Upping Our Game

20 January 2016
US Navy: S-3B Viking

Well, it’s been a hectic couple of months, with our departure from the GE fold and the holidays. Given all that, it strikes me as remarkable that we have not missed a beat in bringing new technologies to the marketplace.

Just this week, we announced the arrival of our latest multiprocessor platform, the DSP282. This features two Intel Core i7-5700EQ CPUs belonging to the family known as Broadwell.

This provides a great technology insertion point for users of our DSP280 board as well as for new designs, upping performance from 260 to 666 GFLOPS peak (core only), PCIe connections from Gen2 to Gen3, and data plane connections from DDR InfiniBand and 10Gigabit Ethernet to FDR-10 and 40G respectively.

The compute performance boost comes from an increased base clock frequency thanks to Intel’s process technology moving from 32nm to 14nm—with the power savings that allows—plus the availability of the AVX2 instruction set and dual MAC units.

There’s a pretty nice bump in graphics performance—whether that be for driving displays or for GPGPU computation, with the inclusion of HD 5600 graphics.

Security gets some loving too. The chipset takes the Trusted Execution of Sandy Bridge, and adds OS Guard (helps protect the operating system from escalation-of-privilege attacks) and IPT (which allows two-factor authentication to protect information). In addition, we have included protection technology in the board design in the form of a programmable device that can monitor the environment and execution to detect anomalous behavior that could be indicative of an attempt to breach the security of the board, and to then take appropriate action to defeat the threat.

Sweet

Pretty sweet for a pin-compatible upgrade—but what of software?

Well, as always, our guys have done their usual stellar job of smoothing that bump out—from the board support packages to compatible versions of math libraries (AXISLib-AVX 2.0), and inter-processor communications (AXISFlow).

We also have a recent addition to the family—AXIS MPI. In a prior blog, Not All Open Architecture Middleware Is Created Equal: Life of MPI, I outlined the reasons why embedded developers might benefit from an MPI that is tailored to embedded systems rather than supercomputers. Well, coincident with the release of DSP282, we have announced that we have unbundled our MPI, and it is now available as a standalone product for those who wish to gain the performance improvements it provides without using our other tools. 

The integration of AXIS MPI with our performance monitoring and visualization tools still provides a compelling match that we believe adds value without breaking the basic rationale for using open architecture software, but it’s now a choice, and choices are always good for customers.

As Chris Lever, our VP of product management, said when we announced AXIS MPI: “Bringing new programs to deployment is not just about hardware—it’s about the software tools available for the hardware that can simplify and speed the development process. That’s what AXIS and AXIS MPI are designed to do. The combination of our industry-leading hardware with our unique software capability enables the development of higher performance, lower SWaP solutions with less risk, cost and time than would otherwise be possible.”

If you want to explore how Abaco’s cutting edge hardware and software can help your tough embedded computing problems, drop me a line or reach out to your local Abaco salesperson.
 

Peter Thompson

Peter Thompson is Vice President, Product Management at Abaco Systems. He first started working on High Performance Embedded Computing systems when a 1 MFLOP machine was enough to give him a hernia while carrying it from the parking lot to a customer’s lab. He is now very happy to have 27,000 times more compute power in his phone, which weighs considerably less.