MLC SSD Controllers

This week I was planning to write once again about the gPhone, but instead have decided to revisit the subject of MLC solid state drives (SSDs) and in particular MLC controllers for SSDs.

On 8 November, an interesting SanDisk  controller patent application showed up and I decided to take the time to look at it more closely. In addition, a friend forwarded me a copy of a Bear Stearns’ flash report which contains some interesting MLC SSD comments from both SanDisk and Intel. Excerpts from this report can be found below after my ramblings.

The SSD market is a key strategic market for SanDisk. It is projected to grow nicely in 2008, accelerate in 2009, and by 2010 or 2011, Sandisk thinks it will become the second largest user of flash memory after mobile handsets.

SanDisk believes that its flash systems expertise, and in particular MLC implementations as well as its extensive system IP, will become important differentiators in this market as it unfolds over the coming years.

The MLC SSD will be a milestone for SanDisk, shaving something on the order of 40% off the cost of SSDs.

According to Bear Stearns, SanDisk recently stated that it would deliver MLC-based SSDs for notebooks in 1H08. Although I might be reading too much into the language, it seems like product rollout is creeping up. Not long ago it was sometime in 2008, then in the Q3 conference call it was midyear. Now it is sometime in 1H 2008.

Am now guessing we’ll get an announcement at least in February, around the time of the 2008 SanDisk analyst day. Personally believe that SanDisk has the goods on MLC SSDs. If so, the time to flaunt it would be on analyst day. Time will tell.

In any case, the SanDisk high-performance flash memory data transfer controller patent looks pretty sweet. Basically SanDisk appears to have figured out a way to achieve high performance data transfer together with low power consumption for flash memory including MLC, both 2 bit and 3 bit. Examples given show close to twice the data transfer speed of conventional controllers, using approximately one half the power. Reading between the lines, it appears that SanDisk has the means to cost-effectively approximate SLC performance with MLC.

Two applications where this controller would be particularly well suited are stated to be high performance digital still cameras and — SSDs.

Bear Stearns also reports that SanDisk’s system level patent IP litigation could result in agreements with significant IP revenue. Bear Stearns estimates that if all the players included in the suits were to obtain a license from SanDisk for all the products named in the lawsuits, SanDisk’s annual royalty revenues could increase approximately $60-$70M and this is before SSDs.

I suspect that one of the reasons behind SanDisk’s new push in system level licenses is to set the stage for SSDs where SanDisk system level IP could be critical regardless of another company’s controller strategy.

I’m inclined to think that SanDisk’s IP approach to SSDs will be two pronged. A general system level license for some and partnerships for a select few. Partners will pay more, but they will get the real deal for their dollars.

The select few in line for partnerships are likely those unnamed companies in the previously discussed Solid State Storage Solution LLC.

As noted, Intel would seem to be a candidate. Intel is a NAND producer targeting SSDs, and probably trustworthy as a potential partner in SanDisk’s eyes.

If Intel is in league with SanDisk for SSDs, suspect we should know before the middle of next year. Bear Stearns is reporting that Intel will be releasing its own MLC SSD around then. Maybe Intel has the MLC controller and system expertise in-house to pull such a product off on their own. On the other hand, maybe they have a friend in SanDisk.

Below is slide 24 “Flash Competitors & Alliances” from Eli’s 8 August presentation at the Flash Memory Summit. Note how the Intel bubble overlaps onto the Toshiba/ Sandisk bubble. No explanation was given in the presentation and no questions were asked. Curious though, isn’t it?

**** Bear Stearns excerpts below ****

Bear Stearns
November 8, 2007

Gurinder Kalra
Andrew Liersch
Dinesh Moorjani

Key Takeaways From The Bear Stearns Flash Tour

SanDisk Corp $40.67 O
O=Outperform; P=Peer Perform; U=Underperform
Securities in this report priced as of:
November 07, 2007 16:00ET

• We hosted meetings with SPSN, MU, INTC, SNDK and SIMO, and a conference call with Samsung, as part of our annual flash bus tour.

• While several of the companies were cautious about near-term NAND supply-demand due to increased supply from process transitions across the industry, SanDisk reiterated its optimistic outlook for 2008 as a whole. We continue to believe overall 2008 industry dynamics will be better than 2007, and currently expect demand to exceed supply in 2008.

• SSDs as a new major demand driver were a common theme in the meetings. SanDisk and Intel expect to introduce MLC-based SSDs for notebooks in 1H08 and 2Q08/3Q08 respectively, which we believe should serve as a catalyst for increased adoption of SSDs in notebooks as we move through 2008 and into 2009. Intel is also seeing strong interest for SSDs from the enterprise market. In the nearer term, SSD demand could increase from applications such as UMPCs and CE devices.

• Continued demand growth from handsets was also highlighted by companies as a key driver for NAND demand in 2008. Not only are the number of handsets with slots increasing, card adoption rates have also been increasing rapidly, driven by increasing multimedia capabilities in phones.

