Memory-making start-up Crossbar came out of stealth mode on Monday, announcing that it has developed what it characterizes as "very high capacity and high-performance non-volatile memory" based on a new approach to resistive RAM (RRAM) technology, and capable of storing one terabyte of data on a single-layer 200mm2 chip. That's …
Hmmm, doesn't that sort of storage density torpedo a boat-load of "Ze CLOUD" business plans?
I mean I've just read into the tech a bit and you seem to be looking at ten-fold plus increases in typical applications. A Nexus 7 with 512GB of storage probably doesn't need a hefty dropbox account too.
Maybe time to short SNDK....
"Hmmm, doesn't that sort of storage density torpedo a boat-load of "Ze CLOUD" business plans?"
Not really. Anything we can use locally, the cloud can use, too, only at larger capacities.
If this mob is for real, say bye byes to the optical media industry.
... but it only makes business sense if you, the user, have the bandwidth to read/write the "clouded" data. Increasing storage medium capacities 10x or more will do nothing for your broadband or phone connection. If, on the other hand, your post-stamp-sized local storage is measured in terabytes, why would you bother with the cloud in the first place?
As for the cloud providers, more capacity means they can service more customers, and with the additional revenues get additional backhaul to handle the extra load. It would look no different on the client end.
There are a lot of pros there, I wonder what the cons are?
Personally I'd wonder about heat generation and dissipation, especially if they are talking 3D stacking.
I'm wondering too.
For a start, I'm unable to find anything about write latency. However read latency provided here is <30ns , so write cannot be too bad either?
If Crossbar started development only in 2010, how can they determine that retention would stretch to 20 years? Is there some kind of stressing that emulates 20 years of wear?
Yes, it's usually calculated by over-stressing the sample parts beyond recommended spec and then using some *creative* math to extrapolate the failure time based on the accelerated failure of the over-stressed component.
Overstressing can be as simple as a temperature increase, which has been shown to be exponentially related to mean time to failure.
Memory-making start-up Crossbar came out of stealth mode on Monday, announcing that it has developed what it characterizes as "very high capacity and high-performance non-volatile memory" based on a new approach to resistive RAM (RRAM) technology, and capable of storing one terabyte of data on a single-layer 200mm2 chip.
The company claims that due to the tech's "simple" three-layer structure, it can be stacked in multiple layers resulting in multiple terabytes of storage space in a single chip "the size of a postage stamp."
My god their postage stamps are HUGE!
To: Stuart Longland
Maybe your math is a bit rusty.
200mm² = 2cm² which is a quite normal sized postage stamp. Besides, that's just single-layer. If implemented as a triple-layer, the total size/capacity ratio would decrease three-fold.
sqrt of 200mm^2 = 14mm, seems pretty reasonable to me.
But how much will it cost?
When you add the interface and packaging? Sounds like SATA is going to be far too slow here.
When will this mythical 1TB device to become available for us consumers?
However, how many times do we hear about these 'super' products from startups only for them to be taken over by a competitior and the product/technology (apart from any patents) quietly burried forever?
Re: But how much will it cost?
It looks like they plan for this to be implemented on top of existing silicon.
So your average ARM SoC (around 80mm^2) could have 256GB of this magic-storage implemented on top of the silicon, with a direct link to an on-SoC magic-storage-controller - no SATA required. Silicon polishing between logic and magic-memory deposition might be required though. It might add $5 to $10 to the manufacturing cost of such a die, but who knows how much that will translate to when that's all integrated into a device.
Availability - I'll be surprised if we see if within 5 years, unless they can find a partner to accelerate development of the real world implementations. I hope that they can surprise us though, but right now I'm sceptical.
It will cost what the market will bear
PCIe flash is about £2000/TB now, so I expect in about five years, you could buy RRAM on a DIMM for £250/TB. The next product to expect is a combined SDRAM/RRAM chip that can go on a DIMM, SO-DIMM, in a mobile phone or (if there is a reason for having one) an iWatch. RRAM on a low power embedded CPU will be easier than on a hot server chip - the life time was probably measured by testing the chip at a range of high temperatures and extrapolating the graph down to sensible temperatures and long time scales.
