Flash Followup

A month back, I made some admittedly strong predictions about what the day-prior unveiling of Apple's iPod nano meant for the future of flash memory versus small form factor HDDs, particularly in portable multimedia record and playback devices (i.e. not in laptop and desktop computers, where I think the HDD vendors' business will remain unthreatened for a long time no matter what Samsung's semiconductor group says…a group that might want to remember that the company also sells HDDs!). It's been gratifying to see other analysts pick up the debate, and I thought I'd offer some follow-up thoughts.

First off, a recently published iSuppli teardown report estimated that the bill-of-materials cost of the $199 2 GByte iPod nano is $90.18. Of that, the two 1 GByte NAND flash memories together cost $54, roughly 50% of what they'd cost on the open market and the probable result, at least in part, of a recently concluded partnership agreement in which Apple agreed to buy 40% of Samsung's capacity. This cost estimate really drives home a point I made in my cover story earlier this year, that flash memory is a disproportionately high percentage of the total cost of a semiconductor-based digital audio player, and that companies who make both players and their memory chip building blocks therefore held a notable cost advantage.

What none of the other analysts I've read seems to have picked up, though, is that the 50% price cut might also reflect the fact that inside each chip's package is a MLC (multi-level-cell) NAND flash memory comprised of only four billion storage transistors (and storing two bits of information within each transistor), therefore with smaller die size and lower silicon cost compared to a true eight billion-transistor equivalent. While Samsung has publicly and vigorously promoted SLC (single-level-cell) NAND versus MLC (offered by competitors such as Sandisk and Toshiba) in the past, the company has published MLC NAND technical papers at conferences such as the ISSCC for quite some time now. In fact, Samsung quietly introduced an 8 Gbit (i.e. 1 GByte) MLC NAND flash memory one year ago. With sufficient EDAC hardware and/or software support built into the system CPU, MLC flash memory would be perfectly adequate for use in a portable audio player, since the absolutely fastest-possible write performance isn't a critical requirement and since each memory storage element likely won't need to be rewritten millions of times over the life of the player.

Here's some more flash memory cost data. I'm on a business trip right now, and before I left I ripped a couple of rental DVDs to 192 kbps MPEG-4 AVC format to watch on my Sony PSP. I was pleasantly surprised to find that five episodes worth of Six Feet Under: Season Three (discs 1 and 2), not including the DVD extras, comfortably fit in just over one half of a 1 GByte Memory Stick PRO Duo. Over the past few weeks, I've been archiving away the relevent RSS feeds from DealNews and Techbargains, and the cheapest price I've seen for the 1 GByte Memory Stick PRO Duo is $61.70 (no rebate required). The 2 GByte Memory Stick PRO Duo is even cheaper on a cost-per-GByte basis, at $116.99 (after the application of a $10-off coupon).

Tell me again why I need a 6 GByte 1" HDD? Well, I had a lengthy conversation two weeks ago on that very topic with Seagate's PR Manager, John Paulsen, and Consumer Electronics Product Manager, Pat O'Malley. Aside from, for example, forecasting the embrace of video (and to ever-higher resolution video) in portable electronics equipment, which I'd already indicated in my month-back blog post were the best hope HDDs had to retain a significant chunk of their current business, John and Pat big-picture pointed out that by nature, human beings are packrats and will inevitably fill whatever storage capacity (in a HDD, or a house) is available to them, with 'stuff'.

I somewhat reluctantly agree with their observation; right now, in fact, my 5 GByte iRiver H10 is stuffed to capacity with 192 Kbps WMA tracks and I wish I could squeeze in even more music, even though that music came from Yahoo's subscription service and, in the extreme case, I could use Yahoo's servers as my archive medium and only copy down a few albums' worth of tunes at a time on an as-needed basis. So while I agree (and said a month back) that a HDD-based system will always deliver higher capacity than its flash alternative (beginning at a system price point that allows the HDD vendor to turn a profit), flash memory will allow systems to break through that HDD-defined price point barrier (along with delivering benefits such as greater shock tolerance and lower power consumption). And will those lower-priced systems deliver adequate capacity for a large-enough segment of the buying public? That's up to the packrats.