Guest columnist Richard Kirk, product manager at GE Intelligent Platforms, describes how the Intel Core i7 processor is challenging Freescale's Power Architecture in military embedded systems
Historically, embedded computing systems designers had a tough choice to make when selecting a processor architecture.
On the one hand, choosing Intel gave them access to raw computing power, the cost-effective PCI Express infrastructure and a range of operating systems support.
On the other hand, Freescale’s Power Architecture gave them the AltiVec floating point processor, an easier-to-manage power envelope and the expectation of the long-term product support and availability that are crucial to the multi-year deployments typical of embedded computing applications.
Equally typical of embedded computing is its growing emphasis on “SWaP” – size, weight and power – and on performance/watt. While the mainstream market has, for the most part, single-mindedly pursued higher clock rates and multiple cores in an attempt to extract maximum computing performance, the embedded market – and especially the military/aerospace market – has been driven to deliver a more balanced approach which makes trade-offs between processing performance, power consumption and heat dissipation.
The requirement is increasingly to pack more capability into the space- and weight-constrained environments that are typical, for example, of unmanned military vehicles.
Intel’s announcement of no fewer than 27 new processors earlier this year – with 12 targeted specifically at the embedded computing market – may have indicated the beginning of a shift towards the x86 architecture for embedded computing applications.
Not only did Intel commit to long-term availability of, and support for, the new embedded processors, but their performance/watt appears to be around a 20% improvement over their predecessors. This gives either better performance for the same power consumption, or comparable performance with lower power consumption.
Three versions of the Core i7 will be available. To the embedded computing industry, of as much interest as their clock speeds (1.06GHz, 2.0GHz and 2.53GHz) is their power consumption (17W, 25W and 35W TDP).
Much of the media coverage at the time of the announcement focused on Intel’s new 32nm fabrication process. While impressive in itself, this is one of the key factors influencing power consumption.
However, it has also enabled what was previously a three-chip design to become a two-chip design, as integration of the memory controller and PCI Express interconnects within the processor has eliminated the need for a traditional Northbridge chip.
The impact of this is to save precious board real estate – allowing the board to become smaller, or alternatively, allowing it to include more functionality. In the case of the GE Intelligent Platforms CT12 and VR12 Core i7-based single-board computers, for example, it has contributed to allowing for the provision of up to two PMC or XMC sites, 16Gbyte of flash memory and enhanced connectivity capabilities.
On-chip graphics
This high degree of integration is further enhanced by the Core i7’s inclusion of an on-chip graphics capability which appears to offer more performance than previously (initial GE benchmarks using the 3D mark test suite indicate a level of performance comparable with single-board computers with a separate graphics capability) and will, in some cases, obviate the need for a discrete GPU – saving not only board space but also power/heat.
EEC memory
As well as its lower power consumption and more compact chip set, the Core i7 is attractive to military/aerospace system designers because of its inclusion of ECC memory. For the typical Word and Outlook user, ECC memory is irrelevant – but for mission-critical applications with their requirement for absolute data integrity, ECC memory is a vital capability.
Integration of ECC memory is not new from Intel – but it is new in a chipset designed with the performance/watt characteristics required by the mobile computing market – characteristics that are, in many ways, a good match for the requirements of the military embedded market.
Of particular interest to those comparing the Freescale and new Intel offerings is that the Core i7 provides a floating point processor, whereas Freescale no longer features AltiVec technology in its latest-generation processors.
Although not new from Intel as such – earlier Penryn processors featured a similar capability – it adds to the attractiveness of the Core i7.
Floating point processor
GE benchmarks on the Penryn floating point processor indicated that its performance compared favourably with that of Freescale’s AltiVec.
For the kind of sophisticated applications that are typical of the military – for example, digital signal processing applications such as imaging, pattern recognition, sensor processing, radar and sonar – a floating point processor provides essential functionality.
Freescale, and the Power Architecture, will continue to be significant players in the military embedded market, not least because of the substantial investment that has been made in it by numerous organisations over many years.
However, Intel’s announcements have unquestionably changed the landscape. The Core i7’s combination of attractive performance/watt, greater functional density, floating point capability and ECC memory, together with the promise of long term support, make it a serious competitor.