Intel’s recent purchase of Wind River Systems is an important move that could have a significant impact on embedded systems design. One possible outcome is better support for multicore system-on-chip (SoC) development.
Intel’s $884 million purchase of Wind River is part of the semiconductor giant’s recent move into the embedded systems market. Two years ago, Intel announced a corporate-wide SoC initiative aimed at mobile and consumer devices. Linux is strong in these markets, and Wind River has a Linux OS offering. A recent EE Times analysis article says more about the strategic reasons for Intel’s purchase.
Wind River, which Intel plans to run as an independent subsidiary, has been the leading independent provider in the embedded software development market. This market includes operating systems and associated development tools. Wind River’s offerings include its proprietary VxWorks RTOS, a Linux OS, development tools, and middleware.
The acquisition is part of an ongoing trend in which major semiconductor vendors are getting more involved in software development. As I noted in a recent blog, recent studies have shown that software costs exceed hardware costs when you get to IC development at 65 nm and below. Semiconductor companies are not only providing reference software and few device drivers – they’re increasingly expected to provide a more complete software stack. With Wind River, Intel is going even further and supplying commercial operating systems and development tools.
The Wind River acquisition comes at a pivotal time for the embedded software development business. Until now, most software developers could get by with cheap or free tools, and many use Linux instead of proprietary operating systems. This is one reason why the embedded software tool business is much smaller than EDA, even though it serves more people. This is also why EDA vendors haven’t ventured far into this business.
But times are changing. Software has become the major differentiator, and bottleneck, in many contemporary embedded systems. Software does not have the kind of disciplined, formalized design and verification methodology that has emerged in chip design. Software development really needs to start before silicon is ready. There are many new attempts to solve these problems, ranging from tools such as Cadence Incisive Software Extensions to the virtual platforms available from various providers.
With their combined expertise in hardware and software, Intel and Wind River can potentially help make embedded software development and verification easier. One interesting question is whether Intel, which has tremendous strength in formal verification, will apply any of that technology to software verification.
But the real inflection point for embedded software development providers is the move to multicore ICs, which will require a new generation of development and debugging tools. And this is where the Intel/Wind River combination could have a significant impact. Intel has been investing heavily in parallel processing research for multicore chips, helping fund such efforts as the Parallel Lab (Par Lab) at U.C. Berkeley. Wind River has been working on multicore development and debugging as well. Intel is also, of course, a major provider of multicore processors. You can bet that Intel will put multicore processors on its own SoCs and put together a programming environment for them.
Analyst Gary Smith said that both Oracle (which recently purchased Sun) and Intel “seem to have a better appreciation of the dynamics of the multicore market and see it as a major growth opportunity. Software is where the money is, and if the EDA industry doesn’t step up to the challenge they will lose a major opportunity to grow their TAM [total available market].”
So what role should the EDA industry play in solving the multicore SoC design and development challenge? This question looms even larger with Intel’s ambitious Wind River acquisition.
It's yet another ancillary purchase by Intel. It'll fail like EVERY ancillary business Intel has bought. Intel doesn't know how to run a business except saving money via off shore tax schemes and twisting customers arms to buy it's product.
It knows how to make microprocessors. Period.
Dan -- I see the acquisition as more of a call to action for the embedded processor, embedded software tool, and EDA communities. It's time to move forward on multicore SoC development, and each of these communities can contribute.
I see the Intel acquisition of WRS as an act of war on the embedded processor community:
That will in fact suppress innovation versus foster it, if history repeats itself once again.