There are some fantastic new non-volatile memory technologies in the lab and working their way toward the market, but they will need quite a bit of time to displace NAND. NAND is just too cheap and too far ahead of the contender to be displaced any time soon.
That was the good news in a talk at Hot Chips last week by Jim Handy, president of Objective Analysis, who follows the memory market. The not-so-good news? Don't expect SSDs to ever get much cheaper. They've broken the $1 per gigabyte barrier and that's about as low as it goes.
"They will always be ten times the cost of a hard drive. The thing that's standing in the way of consumer adoption is the way consumers buy PCs now. [They ask] what does it cost and how much storage does it have? As long as you try putting in a SSD in the PC, something has to give," he said.
There are a number of laptops with SSDs instead of hard drives, but they are usually premium laptops from high-end makers, like Apple. For the low-end PC, especially desktops, a hard disk drive is still standard issue.
There are a variety of next-generation memory technologies in the works. ReRAM promises to offer the speed of DRAM but the storage capabilities of NAND. It's slowly working its way out of the lab and might make it to market next year. Then there's PCM, MRAM and a few other fringe efforts that seek to do the same thing.
"[They all] need to get cheaper to make a major market for themselves. There's a little of the old chicken and egg thing that comes into play [getting cheaper vs. attaining mass market], but it's a matter of when NAND flash stops going through pricing reductions through its scaling limit," said Handy.
NAND flash will eventually reach a point where it can't be shrunk any more and the memory can't be made any more dense. At that point it won't get any cheaper. As it is, NAND is struggling to get below the 20nm manufacturing process. ReRAM supporters are claiming they will be able to get the memory down to just 2nm, but way out in the future. When ReRAM or any competitor can shrink down to a smaller manufacturing process than NAND, it will start to become cheaper and more viable in the market. But that's many, many years off.
Handy figures once NAND reaches its scale limit, the memory makers will find ways around that limitation with other fixes, like 3D stacking. "After 16nm there might be one more process node. Then [memory manufacturers] say they will have three nodes of 3D NAND, but then they will find a way of coaxing that a little further," he said.
Handy thinks we are entering into a period of NAND flash shortage. Prices have been stable and will be flat until 2015, which means SSD prices will not come down any lower than they are. Handy said it's a typical cycle.
"We had a period where no one was making money from 2011 to this year, so no one invested in capacity. Now demand caught up with capacity. So they are building capacity but that won't be fully ramped until 2015," he said.