IBM Scientists Achieve Storage Memory Breakthrough

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
IBM's new optical storage is roughly fifty times faster than flash. According to the press release, scientists at IBM Research have demonstrated reliably storing 3 bits of data per cell and retaining the information at elevated temperatures using a relatively new memory technology, known as phase-change memory.


The current memory landscape spans from venerable DRAM to hard disk drives to ubiquitous flash. But in the last several years PCM has attracted the industry's attention as a potential universal memory technology based on its combination of read/write speed, endurance, non-volatility and density. For example, PCM doesn't lose data when powered off, unlike DRAM, and the technology can endure at least 10 million write cycles, compared to an average flash USB stick, which tops out at 3,000 write cycles.
 
It comes down to price and power requirements.
Unless it's cost competitive, people will stick with flash.
Unless they can shrink this down to the size and power of a laptop drive or USB thumb drive, it's not going to get much market share.
 
There's no mention of "optical" storage in the article, you should fix that Steve!

Phase Change ram has been around for over a decade, but they always have the problem of the cell losing it's data when the temperature changes. The potential density and speed have always been better than Flash, so this changes nothing.

If you can't hold your data after you go through reflow ovens, then your tech is dead in the water :D

Also, because this is temperature-driven there's a limit to how fast you can write cells while not affecting the adjacent cells, which makes write performance no better than Flash.

That's why memristor is the first Flash replacement to actually have legs!
 
Last edited:
Doubt we will see this in the consumer market within the next 6 years, maybe longer. I'm quite happy with flash, but if it proves affordable it will be the next quantum step in storage.
 
Did he really say bit error rates of 1 in 10^-4??

For raw rate, yes. But that goes though correction schemes that greatly decrease the effective bit error rate. We routinely rely on things that have an inherent raw MTTF of 100-1000 seconds and stretch that out to 100+ year. The specification for BER on things like QPI/PCIe is ~10^-12 or so and using simple coding we stretch them to a MTBF of 100-1000 years or more (10^8-10^9 improvement). And that with rather simple 8b CRC/ECC coding. Using multiple BCH codes you can do a whole lot better than that. Most HDs are in the range of 10^-15 for BER. SSD's are in the 10^-17 BER range. Adding 10^13 reliability with complex BCH codes isn't unreasonable. Esp for technologies where you can spend 10+ nS doing detection correction (the error codes used for deep space communication are absolutely insane! but they can take uS to mS to do correction from BER in the 10^-1 or 10^-2 range)
 
Last edited:
Thanks for the explanation. I actually thought that this was the reason (and the way it could actually work in a product) but wanted to someone to explain..
 
Who cares we have intels Xpoint coming out later this year. If that doesn't blow everything else out of the water then it'll be a failure.
 
Who cares we have intels Xpoint coming out later this year. If that doesn't blow everything else out of the water then it'll be a failure.

Right, Xpoint = memristor with some proprietary improvements. The gate structure is nearly identical to memristor diagrams you can find, and just layered in 3D.

Just that Intel figured out how to mass-produce them way before anyone else (or at least their marketing department decided they did :D)
 
Last edited:
Back
Top