I agree that as long as the percentage of 300mm output (using the latest technolog) is low, the benefits will remain low.
I will add one more thing- Once the Spansion-Saifun merger is consummated in 1Q08, R&D expenses will go up disproportionately in the near term, without a matching rise in revs to leverage the higher OE's. So some of the cost savings from 300mm output will be pushed-out.
But Spansion has couple of other things going for them vs. the rest of NOR field in addition to 300mm. These are (in order of importance):
- Mirrorbit has advantage over Intel/STM floating gate MLC that amounts to approximately 1/2 of process node. So 200mm 65nm will be somewhat ahead of 200mm, 65nm Numonyx
Spansion's best chances are in NOR and that too against Numonyx. I doubt they are planning to move to 300mm wafers. Samsung is one company that can ruin Spansion's party though. NOR is still a 8B dollar market and they may see an opportunity, now that STM and Intel have lost interest in NOR.
- 4 bits per cell will be ramping in the near future (sampling now for some applications): investor.spansion.com;
From what I understand there are endurance and performance, especially Write performance issues with it. They maybe ok for applications that do not require frequent Writes, but just require frequent Reads.
- Eclipse - delivering both data (NAND domain) and code (NOR domain) on a single die. The target market are mainstream phones of the near future that will have functionality similar to high end phones of today and yesterday (at lower price).
-Ability to add logic on Mirrorbit die can enable some cost savings (Built in self test) but also open new markets as controllers or microprocessors can be placed on flash die.
Eclipse will have tough competition from Samsung's OneNAND (SoC) and Sandisk's mDoC (MCP) solutions consisting of NAND and Logic+LPDRAM from Qimonda. Realistically, Spansion has no chance here beyond some low-end design wins. I am not that familiar with Eclipse's specs but OneNAND's performance specs are extremely good. OneNAND and mDoC both have their pros and cons and I have written about it on the Sandisk thread.
IMO, Spansion is on a trajectory that the rest of the NOR field cannot match, which gives me some confidence in that regard.
They clearly have advantages vs. Numonyx and other NOR players but NOR market will decline or stagnate. There are just too many legacy designs out there which is extending NOR's lifespan.
But you brought up some good points about competition vs. NAND, BTW, something that I brought up as well some time ago, which is how much large scale helps in competitiveness.
Scale helps a lot! Let me be clear- Spansion's data storage products will not be able to compete with NAND on a cost/bit and/or performance basis for the next 5 years!
My question is: Suppose through combination of all of the factors combined, die size per bit of storage becomes equivalent between large scale NAND players and Spansion. How much does the large scale help in the eventual cost per bit, when the competition is kind of lopsided, like 10:1. What I am talking about the worst case scenario, when all that matters is cost per bit.
Joe, do not waste your time doing this analysis. Spansion has absolutely no chance whatsoever even with Quad bit products because they are at least 2 nodes behind and lack financial resources and experience with immersion lithography. They are too far behind on the learning curve and they don't have much time with Cash burning. Toshiba/Sandisk are about to introduce 3 bits/cell NAND at 56nm which, from what I am hearing, will have good performance. Toshiba/Sandisk are also getting ready to move to 43nm very soon. Toshiba/Sandisk, Samsung are all familiar with and have experimented with CTF and they have poured Billions of dollars into floating gate and will move to CTF when the time is right.
I am not that familiar with cost/bit for ORNAND but 16Gb MLC NAND chips could sell for under $5 before the year is over!
If Spansion continues to execute on its roadmap, IMO, Spansion will be getting ever closer, or even surpass NAND players in die size per bit, but I am wondering how far away will Spansion continue to be in cost per bit due to manufacturing on much smaller scale...
Forget about surpassing, Spansion is way, way behind on the cost curve and that gap is about to widen. There is absolutely no competition between Spansion sampling 65nm Quadbits with 10-20kwpm with someone who has already completed migration to 56nm 2 bits/cell with 170kwpm 300mm wafers at mature yields and ready to begin 43nm production and expanding wafer capacity to 360kwpm 300mm wafers by mid-2009! |
