Hynix develops triple-level-cell NAND flash: 30 percent cheaper than MLC
By Nilay Patel 
posted Jun 4th 2008 10:25AM
posted Jun 4th 2008 10:25AM
The more cells you can pack onto a NAND chip, the cheaper the storage, so we're excited about Hynix's recent announcement of triple-level-cell NAND flash. You might have noticed that MLC-based SSDs are cheaper than SLC units, and TLC keeps the trend alive -- Hynix says the cost of TLC production is some 30 percent lower than current chips. On top of that, the 32GB chips themselves are 30 percent smaller than current NAND chips, meaning device manufacturers will be able to cram more storage into our PMPs and phones. We'll have to wait a little while though: production is going to ramp up in October, and Hynix hasn't committed to using the new chips in SSDs quite yet.
This is perfect.
The "netbook" market is quickly being saturated and their small size means that regular hard drives are out of the picture. Now the SSD market is quickly being saturated to meet the demand of the netbooks - which allows for more R&D on SSD drives.
More competition will lead to lower prices and better/more reliable technology.
I forsee a 100 GB SSD around $99.99 within a year.
I don't forsee that for a couple years at least.
Yeah, seeing that traditional 2.5" 7200 rpm 100 GB drives can still run above the $100 price point, good luck on seeing an SSD get cheaper than that in the next year, Mr. Prophet.
Even at $2-3/GB, SSDs would be great! Once you can get a fast SATA II 256GB SSD for $600-800, They will fly off the shelves in premium notebooks like Apple's Macbook Pro, Dell's XPS and Precision, etc
Interesting timing for this announcement. It would be nice to have that 32gb chip in the next generation iPhone. Sorry, had to make the iPhone reference.
While I don't have any hard facts to prove it I would bet that Apple is the biggest user of OEM SSD NAND chips at the moment.
@Ben
There is no way that is true. Apple may be the market share leader in portable MP3 players and many of those use solid state memory but there is _no way_ they are the biggest user.
Cell phones use it. GPS devices. Computers in cars (not some dashboard kind either but the important ones under the hood). Motherboards. ATMs. I could go on for hours.
Many companies in the sectors I named sell more chips than Apple could hope for. Hell RIM is kicking ass in the cell phone market and their phones have the same kind of memory.
Might also be worth mentioning that MLC currently has a 10x shorter lifetime and is significantly slower. So it will be particularly interesting how hynix will perform on this side.
If they continue that MLC trend there it will not help much since it will only be useful for devices were speed and reliability don't matter - probably not SSDs. So I'm very curious to hear more
This is not news. Sandisk already has its X3 chips in PRODUCTION already. Hynix licensed MLC/X3 technology from Sandisk and will finally attempt to mass produce it.
Kobehuang would be correct. Sandisk has been working on X3 for a while now...
Do these take a speed hit like MLC does compared to SLC?
And why stop at three layers? Are we going to see 4, 8, or eventually 16 layer chips?
Without knowing the specifics its difficult to extrapolate performance. However, assuming they are just using refinements of exiting technologies these new cells will be as slow (or likely) even slower than current MLC.
While I am not an expert on NAND memory, this is the problem as I understand it. SLC cells only have two levels, 0 & 1. They are fast and easy to measure, as you just have to measure a 0 or a 1 when you read. Thus cells can be made small and the charge can be extracted rapidly as there is little charge in each cell.
MLC cells are typically 2-bit cells (thus have 4 possible states). This means they have to be bigger as they must hold more charge so that you can distinguish the 4 states in which they can exist which also makes them slower. Nonetheless, a 2-bit MLC cell is smaller than two SLC cells, thus they cost less to produce as each "bit" essentially is smaller than an SLC equivalent.
TLC would essentially extend the MLC concept to 3 bits, and thus now each cell can exist in one of 8 states. Presumably these 3-bit TLC cells are approximately 30% smaller than two 2-bit MLC cells making them again, cheaper to produce.
Thanks for the explanation. I had just assumed that MLC chips were just two regular NAND layers stacked on top of each other.
The level of growth in this field is exciting. I'm looking forward to getting where this is going.