Japanese researchers showcase hybrid SSD using NAND and ReRAM

Shawn Knight

Posts: 15,306   +193
Staff member

The technology inside that new solid state drive you just purchased could be obsolete sooner than you think. A group of Japanese researchers have developed a hybrid SSD that uses high-capacity NAND flash memory alongside Resistive Random Access Memory, or ReRAM. The combination could allow future drives to write up to eleven times faster than today’s quickest SSDs.

The research was led by Professor Ken Takeuchi from Chuo University and presented at the 2012 Symposium of VLSI Circuits held in Hawaii. At the event, the group proposed a drive that included 256GB of NAND flash as well as 8 Gbits of ReRAM, the latter of which is used as cache and storage.

In addition to the increased speeds, researchers claim the hybrid drive will reduce power consumption by up to 93 percent. And if that isn’t enough to get you excited, these drives could last far longer than traditional SSDs – nearly seven times longer, to be exact. ReRAM, however, is currently expensive but the price versus longevity factor could reduce overall costs to just 1/7 of today’s drives.

Life of the drive is also extended by reducing the number of writes in the NAND flash. Instead, data is written in the ReRAM about 30 times more often than in NAND flash.

An algorithm called MRU (most recently used) is used to store frequently used data in the ReRAM. By using this method, common data can be read from the drive faster than from NAND flash.

Despite its promise, Neowin concludes that we aren’t likely to see this technology come to consumer-level SSDs anytime soon. The reason for this has to do with the price of ReRAM. If the researchers can significantly lower the overall price (none was mentioned) then things could become much more feasible.

Permalink to story.

 
Nice!!! :)

Up to 11 times 6Gb/s, am I doing this right? That would be over 60Gb/s.


Definitely need a new storage standard for this tech. With each SATA standard doubling in speed, we are looking SATA7. Why would we even bother doubling transfer rates for SATA4, SATA5, and SATA6? Even with SATA Express you would need 8 lanes to achieve 60Gb/s.
 
No, you're not doing it right.
Write performance is just 46Mb/s, which is 11x the speed of conventional SSD hard drives.
One thing is theoretical transfer speed and other thing is real write speed...
 
Write performance is just 46Mb/s, which is 11x the speed of conventional SSD hard drives.
I know I don't want any of your SSD's, 46 megabits or 46 megabytes is slow for an SSD. SATA 3.0 (6Gb/s) would translate closer to 600MB/s (megabytes). A standard HDD could transfer 100MB/s.
 
A damn good HDD could transfer 100 megabytes a second, most of them get only 70-80. But yeah, your math seems correct. Today's fastest SSDs have Maximum R/W speeds of let's call it, 550 megabytes p/s, that times 11 would be roughly 5.9 gigabytes a second, which is in fact 47.26 gigabits per second.....
 
The reason I wrote out megabytes, gigabytes and gigabits is because a lot of people get them all confused.

1 GB (Gigabyte) = 8 Gb (Gigabits). 1 GB = 1024 megabytes. If I'm not mistaken, GB are used when measuring size, and Gb are used when measuring bandwidth.
 
Ok sweet. No "Delete" button though, so no point in consolidating all my replies into 1 if I can't delete it anyway.
 
Someone said bigabites are used for size and bigabits are used for speed. Can you also use gigabytes per second or gigabits per second. Is ether still valid. Also the same is true for size you can have gigabits or gigabytes. hehehe confused yet.
 
Technically if you are using base 2 (e.g. 1024), then it's Gib, GiB... Gb and GB are base 10.
 
Now, didn't they make some similar claims about SSDs long before they arrived on the scene? I recall something about ten to twenty year lifespans. As we've discovered, that was bull. I kinda expect this to go the same way. If we're LUCKY we might start seeing five years warranties.

If this new tech will actually have a 20 year warranty, then I'll be first in line.
 
I wonder what primary school is this "research" from. "MRU" Algorithm, my God! This revolutionary approach was called "smartdrive" in DOS some 25 years ago, and now is an integral part of any operating system, it's called "disk caching".
 
Not quite.
6Gb/s is the potential transfer speed of the Data Bus, not the read/write speed of the drive. Most SSDs struggle to utilize the 3Gb/s offered by SATA II. 3Gb/s is GigaBits, not GigaBytes (though we may all wish it was bytes instead of bits). There are 8 Bits in a Byte, so 3Gb/s is actually a data transfer rate around 384 MegaBytes a second (going by pure theoretical limits and math - the rate in reality is usually lower due to system bottle necks). Up to 11x faster would be around 33 GigaBits a second. Again, by pure theoretical work, and that is just the speed the bus would be able to produce (bottle necks and all) in order to make full use of the drive speed - the actual Bus it would be one would probably be operating at a minimum of 48Gb/s (going by the observance that Moore's Law also applies to data bus speed growths, doubling each generation, with each generation being about two years apart)
So, still very fast, faster than most consumer computer can even process data. I'm predicting we're still 10 years off, and will probably have to see electrical data transfers (in-system) replaced with an optical system of some form.
 
Back