ARM Releases Guide to Enable 2 GHz Cortex-A9 Manufacturing
ARM has begun offering a Processor Optimization Pack (POP) specifically for Cortex-A9-based SoCs that are manufactured in 28 nm at Globalfoundries fabs.
The POP enables chips that run at a minimum of 1 to 1.6 GHz and 2 GHz in a "typical" scenario.
"As consumer demand for high-performance, energy-efficient mobile devices increases, Globalfoundries and ARM are lowering the risk for customers by delivering optimized Cortex-A9 cores on a proven 28 nm SoC process," said Kevin Meyer, vice president at Globalfoundries, in a prepared statement. "This latest ARM physical IP solution for our 28nm-SLP process delivers industry-leading performance and energy-efficiency, while also decreasing time to market for customers’ latest mobile products."
ARM is offering the POP as an upgrade for single- and dual-core Cortex-A9 designs to achieve higher clock speeds and enable Smartphone vendors to maintain a "competitive edge". The package comes with ARM's Artisan Physical IP logic libraries and memory instances, a benchmark report that describes the exact conditions and results ARM achieved for its core implementation, as well as an implementation guide that offers detail about the methodology that was used to achieve the improved result.
But the question is, is it just high clock speeds and no performance, like bulldozer? Because the galaxy s2 original on att with the 1.2ghz processor beats the new skyrocket with the 1.5ghz dual core, because the snapdragon 3 is just no good.
You should probably remove that link – I almost reported it as spam (I've seen subtle spam before similar to that).
What does a power efficient architecture getting a clock frequency boost by going on a smaller production process node has to do with Pentium (I'm assuming you mean Netburst technology here) where Intel promised we'd see 10GHz in "a few years". Only to find out that if their architecture indeed reached 10GHz the energy density would be greater than in a thermonuclear meltdown?
This comment is flawed - AMD CPU's have a far, far higher transistor count per core than Cortex A9.
It stands to reason that less complex units are easier to manufacture.
It is just that the 32-bit RISC (by ARM Holdings) is easier to manufacture/fabricate than x86. That's a fact.
Honestly, it's a lot of fun to Intel-troll other fabs
Intel makes you pay for the best