TSMC seem to be busy trying to make faster ARM cores.

http://www.marketwatch.com/story/tsmcs-28nm-based-arm-cortextm-a9-test-chip-reaches-beyond-3ghz-2012-05-03

Unfortunately there are no figures for power consumption. It was made on a high performance node so it is unlikely to be a super low power chip one might expect from from an ARM core so battery life may suffer. Now what we need is for battery technology to advance at the same rate and such high speed chips would become more useful.

Yes, I think I can already see similar pattern in smartphones. Faster CPUs, but not so good battery lifes.

Please give us some context as to why this link is related.

I recently received one of these as a gift, hence my passive interest. It gets about 7 hours battery life.

But that is nothing special in the ARM SoC domain. Those things are everywhere. I though you were suggesting that the Tegra was some sort of high speed chip or something.

You don't think that's high speed?? Tegra 3 is already out (apparently), running at 1.3 Ghz x 4. I don't see how you think that's not impressive (enough) or comparable to your fabled 3.1 Ghz x2 machine. Give it a year or two and they'll be everywhere. Things change so fast, it's hard to be surprised anymore.

No, not really. There are quite few other SoCs around that speed. To me the most important thing is power usage. It is rather pointless to have a hyper-speed CPU (like a Core i7 or Itanium) and have it require 100W+ to keep it running. The ARM cores around today are all consuming ~1/4W per core (not including surrounding SoC stuff here) and this seems to be a sweet spot for the power/speed tradeoff. But a 3GHz core on a high performance node is not going to be all that useful unless it is intended for a desktop or laptop. And so far ARM has not been able to penetrate that market in any meaningful way. Even the tablet market is power sensitive ATM. We need better batteries to support the higher speed cores.

ARM processors seem to be running at around 2/3 the power rating of MIPS which has a higher throughput per Mhz. The Texas C55 range nails ARM on power and throughput.

The trick is to balance intermittent power requirements - ability to switch power hungry sections on and off when necessary, and the balance of peripheral features. Arm from what I can see is lacking in these areas, but SoCs like Atmel and others, using Arm, are making headway into this area controlled by more dedicated chips like MIPs, PICs.. and others.

Arm is now a good data pusher, but as a power processor it's been over taken already. Probably a SoC redesign will bring it back to the fore.

Most of ARM-cored chips are system on a chip and you can switch their peripherals on and off easily.

And now with the upcoming big-LITTLE thing perhaps it will solve the issue?

I think the GreenArray, Inc. model of discrete request for computation is a least resistance, less noisy, path.

Overall Arm power consumption is admirable, though.