• As highlighted yesterday, Samsung provided incremental color regarding their 2008 memory capex. While spending plans have not been finalized, the company currently expects to spend <W6.0T in 2008 on memory vs W6.2T in 2007 (excluding Austin). We expect capex for the Austin facility to also decline significantly. Samsung’s comments are positive for the memory industry as its spending plans demonstrate a balance between protecting market share and impacting industry dynamics.

Overall, our meetings do not make us change our view on SNDK, MU or Samsung. While we expect the NAND market to exhibit typical seasonality in early 2008, we maintain our view that the pricing environment for the year as a whole will be better than in 2007.

Key takeaways from company meetings

Samsung Electronics (005930.KS-KRW570,000, Outperfom)

• Samsung provided incremental color regarding their 2008 memory capex. While spending plans have not been finalized, the company currently expects to spend less than W6.0T in 2008 on memory (excluding Austin). This is less than their revised memory capex for 2007 of W6.2T (excluding the Austin facility which has cap-ex of W1.0T this year). We also expect spending in the Austin facility to decline significantly next year. We see this as a positive for the memory industry as Samsung’s spending plans demonstrate a balance between protecting market share and impacting industry dynamics. The company’s stance was clearly less aggressive compared to its last earnings call, where management emphasized strengthening their market position as an important factor for their spending plans.

• Looking ahead to 2008, Samsung believes NAND industry dynamics should be significantly better than 2007 driven by continued growth in NAND demand from handsets and demand from new PMP models and SSDs. The company does not expect a sharp oversupply in 1H08 as we saw in 1H07. Samsung reiterated their 2008 production bit growth guidance of 120% YoY, and their expectation for the industry to also grow around 120% YoY.

• On the DRAM side, as expected, Samsung’s tone was relatively more pessimistic, and their comments suggested that they do not expect a recovery in DRAM market conditions until 2H08. While Samsung expects their DRAM business to continue to be profitable in 4Q07 due to a higher mix of specialty DRAM, the company suggested that continued losses at its competitors should result in significant cutbacks in DRAM capex next year, which should help industry dynamics in 2H08.

Intel (INTC-$26.90, Outperform)

• Our meeting with Intel was focused on their NAND business. At a high level, Intel’s strategy in NAND is to deliver solutions for their computing platforms, specifically caching solutions and SSDs, by taking the flash components from the IMFT joint venture and adding to it its system expertise. The company also plans to waterfall its NAND solutions to the embedded storage market such as in CE devices.

• Intel (through the joint venture with Micron) has started the ramp of 50nm-based NAND flash, and plans to convert production from 72nm to 50nm process technology over the next few quarters.

• Intel plans to launch SSDs based on the SATA 3.0 Gb/s interface in 2008, for storage in notebooks and for the enterprise server/storage market. Intel currently offers SSDs but with a USB interface. In the notebook space, the company plans to launch MLC-based SSDs in 2Q08 or 3Q08. In the enterprise server/storage space, the company emphasized that despite the higher cost of SSDs versus HDDs, other components of the total cost of ownership are lower, as SSDs require less maintenance costs, and need less space and cooling, and noted that customers have been very keen to use SSDs in the enterprise server/storage space.

• Separately, Intel plans to launch small form factor SSDs with a PATA interface next year, targeted toward the UMPC/MID market This is a package-on-package solution that stacks NAND with the PATA controller to reduce space and weight. It will be included on the motherboard in their Menlow platform, which will use their Silverthorne microprocessor.

SanDisk (SNDK-$40.67, Outperform)

• SanDisk continues to expect favorable NAND supply-demand dynamics in 2008. The company still anticipates the need to increase non-captive production in 2008 to meet increased demand. Near-term demand also remains strong and management commented that they wished they currently had more product to sell. While we believe 2007 will be oversupplied, we think 2008 industry dynamics will be better than 2007, and currently estimate that demand will outstrip supply in 2008.

• The presentation largely focused on the mobile space as SanDisk expects mobile cards to be a key driver in 2008. Management estimates that about 400M phones will have card slots in 2008 and slot penetration will grow to ~900M in 2010. SanDisk is seeing a greater attach rate (cards to phones with card slots) in 2007, estimating that about 30% of phones with card slots have cards. The company highlighted the move to feature rich phones – like the Nokia N95 and Sony Ericsson K850 which both include 5 mega pixel cameras and DVD quality video – as drivers for greater card adoption.

• SanDisk also briefly commented on its license and royalty revenue stream, suggesting that its system level patent IP litigation could result in agreements that are not insignificant to revenue. While SanDisk did not provide any financial information regarding the suits, we estimate that if all the players included in the suits obtain a license from SanDisk for all the products named in the lawsuits, it could result in approximately $60-$70M in annual royalty revenue. As a point of reference, we currently estimate royalties of $532M in 2008 for SanDisk, which implies about $1.42 in EPS.

