In June 2006 msystems’ CEO Dov Moran proudly announced x4, the first and only viable 4 bit per cell flash storage solution, which doubled the storage capacity of current MLC solutions at roughly the same cost.
The announcement was made at msystem’s first, and as it has turned out, only analyst day. x4 appeared to be the technological breakthrough that would put msystems over the top, securing chip supply at best prices from fab partners willing to trade margins and capacity for access to the technology.
x4 was the centerpiece of that analyst day. msystems had been working on x4 for 5 years in secret, but on that day was in full sales mode, even allowing its chief scientist of its Future Technologies Group, Simon Litsyn, to talk about the technology at length. His presentation was webcast live, but never recorded.
He revealed that one of the reasons that other companies had failed in implementing 4 bits/cell NAND is that some of the accepted theoretical Gaussian models for the behavior of NAND memory are not accurate. msystems figured out more accurate theoretical modeling, which in turn led to better design and — success.
x4 is a system solution. That means that there is a special x4 chip, a special very complex x4 controller, and special x4 software. All are essential to the system’s functioning. All were presented as breakthroughs. Suspect the controller/software is the most impressive. The x4 controller has over a million gates, but will only cost “tens of cents”, which I interpret as somewhere between $0.30 and $0.80, depending on volume, etc.
x4 looked like the real deal. Dov confidently noted that he had working x4 chip samples and that x4 would be in mass production in 2007. He also added that msystems was in x4 licensing talks with most of the big fabs.
Although names weren’t named, Dov did say some of the discussions were with fabs that msystems had relations with at the time and some discussions were with fabs that would be new partners for msystems. Reading between the lines, msystems was likely in discussions with Toshiba, Hynix, and newly with IMFT. Samsung looked to be the odd man out.
Then on 30 July 2006, msystems, along with x4, was acquired by Sandisk and info on the breakthrough technology suddenly dried up. When asked, Sandisk would only reluctantly reveal a bare minimum. So, what’s happened to x4?
Likely, just a change of strategy by its new owner.
Personally, don’t think lack of public commentary should be interpreted as lack of progress. Now that it has acquired x4, Sandisk surely has a plan, one that differs from msystems’ original timetable. According to comments made at Sandisk’s analyst day February 2007, it now appears that x4 will be a late 2008 story with significant impact in 2009.
msystems had been a fabless company and with its phenomenal growth needed all the leverage it could get to secure chip supply. Sandisk, on the other hand, has its own fabs and is in the enviable position of possibly already having the lowest costs in the industry. It doesn’t need to offer x4 technology in exchange for production capacity anytime soon. Sandisk can afford to be patient and bring x4 into commercial production at the time of its choosing.
Late 2008 seems optimal, as below 30 to 40 nm, it’s unclear how much smaller it’s industry competitor’s can make viable NAND chips. Sandisk’s strategy appears to be roll out the competitive edge of x4 bit scaling right when the competition is bogged down with the physical limitations/challenges of traditional SLC and 2 bit MLC.
Sandisk has already lined up a supplemental source for x4. In March of this year, Hynix and Sandisk announced a patent cross licensing agreement and a joint venture for x4 technology.
Although no details of the Sandisk/Hynix licensing deal have been announced, my guess is that it will be similar to its Samsung licensing deal in that SNDK will get royalties and best prices on x4 chips from Hynix.
One of the beauties of x4 is that it can be produced with the most advanced NAND fabrication technology currently available, whether it is 56 nm, 45 nm, or below. In its analyst day this year Eli Harrari, Sandisk’s CEO, confirmed that the msystems’ working x4 samples from 2006 were already 70 nm, i.e., then state of the art.
Sandisk’s 2007 analyst day slides reveal that for Sandisk, x4 will kick off as a 2008 story @ 56 nm and x4 is going to arrive as a 64 Gb chip. The significant impact will likely be in 2009 @ 43 nm. By late 2009 x4 should be available @ 32 nm. Sometime in 2011, if all goes according to plan, x4 will transition to 2x nm.
My own suspicion is that if msystems had continued on its own, x4 would have arrived sooner, but at larger geometries. Given the bigger picture at SNDK, it is hard to fault SNDK’s strategy.
My guess is that x4 will eventually be important for SSDs (solid state disks) @ 45 nm. Sandisk has said the first x4 (56 nm) will be used primarily in audio/visual applications.
SSDs are a pretty good fit for x4 as “x4 NAND devices can be configured for two operating modes: SLC or x4 NAND. SLC mode is ideal for small files and firmware that are critical to device operation and require very high performance and endurance, such as the operating system… In addition, x4 NAND gives flash manufacturers the ability to tailor specifications for performance and endurance to meet the specific needs of designers, thereby strengthening its ‘just right’ value proposition.” [from msystems’ x4 NAND white paper]
As a system solution, the speed and reliability of x4 improves with capacity. SSDs seem poised to be the high capacity flash product of the next several years.
In the past, Dov has said that as x4 matures he expected it to be used in just about all applications: memory cards, USB flash drives, embedded flash data storage, MDOC, and SSDs. I see no reason to disagree.
Although x4 has dropped out of the headlines, it appears destined to be rewarding for the patient Sandisk investor. That x4 has a slow fuse doesn’t mean it won’t have a big bang.