Pam, Would it be correct to conclude your perspective on Spansion softened a bit on the technical front?
Rink, let me summarize what I have said-
1. "Traditional" NOR market (code execution) which Spansion (50% of their revs) and others address has had anaemic growth over the last few years and the growth over the next few years will be no different. However, within this growth, Spansion can, and probably will, grow faster than its peers (similar to what CSCO was able to do when the networking market slowed. They grew at the expense of their peers 3COM, Olicom, Madge, Synoptics, etc. to maintain their own growth) because of its cost per bit advantage. One competitor they have to fear from is Samsung. They have plenty of resources and despite their cost disadvantage in NOR, they can give Spansion a tough time or even destroy them and the market completely.
2. Comparing NOR and NAND in some sense is useless. They are two different technologies meant for two different purposes. NORs biggest advantage is in applications that need boot capability with small amounts of storage. If applications require large amounts of storage, ORNAND is not competitive with NAND today, and it is my view that it will not be competitive over the next 5 years!
3. I have never said that Spansion has softened a bit on technical front. What I have said is that despite their 65nm ramp and upcoming 45nm introductions, they are and will remain behind NAND because others are also running as fast as they can (x3, x4, 3D, 3xnm in 2009, etc.) and there are a lot more dollars, significantly nore, at stake compared to what Spansion has on the line. Companies that are competing have a lot of semiconductor manufacturing expertise and experience, a lot more resources and a lot at stake which they would have never invested unless they could see the path reasonably clear over the next 5 years! Again, it is best not to compare Spansion with NAND players. Spansion should be compared with Numonyx, Samsung (NOR), Macronix, Sharp, SSTI, etc. in the NOR space to get a better picture. Although, Spansion is transforming themselves to address different markets, others by the same token are attacking their markets (OneNAND, FlexiNAND, etc).
Refering to 'the defacto standard' discussion (OneNAND + LBA-NAND), and cell size discussions we had earlier. Data from then + recently provided by Buggi might at least show that NROM used by Spansion has good potential to become a good standard (it's not far away from it rev share wise).
Strange things happen in the real world. There have been technologies (Betamax vs. VHS, etc.) which get side stepped for reasons that are hard to comprehend or explain. Just looking at the cell size and assuming cost competitiveness would be wrong because you have to look at a lot of other things such as volumes, nodes, yields, performance, etc. From what I have seen, roughly half of Spansion's revs are coming from cell phone makers and other half from other applications such as settop boxes, media players, etc. Samsung hasn't released much information, especially the dollar revs from OneNAND, etc. so I do not know how they do compare with revs for Spansion from similar products. I do believe that OneNAND is a superior SoC product based on performance specs but have no idea how it compares cost wise with Spansion's products. If Toshiba and Samsung both push this product, there is a reasonable chance that it will become a standard even though it is proprietary architecture. I believe even STM had licensed this technology at one point.
Buggi hasn't provided any new data. We need some real die size data on high density ORNAND manufactured on leading edge nodes by Spansion. Spansion will go into production with a 2nd generation 1Gb MirrorBit Quadbit product at 65nm shortly but we lack the die size data. In any case, this discussion about cost per bit is just academic because what really matters is at what price those bits are selling in the market and not how much it costs just based on cell size discussions. Since MirrorBit ORNAND has a 5f^2 feature size compared to 4f^2 feature size for FG NAND, the effective sell size is 1.25f^2 for ORNAND QuadBit vs. 2f^2 for 2 b/c MLC NAND and 1.33f^2 for a 3 b/c NAND and 1f^2 for 4 b/c! This is just one of the components to actual cost per bit. I do not have information about what happens to peripheral circuitry for these two different technologies which do not shrink at the same rate as the cell itself. There are a lot more things that will ultimately decide which is going to provide lowest cost bits than merely cell sizes! If one just looks at what is out today, QuadBit is not the lowest cost per bit and their performance is nowhere close to 2 b/c MLC's much superior performance. But does it matter? No one makes 1Gb NAND parts any more (well, actually some still do in limited quantities) so QuadBit still has a market in certain applications despite their higher cost/bit and relatively inferior performance. There are products where their performance is sufficient and it is providing an overall cheaper solution.
Would you perhaps agree with me now that:
- Spansion and Samsung(+Tosh) are going to have Numonyx for breakfest,
Have I said anything different from this?
- Spansion might hence become the last pure NOR vendor standing, and
The way I see it, it is trying to move away from being just a pure NOR vendor, but what's your point?
- after that Spansion can't exactly be counted out (there's too much short and long term technical innovation - quad/eclipse/ornand/double density/sonos/etc... going on to discount Spansion on first analysis over the long run).
I am not sure what you are trying to say here but I haven't counted Spansion out from anything.
-Pam |