Aside from demand drivers like handsets, and SSDs the company also commented on the camera market, suggesting that there has been a recent resurgence in demand for high-end DSLR cameras.


6 Responses to MLC SSD Controllers

  1. Poofypuppy says:

    Thanks, Savo, great post (as usual). The MLC SSD controller could be the IP capstone that SanDisk needs to establish dominance in the SSD market, we’ll know in time. Regarding the “Competitors and Alliances” bubble chart listed above, it’s interesting that Hynix (with ST) is depicted by itself on the lower left; I had thought the X4 collaboration and supply agreement between SanDisk and Hynix would make the Hynix bubble slightly overlap SanDisk (as with Intel and Toshiba). Makes me wonder why it was drawn that way. Hynix appears to be the distant step-child in this diagram, is this meant to imply that Hynix is out there on its own (with ST) or is this because Eli doesn’t want to show Hynix as an ace up his sleeve?

    Below is a link regarding SanDisk-Toshiba discussions for Fab 5.


  2. savolainen says:

    Greeting poofy,

    As you say, the MLC SSD controller is a big deal and clearly SanDisk understands the importance of SSD IP. As Eli said in the Q3 cc, he expects SanDisk MLC SSDs and SanDisk system IP to be important differentiators in the SSD market.

    When I posted last night I mistakenly referred to the patent as granted when in reality it is only an application. In any case, I wouldn’t be at all surprised if the approach described in the patent application is the approach SanDisk will be using when SanDisk’s MLC SSDs roll out next year.

    The high speed controller patent application was filed in April 2006, so SanDisk has had a chance to have been working on this for some time. If they have had the technology up and running for some time and can prove it, this would go a long way towards establishing their ownership with or without a granted patent.

    A granted patent would be nice though. At least this one is filed and the clock is ticking.


    Regarding the “Competitors and Alliances” bubble chart, I agree that it would seem to make more sense for the Hynix bubble to slightly overlap on SanDisk. I really have no idea why it is shown the way it is especially when Intel overlaps.

    The relationship between SanDisk and Hynix is likely a work in progress. Toshiba is no doubt sensitive. In the Q3 cc Eli wouldn’t acknowledge a formal agreement with Hynix. He referred only to ongoing discussions and a “ a potential minority investment in one of their (Hynix) 300mm Flash fabs.” He added that “ If that should transpire, the Hynix investment would be on a much more modest scale than a potential Fab 5 with Toshiba.”

    There have been hints that SanDisk will be buying at least a portion of the non-captive flash needed by SanDisk for Q4 from Hynix. I would have thought the non-captive would be coming from Samsung as it has (I believe) in the past.

    I wouldn’t be at all surprised if behind the scenes, SanDisk isn’t pressuring Toshiba to pay royalties for x4 (and for MLC for that matter). One way to do this would be to play the Hynix card. This is probably small stuff though compared to r/w Matrix 3D technology. It will be very interesting to hear who will be SanDisk’s partner for that, if and when the technology hurdles are overcome.

    The obvious choice would seem to be Toshiba, especially in light of fab 5. On the other hand, if Matrix 3D r/w were to live up to its potential as the next nonvolatile memory technology beyond NAND, maybe it would makes sense for SanDisk to have multiple partners.

    One fab 5 item I had wanted to weave into this MLC SSD post and didn’t get to: Eli said that there was no way that SanDisk could be a “significant player in the solid state disk business in 2010 and 2011” without fab 5.


  3. jim dunn says:

    SNDK shares have priced in everything going wrong into Q1.

    I think someone will bid for SNDK in the next few weeks. At these prices and with the condition of this industry its the time for INTC to acquire SNDK or LBOs to buyout SNDK. There’s also sammy and tosiba who could buy them.

    SNDK will not last long at these prices……the IP is worth 42-45 per share away from the products group.

  4. afkkl8_99 says:

    Hi Savo,

    Read an opinion on the YMB saying that any SSD will require use of the MSYS TFFS SWR, hence will have to pay L&R fees to SNDK who now owns the IP.

    1. Is that technically true that any disk emulation requires the use of TFFS software?
    2. Is the TFFS software public domain or SNDK exclusive proprietary ?
    3. If 1& 2 true, is it viable that SSD developers will to develope their own emulation software?

    Appreciate your comments

  5. Sam says:

    Any thoughts on the new deal with Toshiba? Before, they appeared to waffling on the timing for Fab 5. Now they’re talking about Fabs 5 and 6! Needless to say, a lot of money is involved–$6.5b for each, if I read it correctly.

    The only thing I can think of is they think that SSDs are going to take in a big way in 2009, when the new fabs are slated to be operational. Even so, it seems like a huge amount of capacity, especially if x3 and x4 will be function.

  6. Investor says:

    Any thoughts on STEC’s controller technology? Those guy recently announced a nice win with EMC and they keep saying that their SSD’s offre the lowest price/performance ratios in the industry. They also claim to be qualifying with notebook OEMs and to be in a position to ship in 2H 08. Thanks.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: