As the 2011 Freescale Technology Forum wraps today in San Antonio, Texas, here are a few of my initial takeaways. They’re nowhere near comprehensive – it’s a huge (and hugely informative) event with 2,300 attendees and 350 hours of technical presentations. We’ll need to take another look at the conference in the next print Tech Design Forum
AMPing up the control plane
The new QorIQ AMP family really looks cool. The e6500 64bit Power Architecture core is worth a 2X performance boost over Freescale’s previous generation of silicon for the control plane and mixed control/data plane market. But by moving to 28nm, applying aggressive multithreading, boosting frequencies up to 2.5GHz, bringing back a more user-friendly version of the AltiVec instruction set, and offering more packet processing acceleration, the overall relative boost is 4X against the existing eight-core QorIQ P4080.
Plus there’s power management and enhanced virtualization, with an ability to take the 12 cores on board the launch AMP T4240 chip to 24 virtual cores.
We’ll soon take a more under-the-hood technical look at the AMP series, and also how it relates to the QorIQ Converge silicon launched earlier this year.
However, in a nutshell, this is all to do with the non-consumer side of handling the explosion in data traffic, with AMP at the data and control plane, and Converge providing solutions for scaleable basestation development.
The first of the T4240 chips will ship in Q1 2012.
Going from the lead product announcement to the strategic messaging, there was a strong sense of Freescale moving forward after its recent IPO. In the short term, the company is grabbing back market share, up from 40% to 45% of the embedded market according to the latest Gartner data and market leader over Intel’s 27%. In the longer term, the company is setting out its stall for a number of key markets.
The more interesting for me were its comments on the pico and femtocell basestation markets, tablet computing and energy efficiency. And in particular, it’s the company’s recognition that these markets need to see their business models mature although to meet customer demand you still need to be delivering and advancing products for them today. The understanding that they might need three to four years for that to shake out is solid and pragmatic.
Take smaller basestations. There’s still a big question over who should pay for them: the network operator (to assuage complaints about performance – naming no names, but I only get two bars on my iPhone in the office) or the customer (to satisfy the greater demands they are placing on mobile devices). With its Converge design win for Alcatel-Lucent’s palm-sized lightRadio basestation-in-a-box, Freescale gets plenty of kudos, but it’s also comfortable with the commercial issues working themselves out in the medium term.
Energy efficiency is another one largely in the hands of the consumer. Perhaps you can deliver products now that encourage more conservative use but a user panel organized at FTF kept coming back to the issue of getting people to engage in the long term. All the data now shows that people get involved initially but their active participation with today’s products falls off pretty sharply. Automation and education will do a lot here, but it’s going to take time.
And then tablets. CEO Rich Beyer made it clear that Freescale is not going after the big OEM design wins and getting slaughtered on price. Rather it sees plenty of opportunity in the ‘white box’ space, essentially providing the fundamental platforms with some adaptability but with the customer primarily providing the industrial design. Again, that makes a lot of sense.
There was some surprise when Beyer said that across the whole company, the split between software and hardware engineers, still falls very heavily in favor of hardware (at roughly 4:1). After all, the received wisdom is that silicon companies should be recruiting software engineers like crazy and Freescale sees itself primarily as an embedded computing company.
First, Beyer left no doubt that the software headcount will increase, but he also faces the conundrum of getting his customers to pay for that software (as do all embedded vendors). And those customers still expect it for free – given the ever increasing role software plays that simply isn’t a sustainable position today.
Second, there is no ‘right’ or ‘wrong’ about how this should pan out. There is the view argument that says your hardware and software teams should be at least equal in size, although that depends on your market in many cases (if you are heavily dependent on DSP, for example, that balance may be necessary).
But it’s also becoming true that you can do a lot by partnering with third-party software providers, using your internal team for as-needed differentiation and tool development. Look at Android, a market overflowing with developers you can tap into as a result of Google’s open source strategy, and so often capable of providing building blocks at low cost.
Either model can feasibly work – the real challenge goes back to getting the ROI from your development effort, and a highly partnered Freescale (among many others, Mentor Graphics, ENEA, Wind River, Green Hills Software, and QNX Software Systems were all present and correct at FTF) may have the right balance for its objectives.
One other observation that may effect the engineering split. Freescale – even now that it has exited the handset mobile IC business – has a very broad portfolio across a huge number of markets. And that leads into my final point: despite that breadth, the company also seems to have a strong handle on how it’s developing the existing business and how it is going to take it forward.
NOTE: You might also want to check out our article on ‘Migrating from single to multicore processing on QorIQ technology‘ in the recent Embedded special edition.