Chipzilla's 1GB phone
Gunning for OMAP with Flash 'sandwich'
Microprocessor Forum Texas Instruments had a very low profile at this year's Microprocessor Forum. Like Sun, with whom they shared a table outside the press room, they officially weren't there. Or at least not presenting.
But TI's OMAP is the dominant platform for smartphones today, somewhere Intel very much wants to be.
Intel yesterday announced what it described to us as "a Flash sandwich" - the first 1.8V flash memory allowing four 256 MB modules to be integrated with an XScale to create 1GB of code and data in a phone. That's a lot of memory for a phone.
(Those are Megabits - and that's a Gigabit, we should have made clear. The ongoing confusion stems from the historical difference between memory people and storage people, and isn't helped by dunces like us. See I say Petabyte, You say Pebibyte...)
Such phones won't be appearing until later next year, it's safe to assume. Intel has yet to announce a Tier1 design win for XScale, but the StrataFlash package is already sampling with volume production promised for next year.
There are three versions, a basic 200MHz/128Mb PXA261with quantity list pricing of $36.10, and two "put everything on that" club version, the 200 Mhz and 300 MHz/256 Mb versions of the PXA262 at $54.60 and $62.60.
More details and prices can be found here - but you're safe, this detailed release is almost entirely free of gwana-gwana. Apart from one minor detail.
The "System-In-A-Package" is slightly suggestive. It refers to one of the industry's forthcoming industries: whether it's better to build phones with two "CPUs" - as they are today, with one for the radio and one for the data - or one. A two chip phone isolates the baseband stacks from bugs in the general purpose processor affecting the radio.
TI recently announced that it planned to deliver a one chip phone by 2004. Intel has talked about reducing all the analog circuitry required for radio onto one chip in the next ten years.
Last week Bob Hills, a director at Intel's emerging platforms lab told us that the first landmarks in this ambitious plan would be to involve initial analog processing - capacitance and induction - right into the circuit board. The goal would be five to seven years.
He ran through a variety of mobile technologies that Intel is brewing, including display and power supply technologies. For the latter, he said Intel was working on something "broader" than fuel cells, and stressed that today's research wouldn't draw Intel into the component business, such as LED. And he personally thinks people will have more than one device.
No, he didn't think the phone business would ever become quite like the PC business - the Dellification of phones is a running debate at The Register, but what are the odds against phones becoming some kind of personal hub, much as Jeff Hawkins envisaged with the original Handspring?
You can do a lot with one gigabit. ®