New kind of components 100-1000x faster than flash in SSDs?!

[Fi2Pro]

Some progress is happening with a new fast storage technology: ReRam (Resistive RAM).

For some time now scientific companies have been working on developing a new generation of storage components that can be used to replace flash components in smartphones, SSDs, IOT…

This component is “stackable”: it can be built in a 3D format where you could have different layers of storage on the same component.

The company at the forefront of its development at the moment is Weebit Nano and they are doing it in collaboration with the CEA-Leti (a technology developing branch of the CEA (CEA includes the CERN where they discovered the Higgs Boson!!!)).

 

I have a video on the subject of Weebit Nano’s ReRam here:

 

It seems the process is well on its way since they are now in the miniaturisation phase.

https://www.weebit-nano.com/media-2-2/press-release/

 

They claim ReRam is 1000x faster than flash and it shall use a fraction of the power that flash uses. It can even be made to be see-through!!! https://www.weebit-nano.com/technology/

 

 

A different website, Crossbar: https://www.crossbar-inc.com/en/ 

even states: “With ReRAM you get 1/20th the energy; 1000x endurance; 100x read performance; 1000x write performance and terabytes of storage on-chip.”

Crazy!

 

Western Digital have stated in 2016 their interest in this technology for their SSDs -following their own sourcing it seems- which makes a lot of sense. Source:

http://www.anandtech.com/show/10562/western-digital-to-use-3d-reram-as-storage-class-memory-for-specialpurpose-ssds

 

When these components hit the market, we should get superfast storage that enable longer battery life and metric tons of storage. That will definitely be of interest to multitaskers or computer enthusiasts and everyone with a portable device.

The last question to answer will be: How soon will we get it?

They’re talking 2018… Only time will tell. One thing’s for sure, the sooner, the better!

Happy storing!!!

 

[mariushm]

Yeah yeah ... lots of claims... but it's probably 10-20 years to being feasible to be mass produced.

 

We already have MRAM which is much better than NAND flash but so far it can only be made in small sizes and it's expensive : http://www.anandtech.com/show/11184/everspin-announces-new-mram-products-and-partnerships

 

Much faster, huge endurance, can write one byte at a time if you want unlike nand where you must write blocks of a few hundred bytes, can erase one byte at a time instead of 512 KB pages as is the case with SSDs , huuuge erase cycle count ... guaranteed write time for any bit unlike nand, which makes them great for real time critical operating systems (think nasa, nuclear plants  etc)

 

Seagate has a prototype with a 128 MB of MRAM (probably 2x64MB or 4 chips x 32 MB each) used to hold critical data to speed things up and they're thinking of using mram in more ways (scroll down to laste picture): http://www.anandtech.com/show/11721/seagate-demos-new-ssd-features-and-prototypes-at-fms

 

[Ezio Auditore]

Is this counted as self promotion @leadeater @Ryan_Vickers @TheRandomness?

[Bananasplit_00]

Yah sure but this ain't solving the fact that neither power, read or write speed is my main bottleneck in downloading shit, it's my God damn internet connection. We already don't need NVME performance in most cases and while I LOVE tech moving forward this is just kinda overkill as shit for everything but maybe Google and Facebook. Making more power efficient storage for laptops is nice but that isn't a major power drain already. 

[Juanitology]

Just FYI RRAM is nothing new. It's been worked on forever now. It's one of those things that has been "5 years away" for like 20 years now. 

[Fi2Pro]

Thanks mariusm

 

Didn't know about MRAM, interesting!

 

It'd be nice to see something radically new to shake things up though...

 

Claims are high, but you never know...

 

The competition between Intel and AMD restarted and the consumers are benefiting from it. If ReRam was to come to the market soon, that would be a game changer, we've had tiny storage on smartphones for too long...

[Fi2Pro]

9 minutes ago, Ezio Auditore said:

Is this counted as self promotion @leadeater @Ryan_Vickers @TheRandomness?

Understand your concern, that's why I waited (check video release date and number of relevant links in post against the one link  ). Doesn't make the matter still relevant though...

[Fi2Pro]

Regarding the timeline...

 

True that some things can take forever to see the light of day but, some collaborations between CEA-Leti and other companies have produced good stuff.

 

Here's a taste: https://www.globalfoundries.com/news-events/press-releases/globalfoundries-extends-fdxtm-roadmap-12nm-fd-soi-technology

 

http://wccftech.com/globalfoundries-begins-12nm-fdsoi-rollout-product-tape-due-years/

 

http://www.cea-tech.fr/cea-tech/english_old/pages/news/press releases/leti-reports-finfet-feasibility-and-circuit-design-for-better-area-speed-and-power-trade-offs.aspx

 

We never what will take and what not... crazy!!!

[DocSwag]

You may say, "1000x! This is gonna be revolutionary!"

 

Well the thing is, 1000x is the exact same thing Intel touted with 3D XPoint a few years ago. And technically they were correct; at the flash level XPoint is 1000x faster than nand. But once you factor in controller shenanigans, pcie bus limitations, etc., you get nowhere near that much. Don't get me wrong, XPoint is quite an interesting product and has really amazing QD1 results but still.

 

Also you have no idea what the cost of the stuff is gonna be like. Or the density.

[BachChain]

https://www.extremetech.com/computing/163058-reram-the-new-memory-tech-that-will-eventually-replace-nand-flash-finally-comes-to-market

 

A four year old article about a working reram prototype

[Fi2Pro]

Good article comparing strenghts and weaknesse of future possible types of memory:

https://semiengineering.com/new-memories-and-architectures-ahead/

 

"The list of new memories vying for attention include Phase-Change memory (PCM), Ferroelectric RAM (FeRAM), Magneto-resistive RAM (MRAM), Resistive RAM (RRAM or ReRAM), spin-transfer torque RAM (STT-RAM), Conductive Bridging RAM (CBRAM) and Oxide-based resistive memory (OxRAM)."

 

As the article states, the IoT market will be hundgry for cheap memory with low power consumption hopefully that will help bring costs down. (Talking about prices, I know ReRAM is silicon oxyde based and could be produced by the same factories as now with the same type of material, so they wouldn't have to change everything to be able to produce that kind of memory, I'm not sure about others though).

[Doobeedoo]

This will be amazing in every aspect. I remember reading about it before. Somehing of closer to RAM speeds with way better endurance and capacity along. 

SSD storage is evolving at a great pace! Even capacity, price got down in general but still more to go to close in on HDD price. That will be exciting as well. 

[Misanthrope]

On 8/17/2017 at 10:03 AM, mariushm said:

Yeah yeah ... lots of claims... but it's probably 10-20 years to being feasible to be mass produced.

 

We already have MRAM which is much better than NAND flash but so far it can only be made in small sizes and it's expensive : http://www.anandtech.com/show/11184/everspin-announces-new-mram-products-and-partnerships

 

Much faster, huge endurance, can write one byte at a time if you want unlike nand where you must write blocks of a few hundred bytes, can erase one byte at a time instead of 512 KB pages as is the case with SSDs , huuuge erase cycle count ... guaranteed write time for any bit unlike nand, which makes them great for real time critical operating systems (think nasa, nuclear plants  etc)

 

Seagate has a prototype with a 128 MB of MRAM (probably 2x64MB or 4 chips x 32 MB each) used to hold critical data to speed things up and they're thinking of using mram in more ways (scroll down to laste picture): http://www.anandtech.com/show/11721/seagate-demos-new-ssd-features-and-prototypes-at-fms

 

We're probably also 10-20 years away from computational power to generate large amounts of content 100 to 1000 times faster so it actually lines up well.