SanDisk bigs up its flash postage stamp
Just keep taking the tablets
Magic Quadrant for Enterprise Backup/Recovery
SanDisk has announced a postage stamp-sized flash iSSD product for tablet computers at the Flash Memory Summit, with capacity ranging from 4GB to 64GB.
IDC has conveniently defined a new flash product category, embedded flash, for what are called "highly portable consumer electronics devices", presumably meaning anything from music players through mobile phones to tablet computers. SanDisk says the iSSD is the first product in this new category, which IDC seems to have defined so as to rule out all flash products shipped in mobile phones, music players and tablet computers to date. Frankly, that seems idiotic.
Jeff Janukowicz, IDC's research manager for solid state drives, is quoted in the SanDisk release talking about: "The ultra-thin tablet and mobile computer markets" and "stringent size requirements of small and light devices". Even if we grant him that ultra-thin tablets are a proper computing category compared to thin tablets and, logically, thick tablets, and that mobile phones which, denying common sense, must be supposed not to use embedded flash, the distinction between this and non-embedded flash used in all the other mobile internet devices to date seems artificial.
Some observers may think dark thoughts about IDC inventing a market category to suit SanDisk but we couldn't possibly comment on that notion.
Maybe the category's real distinction is that it refers to BGA-packaged flash? Samsung's OneNAND (pdf) flash came in a BGA package so if BGA-ness is the defining factor then Samsung got there before SanDisk.
The iSSD product is 1.85mm thick and claimed to be smaller than a postage stamp. We're all so blasé about physical size that we're not that impressed - SanDisk announced a finger nail-sized SSD only last month. That one went up to 16GB of capacity but the iSSD holds four times more data. The iSSD uses multi-level cell (MLC) flash, probably two-bit, and if it is used in devices that are used more for content consumption than content creation the well-known MLC write endurance problem should be much less of an issue.
The other measurements are 16mm x 20mm and the little tab weighs less than a gram if it's a 4GB jobby and a little more if it's a 64GB version. It comes with a SATA interface (SATA II we reckon) in a Ball Grid Array (BGA) package (157 ball BGA with 0.5 mm pitch). It offers 160MB/sec sequential read and 100MB/sec sequential write speeds.
SanDisk says the packaging means manufacturers can "avoid the additional weight and thickness required by a connector-based SSD solution". It also means that the SSD sits snugly against its host board and impedes any cooling airflow less - which only really matters if it is used in a device with a fan, such as a notebook computer.
It supports SMART diagnostics and the TRIM command and its write endurance is defined by SanDisk as the total number of terabytes that can be written (TBW) before the drive fails. For a 4GB model it is 2.5TB, 5TB for the 8GB version, and so on linearly to 40TB written for the 64GB version. The predicted lifespan requires customers to know the average data write rate per time period and this is used to divide the TBW figure and so produce a working life period.
The iSSD, available for sampling now, is being evaluated by some unidentified manufacturers and its price will vary with the order quantity. ®
COMMENTS
Write endurance
The ratio of the total amount written to the capacity is 625:1. For many purposes that will be fine, provided that the wear levelling works. I would suspect that it will see out the lifetime of most notebooks on moderate usage. Fitted with a 128GB drive, this implies a lifetime write allowance of 80TB. If the notebook lasts 5 years, then that's 44GB per day used 365 days per year. Of course many notebooks won't be used that frequently or consistently,.
However, it the 625:1 the full story here? Flash memory is written out in large pages and a relatively small 4KB write can involve the controller writing out a very large page - maybe 128KB or more, depending on the page size of the memory. Controller optimisation with write caches and so on can reduce that overhead by "rolling up" writes into a few, larger ones. However, the issue here is does the 625:1 ratio apply at the back end (what the controller writes to cache) or the front end (what the computer writes to the drive)? If your "write allowance" is rapidly used up by many small writes (quite a common thing when managing file systems), then the drives might fail rather faster than the specifications would indicate.
Yeah, 157 and for good reason.
The 157 pins are there for a reason. In an "End-User" scenario you would find most pins either connect to the power planes for improved power distribution inside the chip, or are non-connect (NC).
Now with only Power and SATA signals documented, you will probably find a lot of the NC pads are actually connected to stuff like JTAG for the initial programming of the SATA controller's firmware, Direct access to the flash array's address and data lines for factory testing of the array, and any number of other functions.
One way to tell if such hidden functions exist is if some pads need to be pulled high or low for no other reason than the datasheet says it's required for proper operation but with no other reason.
157 pins, indeed.
157 pins does seem rather a lot. Perhaps the little BGA package doesn't connect to the outside world via SATA. Probably there's a connected supporting chip which is the SATA interface.
IT infrastructure monitoring strategies
Requirements Checklist for Choosing a Cloud Backup and Recovery Service Provider
Data control in the cloud
Cloud based data management
Enabling efficient data center monitoring
