Perspectives From the Leading Edge

Evolution of 3D Market Applications

Over the past year we have watched CMOS image sensor (CIS) manufacturers follow the lead of Toshiba and one after the other announce new packaging designs that offered volumetric efficiency advancements  by incorporation of TSV. [ see “...on Mechanical Bulls, Rollercoasters and CIS with TSV”, PFTLE, 9/26/2008 ] The initial announcements of TSV based production ramps actually came last fall during a significant downturn in the CIS market, the first decline since the 2001 downturn. [ “Image Sensor Market Set to Rebound”, EE Times 4/11/2008].

I, like most other prognosticators, believe that memory will be the next application space to adopt 3D Integration. Some marketing firms initially proposed FLASH would come first,  but then retracted that prediction in favor of DRAM, when it was clear that others did not agree. Memory leader Samsung was showing prototypes of both TSV stacked FLASH and DRAM in 2006, but no products or announcements have come since then. The evolving recession that we will soon find ourselves in will most assuredly hammer the entire memory market over the next few years. This will offer a great excuse for why the inflated 2010 – 2015 market numbers, proposed by some in 2007, will not come true. In fact something tells me we are about to see some Olympic quality back peddling from those market gurus !

I have openly expressed my opinion that FLASH is so price sensitive, that there is just no way to introduce 3D TSV technology into FLASH  until a very low price point is reached, or at least  in sight. With FLASH manufacturing prices dropping yearly (see later discussions) I do see this happening, but not starting until 2012. My commercialization timeline (pre recession) is shown in the figure below and has not changed since I started writing these blogs. Whether this time line gets pushed out or stays firm really depends on whether TSV 3D drives the new product introductions that will eventually pull us out of this downturn and whether evolving markets like solid state drives (SSDs) which are using available technology for their first generation products try to use 3D TSV to get a performance / volumetric efficiency leg up on their competition. History tells us that it will only take 1 company announcement, like Toshiba in the case of CIS, to cause the others to move forward (remember – most are followers not leaders !). The performance and volumetric efficiency motivations for 3D integration are beyond question, yet we all know that this won’t happen until the economics are right.

The likely road forward for the  memory market does not look rosy. The worst enemy of the chip industry is an economic slowdown. Semiconductors are pervasive in discretionary consumer electronics (Cell phones, PCs ,cameras, MP3 players etc. ) and the automotive sector which make up a significant portion of total chip sales. Weakening consumer demand will mean oversupply and excess capacity.

According to iSupply  DRAM prices are falling  and they  don't expect a recovery till the second half of 2009. That's bad news for DRAM producers like Samsung, Elpida, Hynix, Micron and Qimonda many of which are already bleeding red ink. As bad as DRAM is FLASH is worse. Here are some recent “memorable” events (pun intended) :

In March Intel and  STMicroelectronics combined their money-losing businesses in memory and launched  Numonyx with a combined annual revenue of ~ $3 billion. 

In March 2008 Elpida entered into a partnership with UMC . The foundry venture will use Elpidas fabs in Japan.

In May Intel and Micron announced a joint venture to shrink their NAND technology from 52 to 34 nanometers.

In August of 2008 Hynix and Numonyx signed a 5 year deal to produce  NAND flash memory for high end  mobile phones , MP3 playuers and  SSDs (solid state drives) [we’ll get back to these later]  and enhance the mobile DRAM used in cell phones. Hynix had posted its third consecutive loss of $700m for Q2.

Last week Micron reported that it would l cut its global workforce by 15 percent over the next two years due to a massive slowdown in demand for memory. The oversupply and decrease in demand  has led to seven straight quarters of losses at Micron.

Also last week, Samsung “the big kahuna” reported that their memory business ( both DRAMs and FLASH), “...remains challenging” .

We have seen reports that Infineon is desperately trying to sell its stake in sinking Qimonda. It is reported that they  will most likely sell the DRAM operation to Micron or Samsung.

NAND FLASH for SSDs

While many of the traditional NAND markets are maturing, it is expected that the next NAND market driver will be solid-state drives (SSDs). It is likely that solid state drives ( SSDs) will be a large future market for NAND Flash but the question is when ?

This past April, Toshiba Semiconductor forecast that 25% of all notebook PCs will use SSD in 2011[ “1/4 of Notebook PCs to have SSD in 3 years, Toshiba Semiconductor says”, Nikkei Tech-On, 4/21/2008]. They further indicated that evolving technology would lower NAND FLASH memory manufacturing costs by 40-50% per year. They stated the current SSD pricing disadvantage vs HDDs would be gradually reduced from the current 2.9 and 6.4X that of 1.8- and 2.5-inch HDD respectively to 1.4 and 3.2X by the 2010-2011 timeframe. They showed a 256 Gb SSD drive and said they would be introducing a 512 Gb SSDs in the future.

In support of such predictions, IDC projections for 128 Gb SSD ASP and Web-feet research ( chartered by Toshiba) unit demand projections are shown below. They project ~ 200MM units / yr demand by 2013, 40% of which is expected to be SSD.

In September Intel introduced SSDs aimed at replacing hard disk drives in computers. The products feature controllers that Intel designed specifically for the PC architecture. There have been reoprts that by 2012 we could see 256 GB SSD for < $200. [ “Replacing hard-disk drives with solid state: A challenging but opportune prospect for vendors” EDN, 10/14/2008]

Not that there aren’t still issues with SSD. For instance there are reports that Samsung is working with Microsoft  to improve the performance of SSDs on the Windows operating system. [ see “Samsung and Microsoft in talks to speed up SSDs on Vista”, Computerworld.com, 8/7/2008 ] Windows is currently designed to work with data from rotating media, but by working with Microsoft, Samsung wants to distinguish SSDs from hard drives on the Windows OS (operating system). OS currently recognize HDD data in 512 byte sectors whereas SSDs need a 4 Kb sector size. It is assumed that such issues will be resolved quickly as SSDs gain more market share.

In the table below I have listed some of the big players and shown their overlapping activity in CIS and memory . It is obvious that development of TSV stacking and bonding technology, is relevant to both product lines.

So the question for 3D Integration is not whether the SSD market will happen but rather when one of the players will pull the trigger on 3D Integration for SSD, which has been shown to result in improved electrical performance and vastly improved volumetric efficiency for stacked FLASH. 

Stay tuned......we will be finding out together..................

Status of XSil ?

I’ve been a strong supported of the great laser research and development work that XSil has been doing / presenting in the past year or so. They have pretty much single handedly put lasers back on the map when it comes to technologies for TSV formation.

So, I was extremely disappointed when I recently saw (mid September) reports in Irish newspapers that that XSil had put all its staff (of 44 including Sr management) on protective notice after a sharp deterioration in business. The company’s staff were told by Brian Farrell, CEO that they “…have launched new products, but the uptake has been much slower than anticipated and the downfall [in the existing business] has been much steeper.” and that its business with one large customer had ‘‘…deteriorated much quicker than anticipated’’.  Newspaper reports indicated that that the firm’s revenues had collapsed and it was facing a liquidity crunch. If this ominous report is true, best of luck to all those involved. ……..

Coming up in PFTLE..................

....highlights of last months Sematech 3D Integration workshop “Manufacturing and Reliability Challenges for 3D Interconnects using Through-Silicon Vias” held  at the Advanced Metallization Conference

For all the latest on 3D IC Integration stay linked to Perspectives From the Leading Edge [PFTLE] ..........