This post is something of a follow-on to the last X-Files post.
It began as a response to comments on X-Files and then wandered on its own way.
On Thanksgiving, 27 November, a most interesting and relevant article was posted by Tech-On!, a Japanese site for engineers in the electronics/ machinery fields. This article, “SanDisk Eyes Bringing 3bit/cell NAND to SSDs”, is the most informative to date on SanDisk’s business strategy for SSDs. The article ends with a tantalizing hint of a mystery technology soon to be revealed.
This talk of mystery technologies reminded me of Solid State Storage Solutions, LLC (SSSS LLC). Decided to check up and four more patents have appeared.
By the way, for anyone reading this not all that familiar with popular USA culture of the last couple of decades, the “X-Files” was an American sci-fi TV series which first aired on September 10, 1993, and ended on May 19, 2002. A big deal in my house. Probably a bigger deal than psychologically healthy.
For x3, I anticipate a replay of the SLC vs MLC debate of 2004/2005. Samsung (as a secondary MLC player at the time) spent a lot of time and money campaigning against MLC as both slow and unreliable.
Shortly after Apple came calling for MLC in 2006, Samsung finally stopped beating that dead horse. Today x2 MLC is accepted as being just fine for the lion’s share of NAND applications.
Am expecting deja vu all over again for x3 MLC. x3 is going to be slower and less reliable than the better x2 MLC- certainly in these early days. Just how much of a drop-off will likely depend on the manufacturer.
That said, SanDisk seems to feel that its x3 does the job at least in some applications. SanDisk is shipping x3 in commercial products today. These x3 products are not being sold cheaper. The x3 is simply standing in where x2 once held forth.
And there are no complaints that I am aware of. SanDisk x3 is not considered a reduced capability part.
Which brings us back again to ABL. SanDisk has attributed its x3 success directly to ABL.
Apparently the internal chip layout is very efficient. The page buffers are along the side where the bonding pads are and then the row decoders are between the arrays. The row decoders, page buffers and all bonding pads are on one side and voila– an efficient floor plan. Or at least so they say.
While every square mm is precious, to my mind, the bigger ABL deal has to do with performance. ABL column logic is two-sided where conventional MLC is only one sided. In conventional MLC each sense amplifier has to deal with 2 bit lines. With ABL each bit line has its own sense amplifier and therefore the chip is a lot faster.
The example given in the 2008 ISSCC presentation is for 4 KB pages programmed in parallel. Conventional x2 MLC can program 2 in parallel while ABL full sequence can program 8.
For reference: Conventional x2 MLC clocks in around 10MB/S. x2 ABL MLC full sequence can achieve 34 MB/S. SanDisk/Toshiba x3 (56nm) achieves 8MB/S.
By 2012 x3 (3 bits/cell) is expected to take over as the industry workhorse accounting for over half the market at 53% (numbers from Forward Insights).
To hit 53%, x3 will have to have found a home in SSDs, the looming megamarket. 8 MB/S, in and of itself, hardly seems adequate.
SanDisk, at least, appears to have a plan, and maybe a solution.
It appears that SanDisk has a new unannounced technology waiting in the wings. Rumors are that this technology will enable “ 3- and 4-bit MLC flash acceptable in the performance stakes.”
This technology is expected to be announced January 7-8, next year at CES in Las Vegas.
Purportedly this mystery technology together with x3 ABL NAND and ExtremeFFS will enable viable SSDs using 3 bit/cell NAND.
Extreme Flash File System (ExtremeFFS)
The ExtremeFFS slide above is from SanDisk’s Don Barnetson’s press conference in Tokyo November 26, 2008. Suspect we’ll see a lot more of it.
ExtremeFFS was announced by SanDisk on 5 November of this year. It is advanced flash file system software for SSDs. Purportedly a solution to the SSD problem of random writes- a big, big problem.
ExtremeFFS is patented and SanDisk also owns the IP to its predecessor TrueFFS, developed by msystems in 1994.
