Brian's Brain

This blog post references my cover story 'Solid-State Storage: Feasible Plan Or Flash In The Pan?' in EDN's September 27, 2007 edition. It's one of a series of web addendums to the print writeup.

In that article, I discussed the capacity, constructed cost and cost/GByte trends of magnetic and solid-state storage (along with the hybrid intermediary), and calibrated them against the capacity and cost expectations of various computer form factors and embedded applications. In light of those factors, Western Digital's unveiling yesterday of a two-platter 5400 rpm 2.5" 320 GByte HDD certainly caught my eye. WD's announcement came on the heels of Toshiba's 5400 RPM 320 GByte 2.5" HDD release in late August (PDF). One day earlier, Samsung had also unveiled a 4200 RPM 160 GByte 1.8" drive, followed by Toshiba's near-clone (3,600 RPM) in early September.

Back in April of last year, I predicted (based on to-date PMR trends) that 1 TByte 3.5" HDDs would soon appear. Within 9 months, in fact, Hitachi had released its Deskstar 7K1000, a 7200 RPM HDD notable for incorporating five platters' worth of storage (at 200 GBytes per platter, alternatively stated as 100 GBytes/side). HDD market share leader Seagate chose to wait until late June to unveil its four-platter 1 TByte design (at 250 GBytes/platter).

Roughly one week prior to Seagate's announcement, Samsung unveiled a three-platter 1 TByte product (334 Gbytes/platter). In the interest of full disclosure, I feel compelled to point out that although Samsung trumpeted that the drive was 'currently shipping', Google searches on the HD102UJ (16 MByte RAM buffer) and HD103UJ (32 MByte buffer) suggest otherwise; the few vendors that even bothered listing the products in their catalogs report no product inventory. WD introduced its own four-platter 1 TByte HDD in late-July standalone and external storage appliance announcements.

The contrasts in these four vendors' strategic approaches to the 1 TByte 3.5" HDD opportunity are very intriguing. Hitachi chose (and to date, per my PR contact, still chooses) to service the market with a bulky, power-hungry and (potentially) costly five-platter configuration...but in doing so, thereby not pushing per-platter capacity as aggressively as its competitors did, it garnered a nearly six-month proprietary position at this particular capacity metric. Hitachi's sole-sourced stance in an otherwise commodity market probably didn't cultivate tremendous demand, but what product the company did sell during the first half of this year probably came with an estimable profit margin.

Note that I'm qualifying my 'costly' description with the 'potentially' modifier. I've been collecting Dealnews and Techbargains listings on the Hitachi, Seagate and WD drives over the past few months, to monitor price trends at this particular capacity point. None of these 1 TByte drives have yet hit the 20 cent-per-GByte threshold of today's most cost-effective HDD capacities, but Hitachi's five-platter product has been as low as $300, with WD's four-platter drive hitting the $260 mark. Interestingly, though, Seagate's four-platter drive has only dipped to $330, and befitting my earlier comment, Samsung's drive has been AWOL from both price-monitoring sites. From the data I've seen (although I admittedly don't have insight into Hitachi's profit margins), Hitachi's downward price-pushing abilities haven't been hampered by its one-extra-platter approach, at least at retail.

This tug-of-war between leading-edge and mature storage technology generations is reminiscent of the tradeoff that many of you encounter when planning chip designs; do you go with a leading-edge lithography that delivers smallest possible die size but at yield and other risks, or do you instead rely on a more silicon-hungry but also more mature, less-aggressive-dimension process? And speaking of chip designs, the flash memory camp isn't wilting in the face of the HDD capacity challenge. Samsung just unveiled first proof-of-concept prototypes of its 64 Gbit (multi-level cell, aka 32 billion storage transistor) NAND flash memory chip, based on 30 nm process technology. Here's one translation of that achievement; using only 16 MLC-configured chips, it'll be possible to construct a 128 GByte SSD. Granted, Samsung doesn't promise mass production until 2009, but given the company's track record on flash memory, I wouldn't bet against it:

The 30nm-class 64Gb NAND flash marks the eighth consecutive year that the density of memory has doubled and the seventh straight year that the nanometer scale has improved for NAND flash since the 100nm 1Gb NAND chip was developed in 2001.

Put it this way; I trust Samsung's semiconductor group's projections more than the pronouncements of the company's HDD group ;-)

Followup: Samsung 're-launched' its three-platter 1 TByte HDD at the end of November. Should we believe the company's 'now shipping' claims this time?