ARM Cortex-A processor roundup

ARM, maker of the low-power processors in cell phones, has recently announced the Cortex-A9 and Cortex-A5 processors. Here's the brief summary of each Cortex processor and what it all means:

Cortex-A5

The low-end of the Cortex line. This replaces the older ARM9 and ARM11 processors which are the current mainstream. It's cheaper, faster, and lower-power than the ARM9 and ARM11. Most of that is because it's manufactured using the newer 45 nm process. This is the entry-level Cortex processor, meant for low-power, low-end devices. Five years from now, the really cheap cell phones will all be using Cortex-A5 processors.

Arrival date: 2011. Speeds will be around 1 GHz, and power about 80 mW.

Cortex-A8

This one is actually out today. It's used on the iPhone 3GS, and can run at over 1 GHz. It's pretty fast. This is the middle of the road for the Cortex family. The higher-end cell phones are switching to this one as we speak, and it'll be around for a while.

Power: about 500 mW. More than the 'A5, but still pretty low.

Cortex-A9

The big one. Supports one to four cores, up to 2 GHz, and is very fast. How fast? Benchmarks are hard to interpret reliably, but it looks to be about twice as fast as Intel's current-generation Atom, but with quite a bit less power, and cheaper. This is the one that will be used in netbooks that don't need to run Windows, like Google's upcoming Chrome OS. It'll also be used in high-end phones, and really power-conscious computers.

Power: depends on the configuration, but about 250 mW/core.

We're still waiting for these to come onto the market, but it looks like it'll be released before the 'A5.


What does it mean?

The first good thing about this is that the Cortex series is a full replacement for ARM's older processor lines, even on the low end. All the Cortex processors have some useful extensions that will be nice to have universally available. It supports some SIMD instructions for speeding up some multimedia code. They support the Thumb-2 instruction set, which lets you mix 16-bit and 32-bit instructions to get better code density without sacrificing speed. They have virtual memory support.

Good stuff.

The problem with photography

Cameras have trouble taking big enough pictures, and the largest computer monitors are nowhere near big enough to do some subjects justice.



I took this picture in Canyonlands National Park this summer, and no matter where I pointed my camera, I couldn't capture how staggeringly big that landscape is. To do it properly, I would need to take a panoramic picture the size of a wall.