test R&D units of the memory hardware with shipping product.

This is on the near horizon.

Rocketman

What happens when you have "big data" at reasonable speeds on your iPhone 8 handtop?
In 1995 the leading supercomputer was 10^11 FLOPS. What is an iPhone6?
When the iPhone was released in 2007, this was the top 500 list. http://www.top500.org/list/2007/06/

 

[subsonix]

Certainly, it'll likely replace all current SSD technology if it really delivers, but ultimately your bus speed is going to limit it. Put succinctly, SATA III isn't going to cut it...not by a long shot. Even Thunderbolt III isn't going to be enough. NOTHING IS, although clearly Thunderbolt III is preferable. This stuff will be inherently limited as external (or even external to main memory) storage by the bus connections to it.

SATA isn't used anymore for the fastest SSDs but PCIe, and since there are busses fast enough for DRAM, there will be busses fast enough for this stuff. Assuming it's actually performing that good. Most of it looks like a PR puff piece with little information about real performance. For example, the 1000x figure refers to a best case comparission to NAND latency, not throughput. And is this comparisson made to NAND or NAND hidden behind a complex controller and a cache on a modern SSD which hides some of this.

 

[MagnusVonMagnum]

I should add, not just bus speed improvements, but improvements in CPU speed, and many other components, as well as their busses. A 1000x improvement in storage/memory speed literally gives you nothing, if your system busses and other main system components can't take advantage of it. Systems need to be designed as... systems. Go figure!

I don't think some people are seeing this for the potential game changer it is. SO WHAT if some areas aren't used to their absolute maximum when it's first released? Just because you only have a basketball and not a regulation hoop and waxed wooden court doesn't mean you can still play with the ball. In other words, it's a product that will be immediately useful (it will go as fast as your bus will allow it and you won't need to RAID0 2-3 SSDs to saturate a Thunderbolt II or III bus. One will do just fine. One should note that SATA3 is worthless already for a RAID0 configuration with SSDs (6Gbps compared to 20Gbps for Thunderbolt 2 and 40Gbps for Thunderbolt 3). Fortunately, there is "M.2" that connects SSDs DIRECTLY to the PCI bus internally (and PCIe for external). That will get you 2GBps (yes Gigabyte, not bit) and Thunderbolt 3 can do 5GBps. That's a big improvement for ONE drive, but still not nearly fast enough. They need a faster PCI bus and they need it sooner rather than later. Still, the point is that this solves one problem for "next gen" technology and puts the emphasis on BUS speeds. I don't think many saw a NEED for a faster bus until now.

Unless they solved the issue with the finite number of read/write cycles, it won't be useful as system memory.

That is a good point. This supposedly greatly increases the number of cycles (1000x more, was it?), but given the amount of cycles involved in actual computational work, it might not be suitable for computationally intense environments (i.e. might be OK for a Phone designs to last a couple of years, but perhaps not a computer doing heavy data work 24/7). How much work is done in the CPU cache versus actual writes to memory, though? I'm not in a position to guess even. I would assume the more memory you had the longer the machine would last if it followed SSD patterns of assigning new cells rather than used ones first. Thus, one could potentially get a usable lifetime by buying more "memory" than they actually need for a given time frame if the price came down enough. (e.g. Buy 1TB when you only need 250GB of actual disk space and it will cycle through the empty space for RAM use). Certainly, in the long run something like that might make it usable along with improvements in the tech itself.

 

[MagnusVonMagnum]

So fa


So far the dumbest comment I have read in this thread but hey only on page 2 so far.

Given how much you just said (aka NOTHING except the insult to my post), I'd say you just took over the lead, guy.

 

[spatlese44]

Not "1000x speed" but "up to 1000 times faster". I have the impression that they are capable of reading single bits. An SSD drive will always read 4,096 bytes at a time. So if you want to read a single byte they might be 1,000 times faster because your SSD drive must read another 4,095 bytes. But for reading the whole 4,096 bytes, or a 100 megabyte file, the difference might be a lot smaller or non existing.

It was difficult to tell from the article, but I watched the video on Micron's site and they were fairly clear that both the technology involved both access to individual cells and faster individual cell read/write times. It's faster.

What I really found interesting is how much room they left for improvement. The technology allows stacking of layers and they only are implementing the lowest possible level of stacking (2). All they have to do is put 16 layers on a chip and it's 128 GB not Gb. They also weren't shy talking about potential for shrinking the process size. I was left with the impression that 1 TB on a chip is around the corner. This thing is the real deal.

http://www.micron.com/about/innovations/3d-xpoint-technology

 

[Bstephens]

Gotta wait and see how durable it will be because conventional ram is practically bullet proof.