As the ExtremeFFS slide above indicates, ExtremeFFS adds three important integrated functionalities to TrueFFS. Each tailor-made for SSDs. From Tech-On!’s coverage of SanDisk’s Don Barnetson’s press conference in Tokyo Nov 26, 2008 [points corresponding to slide]:
“First [Page based Data Allocation], the new system can manage data by page unit, which is a smaller data storage unit compared with the prior block unit.
Second [Fully non-Blocking Architecture], it can perform different actions in parallel via multiple channels. For example, while the user is writing data and doing “garbage collection” (liberating unused memory) via one channel, data can be read out via a different channel.
Third [Usage Based Content Localization], the new system can “learn” the user’s pattern of using data and localize data storage areas in accordance with factors including how often the data has been used.”
So how good might ExtremeFFS be?
SanDisk claims that ExtremeFFS has the potential to accelerate random write speeds up to 100 times over existing systems. Translating this into “net performance next year”, SanDisk feels we’ll see performance “four times faster than the current generation of SSDs, and nearly six times that of the latest 2.5” HDDs.”
SanDisk “has no plans to license the ExtremeFFS file system so far.” If SanDisk were to have a change of heart about licensing, the licensing vehicle might already be in place.
Solid State Storage Solutions, LLC (SSSS LLC)
SSSS LLC is about as obscure as obscure gets for those watching SanDisk closely. It may or may not amount to anything. It may be irrelevant already. On the other hand it may not.
All this recent talk about mystery technologies related to SSDs got me thinking that it might be time to check again on SSSS LLC. Nothing new in SEC filings. But 4 new USA patents have shown up since March: #7450457, #7447072, #7379379, and #7366016.
The slide above is from SanDisk’s 12.04.07 Nasdaq 20th Investor Program presentation. To my knowledge this is the one and only time SSSS LLC has been discussed publicly. At the time, Eli said:
“We have formed a partnership with another major owner of system level flash IP, [the partnership is] called Solid State Storage Solutions.”
Today, it seems pretty clear that SanDisk’s SSSS LLC partner is likely Toshiba. Intel, Hynix and Samsung can be eliminated for obvious reasons. Not much left after that.
Note the language on the slide: “Solid-State Storage Solutions licensing entity formed: captures third party flash systems IP portfolio.”
It also appears a pretty safe bet that the “third party flash systems IP portfolio” captured, comes through Renesas, from Hitachi.
6 patents have been assigned to “Solid State Storage Solutions LLC”. All this year. All are continuations of patents assigned originally to Hitachi and then to Renesas with one exception- U.S. Pat. No. 7450457- which was assigned to both.
Maybe the exact name, “SSSS LLC”, is just a coincidence. Given that the patents have to do with clever ways of managing flash memory, am inclined to think not.
The first U.S. patent assigned to SSSS LLC is No. 7324375, which has a publication date of 01/29/2008 and a filing date of 05/08/2006. It is a continuation of Pat. No. 7,286,397, assignee Renesas. This patent started life out as U.S. Pat. No. 5,889,698 filed way back in Jul. 9, 1997. Date of Patent 1999. Assignee Hitachi. Foreign Application Priority Data goes back to 1995.
The most recent patent assigned to SSSS LLC is U.S. Pat. No. 7450457 which was published 11/11/2008. I suspect there are more to come. No reason not to.
So what systems IP might SSSS LLC, the licensing entity, have, that others might want/need?
As SSSS LLC IP is a combination of specific SanDisk IP, specific IP of the unnamed partner (Toshiba?) and the specific captured third party IP (Renesas/Hitachi?), its all guesswork today.
My guess is that SSSS LLC is somehow related to SSDs. If so ExtremeFFS would be a prime candidate.
If, as I suspect, its focus is also MLC SSDs: x-files stuff: x2, x3 and maybe even x4, the unannounced technology waiting in the wings is also a prime candidate.
As the X-Files tag line goes, “I want to believe.”
That said, SanDisk still has to deliver.