First of all, there’s the power consumption and management. According to Terry Ngo, senior manager of the consumer connectivity business for Atheros, there’s a lot going on. “These are all CMOS mixed signal implementations, with no GaAs,” he pointed out first. He also indicated heavy usage of power domains to be able to lower voltages, and a lot of clock gating. Also, Atheros has spent a lot of time on things like a linearized power amplifier and an efficient implementation of the SDIO interface in these designs. The power consumption of the AR6003 is 20% lower than its predecessor, with more performance from 11n rates instead of just 11g rates.

A second point is the 1-stream 802.11n implementation, again balancing low power consumption with good performance. Also, in the case of the AR6133 shown below, Bluetooth and Wi-Fi are combined to work with a unified antenna, reducing device size and cost.

While designed for phones, designing around an SDIO interface means these chips are able to interface with a wide range of SoCs, working with more devices. Drivers are available for Linux under GPL, and for Windows Mobile, Android, Symbian, Chrome OS, and others.

There’s a lot more to the technical details of these announcements. As we see more and more devices going to Wi-Fi, and the demands of power and performance getting more challenging all the time, parts like these can make the difference in tough designs.

They claim the new microMIPS instruction set keeps most of the performance while reducing code size by a bit more than one-third. Both devices are fully compatible with the regular MIPS ISA. Using the new ISA can cut system costs through reduced memory requirements.

Along with a few other features, MIPS also took on one interesting feature that mostly gets overlooked: a flash accelerator block in the M14K.

One of the first things programmers often do is copy code over from flash into RAM for faster execution. It seems most architectures just let this exist instead of fixing the weakness. MIPS is taking a shot at accelerating flash performance with some pre-fetch capability - we didn’t get a lot of details but it’s worth a closer look if you’re interested. Obviously, if you’re generating a whole lot of misses, this might not help and you might still have to shadow some sections of code, but the idea of being able to execute out of flash better is an interesting one.

Also around these cores is extension capability. Optional units include CorExtend, which is a way to bolt in a co-processor in an SoC environment, and a User Defined Instruction (UDI) capability. Off core interfacing to peripherals is with the AHB-Lite interface.

Lastly, there are the iFlowtrace extensions for debugging - something we’re about to have a viewpoint from MIPS on in our November print issue.

These enhancements along with the new microMIPS ISA make this an interesting development.

This week, Wind River’s Ken Klein issued his first blog post on “Wind River’s Next Chapter“. Interesting reading. A couple other editors have commented on this post, so I thought I’d add maybe more than two cents.

The core of the conversation goes something like this: Wind River is part of Intel, therefore eventually only Intel technology will be supported by them. Anything’s possible, but that is a very highly unlikely outcome.

Ken used the word “firewalled” and we talk about that a lot in our industry, but if there are two companies that understand it, they would be Wind River and Intel.

Somewhere in my wayback machine:

• I am a recovering engineer, and I used VxWorks in the beginning when the kernel inside was VRTX. Yeah, Ready Systems. It wasn’t obvious until you really started digging around under the hood. The VxWorks we know today would have never existed if they hadn’t embraced VRTX and built on it, making a better development solution.
• One day years later, my boss called me into his office and said I should call this guy at Intel (who now goes by @Intel_Jim , you can follow him and me at @dondingee ). Jim and I spent many days and more than a couple nights on phones, in conference rooms, and in hotel lobbies and restaurants trying to figure out how to get Motorolans and Intel-ers to not only talk, but work together on embedded (er, applied) computing. It was total sacrilege at the time. We avoided more than one airstrike from Schaumburg, Santa Clara, and Austin. But the results were worth it (I think, ask Jim).

Things never get better if people aren’t bold enough to embrace an innovative solution.

There is no one company that can keep up with all the aspects of both the technology and the business acumen needed, and let’s talk the biz side for a second. I happen to think that Wind River will remain a independent subsidiary for one big reason: the Intel sales force and the Wind River sales force are two different animals, and will probably stay that way. Sure, competent sales people share some traits and skill sets. But the two selling cycles are completely different. While the sales teams might roll up strategies to the Intel side (Intel is big on the two-above-your-box matrix boss concept, so Wind River types will find themselves matrixed to an Intel type soon) they won’t integrate into one entity easily - and their customers probably won’t allow it, because of the types of support they’ve grown used to.

There is no doubt the Wind River offering will get a lot better on Intel architectures, and Intel’s software strategy gets more important every day. But as long as Wind River plays the business side right and develops the right type of technology for the markets, I see no reason to disbelieve they can remain an effective provider for both Intel and non-Intel processor technology.

If you think beyond these two companies for a second … consider this. Adobe could not possibly exist unless firewalls worked. Neither could Mentor Graphics. Neither could TSMC. For every Apple Computer that builds a self-contained environment, there are 100 companies that figure out how to work with multiple competitive camps in their customer bases and ecosystems.

For the competitors that want to try to spin this story that a customer can’t possibly work with Intel/Wind River because they can’t play fair … be careful. If you’re ready to compete on your technology and business relationship, that’s the right place to go.

For customers, do your homework. Consider both the technology and the business relationship you want. You might not be comfortable with this relationship, and it’s ok. But don’t rule something out if you see this relationship providing the right technology but are squeamish about the biz side. If you need legal advice to increase your business comfort level, get it. If you need more information, ask for it. And if you like the relationship, tell them that, too.

Ultimately, everyone has the right to choose between Intel/Wind River solutions and many others, and the right providers listen and will win - and so do their customers.

I might be wrong, and I’ll see you in the sanitarium if I am. But I’ve seen (and lived in) this show, and I think I can guess how it goes. Change can be fun, deal with it. Shoot back a comment with your views and experience with how this change is going from your desk.

Also news from TechCon3 by ARM on the new Cortex-A5 MPcore processor, targeting embedded devices. Better perf than ARM1176JS-S, but power and footprint of AMD926EJ-S - twice as power efficient. Includes TrustZone security and NEON multimedia engine. MPCore increases multicore scalability while containing power consumption. Fully Cortex-A8 and -A9 compatible so lots of software.