 

[SactoGuy18]

I think with the new 3D XPoint technology, if the storage drive using this technology is connected to the PCIe 4.0 bus, we're talking a full boot from cold start of Windows 10 or MacOS X that could be completed in 2-4 seconds! Now that would make it possible to have "instant on" computers with vastly more computing power than any iOS or Android-based tablet.

 

[Derived]

1000x faster at what, writes, reads, setup or latency. We don't have all the information yet, they could be comparing it to some USB dongle flash from a decade ago.

No they couldn't be, why would they do that. Also, it mentioned "Current NAND technology" so...no, I doubt they're comparing it to a cheap USB flash drive from a decade ago...

 

[Derived]

I hate vaporware news releases like this.


Can't wait to see the speed enhancement in 2028.

You hate when people talk about technological progress? That's quite odd, on a computer forum.

 

[beingcurt]

This all sounds amazing but reminds me of the claim of unbelievable battery tech a little over 2 years ago (1000x faster charge & 30x more power)

http://www.engadget.com/2013/04/17/researchers-devise-fast-and-powerful-microbattery/
http://techcrunch.com/2013/04/18/re...hat-charge-instantly-and-hold-hours-of-juice/
http://mashable.com/2013/04/17/battery-breakthrough/

 

[ardent73]

"The companies invented unique material compounds..."
Definitions of "new" and "flavor" may vary, but this sounds like more than just changing the doping profile.

New is Nobel Prize worthy, while unique just bumps the stock price and the bottom line.

 

[Analog Kid]

New is Nobel Prize worthy, while unique just bumps the stock price and the bottom line.

Hm. With nothing useful in between?

I thought this might be a high bar for such a common adjective, so I did a quick tally. Assuming we aren't talking about economics, literature, physiology and peace (leaving chemistry and physics), there have been somewhere around 218 new things in the world since 1901.

I think the tone you were going for was cynical, but you pushed it across the line over to silly.

 

[maxsix]

It will certainly shrink the cloud market, but not necessarily kill it. People will still want to backup in case of, say, robbery, or fire.

Backing up my data has been a habit, something I've practiced for many years. I even keep an off site copy in case of fire. It's so routine often I don't even give it a second thought.

A few weeks ago I came across a recent study that revealed less than 23% of the American Public actually back up their data at all, much less doing so regularly. While I wasn't surprised that many don't, I must admit that I'm shocked at how low the number is.

As I see it, Time Machine is such a valuable and easy way to backup ones Mac, there's no reason why people wouldn't use it. Oh sure they've got lots of excuses and such but I'd bet they either live in denial or don't value their work.

 

[retep42]

This all sounds amazing but reminds me of the claim of unbelievable battery tech a little over 2 years ago (1000x faster charge & 30x more power)

http://www.engadget.com/2013/04/17/researchers-devise-fast-and-powerful-microbattery/
http://techcrunch.com/2013/04/18/re...hat-charge-instantly-and-hold-hours-of-juice/
http://mashable.com/2013/04/17/battery-breakthrough/

Ouch!! And these days aluminum (instead of lithium) batteries are the next big hype...

But there is a difference between a professor at a US university trying to hype up his research funding and two companies shopping a new product around to customers. If nothing else, Micron and Intel would not want to give their competitors ideas before they themselves have a real product ready, so I am more inclined to believe them.

 

[anthonylambert]

"This is going to literally KILL THE CLOUD for significant data storage

Was the cloud ever alive for large data storage? If I wanted to back up my 2TB drive to the cloud. Basically it would be quicker to take my disk drive and get on a plane from anywhere on the planet to any where else and deliver it personally !

This is true today.... This RAM makes no difference.

 

[goinskiing]

Pointless until a bus, ports and other I/O can feed data at that rate.

Actually, that may be moot because those busses assume you have separate CPU, Memory, and Storage. The idea here is that you would eliminate the bus between memory and storage (one of the limiting buses you speak of) and tie this directly to the CPU.

 

[MagnusVonMagnum]

"This is going to literally KILL THE CLOUD for significant data storage

Was the cloud ever alive for large data storage? If I wanted to back up my 2TB drive to the cloud. Basically it would be quicker to take my disk drive and get on a plane from anywhere on the planet to any where else and deliver it personally !

This is true today.... This RAM makes no difference.

Actually, with Google Fiber backing up 2TB to the cloud isn't much worse than backing up to something like a 2-4TB external WD hard drive (I use a 3TB USB 3.1 right now and 120MB/sec is about what I can expect at best and Google Fiber can do that so Cloud backup in that regard, at least looks viable. I personally wouldn't want my data all going over the Net, but I doubt the average person even considers it.). My comment was based on all the articles saying the Cloud is the future for data and more backup services have been becoming available since most people are connected all the time. Even my own ISP is pushing Carbon backup by offering a free initial offer. But this speed increase is potentially so severe as to make it moot entirely once again. The price of the new storage may be cost prohibitive for large drives, though. Only time will tell.

