Ahhh. Let's see whose memory is short and volatile. Mr. Doller is on record claiming that his 45nm (1Gbit) PCM chip would be available by mid 2009:
It is February 2010 now, and the 1Gbit PCM is not in volume production. Not at all. Can Mr. Doller explain that failure to deliver a commercial 45nm PCM product? The only "commercial" PCM chip is the 128Mbit 90nm disaster formerly known as Alverstone (now Omneo P8P?) which writes at less than 1 megabyte per second, that is, 2x slower than Numonyx's own NOR and up to 15x slower than NAND. So, no, nobody in their right mind would use PCM in SSDs, even when costs per gigabyte are ignored. Those costs, by the way, appear to be about $200 per gigabyte for the Omneo P8P, or 100x as expensive as NAND, and about $60 per gigabyte for the "upcoming" 45nm part, 30x as expensive as NAND, assuming that 45nm part ever gets delivered in volume, based on a chart from a recent Numonyx presentation.
Just watch this webcast and calculate the horrible write speed yourself:
In the mean time, Fusion-io has been delivering for months a reliable NAND-based non-volatile "SSD" that writes at 1.4 gigabytes per second and reads at 1.5 gigabytes per second. And, yes, it is perfect for the enterprise. Unlike the PCM, which has been promised since 1970s, but never arrived.
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...