The Great TLC flash cash-slash: Coming soon to tablets, Ultrabooks
Super-dense storage tech moves towards 20K cycles
SMART recently announced that its controller technology, the Guardian Technology Platform, could get MLC flash to do 50 full drive writes/day for five years. Now SMART's president has said 1Xnm TLC could be driven to 15,000 to 20,000 PE cycles using SMART's Guardian controller technology, representing a 30X - 40X increase – implying 10-11 FULL DRIVE WRITES a day for five years...
That means 1Xnm TLC could be usable in some enterprise storage scenarios and should be a candidate for tablet computing storage. Ditto 2Xnm TLC. We'll remind ourselves here that Apple bought the Anobit flash controller firm for its MLC and TLC endurance extending technology. Seagate bought DensBits for the same TLC reason.
SMART's president John Scaramuzzo told El Reg: "The Guardian Technology Platform can deliver an SSD that achieves exponentially greater endurance than off-the-shelf TLC Flash. At the 1x process node, TLC with raw endurance of 500 PE cycles can potentially be extended up to 15K-20K PE cycles."
So how did we get to that 10-11 drive writes a day figure? Well, 20,000 divided by five gives you 4,000 a year. That number divided by 365.3 (days in the year) gives you 10.95. We'll round that down to 10-11 PE cycles a day, implying 10 drive writes a day for five years. Not bad, not bad at all.
Here's the background: 3-bit flash, Three Level Cell (TLC) flash has a much shorter endurance because cells can be written to in a program erase (PE) cycle far fewer times than 2-bit multi-level cell (MLC) flash which has a shoer endurance than 1-bit (Single Level Cell or SLC) flash. Each step down in cell size, described by the process geometry size in nanometres, reduces endurance some more.
Thus a 3Xnm TLC cell – meaning 39-30nm process geometry – could do 1,250 PE cycles; a 2Xnm cell (29-20nm) could do 750; and a 1X cell could do 500 (AnandTech numbers).
A flash drive controller can increase endurance through having spare cells on hand to replace worn-out ones; over-provisioning by detecting and correcting errors in the read signal from cells that are wearing out; via ECC; by better interpretation of the read signals; digital signal processing; as well as other techniques.
El Reg's conclusion: TLC flash prospects for tablet and enterprise use are real. Product is coming. ®
Don't forget to post your comments here, please.
Sponsored: What next after Netezza?