Atom-based system-on-a-chip devices will start pumping out of TSMC's fabs in volume in the fourth quarter—or so says a DigiTimes story that cites anonymous industry sources. As we reported on Monday, Intel officially plans to outsource Atom SoC production to TSMC's fabs in Taiwan.
The Q4 2009 time frame sounds about right, because Intel doesn't plan to release Atom-derived SoCs (code-named Lincroft) until late this year or in early 2010. If DigiTimes' sources have their facts straight, then Lincroft might be available in Intel- and TSMC-produced flavors right from the beginning. DigiTimes does append "at the earliest" to the Q4 time frame, though, so we wouldn't necessarily count on it.
In related news, DigiTimes has spoken to market research firm The Information Network, which believes Intel might have chosen to outsource Atom SoC production because of cost-cutting concerns. ARM President Tudor Brown apparently shares the same view, too. (Intel itself says it's teaming up with TSMC to "accelerate deployment of the [Atom] architecture through multiple SoC implementations.")
The Information Network stated in January that Intel lost $1.14 billion in revenue by making cheap Atom processors instead of higher-priced Core 2 chips. The research firm expects Intel to miss out on another $2.16 billion in revenue this year for the same reason. If those numbers are accurate, then increasing margins by cutting production costs could be a priority for Intel—although who knows how much of a margin increase the company can squeeze out of sub-$50 processors.
|The TR Podcast is live, so come ask us stuff!||0|
|AMD shows off DirectX 12 performance with new 3DMark benchmark||40|
|Intel and Micron sampling 3D NAND based on floating gates||20|
|Report: Microsoft to build an Intel-powered, non-Pro Surface tablet||49|
|Toshiba's 3D flash spreads 16GB over 48 layers||6|
|Cougar's 300M gaming mouse looks awfully familiar||15|
|Wednesday Night Shortbread||69|
|EVGA Hybrid liquid-cools the GeForce GTX 980||20|
|Rowhammer attack exploits shrinking process size in DRAM||40|