If existing flash foundries can make RRAM, then I expect my entire DVD collection will be on a micro SDHC card in a pumpkin pi by 2020.
Re: But how much will it cost?
SATA probably will be far too slow. Expect devices direct-coupled to the board, on the address bus, perhaps with some cache RAM.
Re: But how much will it cost?
so much guessing, so little concrete information ...
Yes, how much does it cost?
That's what I want to know!
There are already several NVRAM technologies on the market that I could buy off-the-shelf right now, but the prices of everything other than MLC-NAND Flash are just way too high for anything other than specialist, high-value products. Military and aerospace, basically.
How ridiculously fast?
Ludicrous Speed !!!
Re: How ridiculously fast?
They've gone plaid!
Sounds all good but its V 0.9 at best
And the price is a big one.
Thumbs up so far. It really does sound too good to be true.
And historically we know what that means......
Re: Sounds all good but its V 0.9 at best
>It really does sound too good to be true. And historically we know what that means......
I don't know what that means.... I'm surrounded by bits of technology that once sounded too good to be true! (But yeah, point taken, pinch of salt on standby)
Apple (and a few others) need a new business model, and soon.
If they can't charge an extra £200 to throw in £30 more flash memory, what will they screw us on next?
Oh look, another memory manufacturer purporting to be just about to change the world, real soon now. Do they make batteries too?
Okay, maybe that was a little petty, but there are so many empty promises like this one floating around, it's hard not to become a little jaded. In short, I'll believe it when they actually start shipping product.
Re: Sounds Familiar...
Whilst fair enough, I suppose, one cannot doubt that the current state of affairs is an open goal waiting to be shot at, and I like reading about people who think they have the ball placed on the ground in front of the net and are ready to kick.
It's a matter of time before someone works out how to make non-volatile memory really fast, cheap and reliable, and that tech is likely to disrupt things very seriously, to my personal advantage!
So good luck little startup!
Re: Sounds Familiar...
Wasn't petty, because this hyped up press release immediately brought "power breakthroughs" (a la uber batteries or micro fuel cells that will power an Alienware latop for a week at full load blablabla) to mind.
Wake me when I can buy one for <$500/TB and I'll jump on the early adopter bandwagon.
We've been waiting for this for sooo long!
Here's hoping that the inventors don't drag the market by charging larcenous prices, and instead price it to become the dominant persistent memory model so fast that it makes flash obvious as the outmoded, clumsy, unrealiable technology it already is.
How about a gig of L1?
Another tech site has this news and refered to the 1tb chip as being 200mm x 200mm. 200mm2 is a bit more impressive :-)
Not to mention more realistic. What kind of system can take an integrated circuit 20cm to a side?
I'm suspecting 200mm^2 is closer to the mark (and more believable, a little smaller than the size of a full-size SD card), though I could still be wrong.
Who was it that said every chip costs $1 apiece eventually
Who was it that said every chip costs $1 apiece eventually[?]
HW Lay, I believe.
Its a game changer...
200mm wafer, if it can be manufactured cheaply, then you have a disruptive technology.
Even if they put it inside a 2.5" form factor w SATA interface, you're talking about 16+ TB in a drive.
Now your existing hardware could benefit from this.
Put it on a PCI-e card and you have again lots of storage.
And it gets even more interesting...
Assuming you could create a WORM device... (write once, read many), you could create a very small package capable of holding an entire movie + features, and stuff all on an incredibly small form factor.
Assume you use something like a thunderbolt interface.
You could then create a video player where you 'slot' the movie, or plug it in to your current Mac product.
(Intel is supporting this interface...)
So if you have a TV or Home AV Theater package that doesn't offer Thunderbolt, you just have to create a very small and inexpensive player...
Storage winner, now just need faster and larger bandwidth pipes... ;-)
- Mounties always get their man: Heartbleed 'hacker', 19, CUFFED
- Analysis Oh no, Joe: WinPhone users already griping over 8.1 mega-update
- Leaked pics show EMBIGGENED iPhone 6 screen
- AMD demos 'Berlin' Opteron, world's first heterogeneous system architecture server chip
- OK, we get the message, Microsoft: Windows Defender splats 1000s of WinXP, Server 2k3 PCs