AMD confirms 'Atom-smasher' chip on its roadmap
But Intel's Atom issues mean there's no rush to release it
AMD will unfold its plan to take on Intel's Atom in November, newly promoted CEO Dirk Meyer said last night.
As Register Hardware has reported in the past, AMD's is currently working on a chip codenamed 'Bobcat', a single-core, 64-bit processor designed for low-cost laptops and Small, Cheap Computers.
Past leaks have indicated that Bobcat will debut with a 1GHz clock speed, 128KB of L1 cache, 256KB of L2 cache and an 800MHz HyperTransport link out to the rest of the system. Its on-board memory controller will handle 400MHz DDR 2 chips.
The whole thing is set to consume no more than 8W and sit inside an 812-pin, 27mm² BGA package. 8W is more than the Atom N270's 2.5W, but then the Atom has to be accompanied by a big old northbridge that consumes up to 6W.
Speaking after the announcement that AMD's made a $1.19bn loss and ousted the previous CEO, 'Silly Old' Hector Ruiz, Meyer mentioned the new chip, though not by name. He said the company will say more in November.
Why wait? Intel's having problems getting Atom chips out the door - the problem is not production per se, but limited testing facilities, the company says - and that's holding suppliers up. Asus is still using Celeron M chips in some of its Eee PCs - the 900 and 904, for example - while HP turned to VIA's C7-M chip.
Still, AMD's pause will allow it to see how Intel manages Atom alongside its Celeron line. This week, Intel CEO Paul Otellini said Atom will not replace Celeron, at least not for the moment.
His argument is that Atom is aimed at Small, Cheap Computers - Intel calls them "netbooks" and "nettops" - while Celerons are meant for low-cost machines.
There's clearly plenty of overlap, but we can see Intel wanting consumers to associate Atom with ultra-compact systems and Celeron with full-size, low-price boxes. But then isn't that what the revived Pentium brand was all about? And isn't the Core brand becoming ever more low-end?
Sponsored: Benefits from the lessons learned in HPC