 

[Rocketman]

The price of the new storage may be cost prohibitive for large drives, though.

This tech is for "unified memory" for mobile devices where the power and throughput requirements are low. It may have purpose in dedicated drives for servers for momentary offline uses. The current trend in live services is massive ultra-fast RAM. So this could be cache for that RAM.

Mobile. At least for the first three years.

Ironically tape still does not suck.

Rocketman

 

[MagnusVonMagnum]

This tech is for "unified memory" for mobile devices where the power and throughput requirements are low. It may have purpose in dedicated drives for servers for momentary offline uses. The current trend in live services is massive ultra-fast RAM. So this could be cache for that RAM.

Mobile. At least for the first three years.

Ironically tape still does not suck.

Rocketman

Numerous articles (including Intel's own) suggested one of its first uses will be to take over the SSD market with much faster, longer-lasting drives (as in 1000x longer lasting and 1000x faster). So no, I wouldn't peg it as only for unified memory stuff. It may or may not take off for that, being slower than DRAM makes it undesirable for high-end computers in that task as it will slow it down. But IF they can offer it at a reasonable size/price, it will kill all current SSDs on the market in short order. Who would want a drive that is 1000x slower and lasts 1000x less time at the same price? NO ONE.

 

[macduke]

7
7s
8
8s
9
9s
10
10s
11
11s
12
12s
13
13s
14

...so however many years that is.

Well the 6 was the eighth iPhone and the first to offer 128GB. The high-end original iPhone was offered in 8GB capacity. Going by that logic, eight years after the iPhone 6 should give us the iPhone 10 (X?). That device should max out at (128GB / 8GB) * 128GB = 2048GB, or 2TB. Realistically most people won't need that much storage space, and the cloud will become even more important over time. But we might see 1TB by the iPhone 10? Although I get the feeling that we're far more likely to see much faster storage before higher-capacity as currently, speed is a big issue with mobile storage.

 

[flat five]

Well the 6 was the eighth iPhone and the first to offer 128GB. The high-end original iPhone was offered in 8GB capacity. Going by that logic
[...]
But we might see 1TB by the iPhone 10?

maybe.. but Col4bin is specifically wanting an iphone 14.. so that will still be a few years after iphone10.


(though in reality, i expect there to never be an iphone 14.. it will be called something else or replaced with something else by that time.)

 

[G4er?]

Apple won't use it without a name change.
Anything called 3D doesn't fit with Apple's thinner at every cost mantra.
*It might not actually be thicker, but it sounds that way.

 

[Mattsasa]

Interesting:


There will always be memory, but if they use the same storage pool for RAM and (P)ROM then the contention for storage is going to become even tighter. Your available memory for execution is going to start to depend on how many apps you have installed, and how many pictures you've taken. How much free space will you need for your app to run? Have you taken too many pictures since you last checked?

I suspect Apple will still keep the memory segmented, just to avoid having to educate users. Most of my family doesn't understand the difference between their 8GB of RAM and their 2TB HDD...

But say you had 2TB of xpoint memory,
Even if you have 1.5Tb used up. Then you would still have 500GB of memory.

Software could be designed to not let you save any more space on your storage so that there is not enough for memory.
Like say you have 10gb of space left, the OS would not let you install any more apps or take any more pictures because it is automatically leaving room for memory

 

[Data]

It is not clear if Apple will adopt 3D Xpoint when it first becomes available, but the company does often aim to implement cutting edge technology in its devices for maximum speed and performance.

Article Link: Intel and Micron Unveil New Memory Class That's 1,000 Times Faster Than Current Technology

LOL, Sure just look at the thunderbolt display with its superfast USB 3 connections, and they were one of the first with that too.....Oh wait .....

 

[Jvanleuvan]

this could be revolutionary;
1) Get Rid of Dram/Storage, and just have 1 memory to keep everything.
2) Reduced power consumption as DRAM memory strobes would be gone (power to keep dram alive)
3) Hibernation in the tradition sense would be gone, Not need to transfer volatile memory to permanent storage.
4) Loading would be gone, no need to move data and files into memory, just execute right off of permanent storage

5) New type of power saving mode, move network and sensor functions to a discrete, low power chip (i.e the m7), then when a device isn't in use simply turn every thing in the device off, CPU, GPU, Bus, everything except that network chip which would wake the computer up when needed (i.e. a Phone call); now "waking up" isn't something that needs to happen, the CPU can just pick up right where it left off without transferring data at all.

-this is different than current sleep mode, that simply puts the system in "low power" mode, I'm talking about TURNING OFF THE "COMPUTER" completely, this might even be doable in periods of no-activity between tasks, instead of having the CPU cycle waiting for an instruction, just turn it off and wake it up when an interrupt comes in.

just some possibilities.