Toshiba's new 19nm flash

Toshiba has announced 19nm NAND flash. Does this mean that even smaller feature sizes are on tap?

The end is near! It's an annual ritual in storage: "the end is near!" The "end" being the end of shrinking feature sizes - which means the end of higher density, lower cost-per-bit and ever-growing demand.

For at least 10 years the disk drive folks invoked the paramagnetic limit as the reason the density ride would soon stop. The limit hasn't gone away, but engineers figured out new ways to work around it.

And the disk technology race continues with HAMR and bit-patterned media.

The flash limit A similar flash limit has been bandied about for years: ≈20nm feature size. In 2008 SanDisks CEO questioned the industry's ability to scale below 20nm:

NAND flash technology is expected to scale down to the 20-nm node, but there are some questions beyond that process . . . .

Toshiba's 19nm part So what are we to make of Toshiba's new 19nm 64Gb part announcement? They're shipping samples this month and expect volume production in Q3.

The Storage Bits take It's good news for us: smaller, cheaper and higher capacity flash when 16 die are stacked into a single chip package.

And it also means that competition is alive and well in the NAND flash market, even as it slows down in the disks. Less than 2 weeks ago Intel and Micron announced 20nm flash.

The limits that storage people talk about aren't just physics. There's the economic cost of achieving a new shrink.

30 years ago you could build a small but economic semiconductor plant for $50 million. But today's economic plants are in the multi-billion dollar range - and the price tag keeps rising with each new generation.

The costs of developing HAMR and bit-patterned media are one reason for disk consolidation. The same force is driving semiconductor consolidation as well.

Our ability to afford new technologies will fail before the ingenuity of our engineers and scientists does. Expect stepwise enhancements that get us to 10nm flash. Beyond that, nobody knows.

Comments welcome, of course.