Read an article today in ARS Technica titled NAND flash gets baked, lives longer that researchers at Macronix have come up with a technique that rejuvenates NAND bit cells by heating them. The process releases the bit cells captive electrons and returns it back to a fresh NAND cell.
As discussed previously in this blog (e.g., see The End of NAND is near, maybe… and What eMLC and eSLC do for SSD longevity) as NAND technology shrinks to smaller transistor diameters, their longevity and durability decreases proportionally. Which means that with denser NAND chips coming out over the coming years, they will become increasingly short lived.
With this new approach and an awful lot of engineering to zap a NAND bit cell with intense heat (800C) can take dead NAND cells and bring them back to life. Apparently this rejuvenation process has been known for some time and had been in use for phase change memory but had not been applied to NAND cells in memory. They were heating batches of NAND cells for hours at 200C but this wouldn’t be very practical in production.
The new NAND memory cells are designed with a resistive heating element on top of them, which when enabled can heat the NAND bit cells beneath them. According to some news reports I’ve read this enables the NAND cell to go from 10K P/E cycles to 100K P/E cycles. And the heat only needs to be applied in occasional pulses to keep cells operating within parameters. As such, it can be used sparingly and not cost too much energy in the process.
Another side effect of heating is that erase cycles operate faster than at normal temperatures, which now adds the possibility of heat assisted NAND cells. Erasure being one of the key bottlenecks to NAND write performance anything that can speed this up would help.
The smartphone research was done by NEC. They took an Android phone and modified the O/S to use an external memory card for all of the App data needs.
Then they ran a number of Apps through their paces with various external memory cards. It turned out that depending on the memory card in use, the mobile phones email and Twitter Apps launched 2-3X faster. Also, the native web App was tested with over 50 pages loads and had at best, a 3X faster page load time.
All the tests were done using a cable to simulate advanced network connections, above and beyond today’s capabilities and to eliminate that as the performance bottleneck. In the end, faster networking didn’t have as much of a bearing on App performance as memory card speed.
(NAND) memory card performance
The problem, it turns out is due to data writes. It seems that the non-volatile memory used in most external memory cards is NAND flash, which as is we all know, has much slower write time than read time, almost 1000X (see my post on Why SSD performance is such a mystery). Most likely the memory cards are pretty “dumb” so many performance boosting techniques used in enterprise class SSDs are not available (e.g., DRAM write buffering).
Data caching helps
The researchers did another experiment with the phone, using a more sophisticated version of data caching and a modified Facebook App. Presumably, this new “data caching” minimized the data write penalty by caching writes to DRAM first and only destaging data to NAND flash when absolutely necessary. By using the more sophisticated “data caching” they were able to speed up the modified Facebook App by 4X.
It seems that storage sophistication matters even in smartphones, I think I am going to need to have someone port the caching portions of Data ONTAP® or Enginuity™ to run on my iPhone.