Steve,
Thanks for your comments.
I think you need to divide all your market share projections by a factor of 10 (OUM taking 75% of the flash memory market inside 4 years ? hmmm), and increase the market size numbers by about 20% (because chip market size projetions are _always_ low).
By a factor of 10!!? Why would you say that? I think the Flash marketshare forecast is very realistic. Intel presently owns 25% of the market and their aggressive capacity expansions guarantee them to at least maintain that share. Because of its vast performance/operating advantages over Flash (speed, cycle endurance and true RAM), market share should expand rapidly, as long as the manufacturing process is not too incredibly difficult to adopt. Maybe push the Flash penetration rate back by 1/2 year to a year: We could see Intel convert half their capacity in the first year, and the rest in the second year. Meanwhile, some licensing to other Flash producers.
OTOH, if they are able to make OUM 2 bits/cell without sacrificing a lot in terms of cost and performance (i.e., operating temp. range), then OUM will be a stunning success in the Flash market. Intel's 25% share of industry capacity would become 50%, or more, since not all of their present capacity is multibit.
Regarding the embedded and DRAM markets, I would be willing to push the penetration rates back by a year, but I don't think I would have to decrease the actual numbers. A 10ns or faster nvRAM will have great performance advantages as system memory (in latency and bandwidth), not to mention the instant-on crowd-pleaser aspect. For on-die cache or in a multi-chip module with the cpu, fast/dense RAM will be very valuable. Then you have all the networking-type chips that are always looking for cheap embeddable RAM. This is a very large market. Again, if OUM achieves multibit per cell, it could be an awsome success in these areas.
Even without multibit, OUM has the smallest cell footprint of any memory technology out there now, save 2 bit/cell Flash. Here's a table of these footprints:
2 bits/cell cost-adjusted
DRAM 8
NOR Flash 12 6 7.8
NAND Flash 10 5 6.5
SRAM 90-160
OUM 6 3 ?
It costs the Flash makers 30% to do 2-bits/cell because yields suffer. I don't know if OUM would suffer the same penalty making multibit (or if it's even possible).
Cost is a HUGE factor in determining penetration rate of a new technology. Rambus' yields were horrible, making it several times as expensive as SDRAM, thus derailing its penetration plans. (It didn't help that the chipsets didn't work either).
OUM promises to be the opposite in this respect, by virtue of its monolithic thin-film design and very small footprint.
The major assumption in all of this is manufacturability: There could be a fineprint catch in all of this. I don't have much of a handle here (but I'm gonna keep on pryin').
wily |
