GE’s HPEC CoE is Open for Business



How would you define HPEC?

The term HPEC - High Performance Embedded Computing - is a broad one that can encompass a number of paradigms. We at GE Intelligent Platforms use it to describe computing sub-systems in which both hardware and software  are combined to deliver very high compute and/or communication performance that is deployable on air, land or sea vehicles in the harshest environments. These harsh, constrained environments normally dictate the size, weight and power (SWaP)of the systems. Typical applications for HPEC systems are radar, signals intelligence, electronic warfare and image processing.


What was the rationale behind opening the HPEC CoE?

Since 1998, we have been developing multiprocessor products that are aimed at HPEC systems. In 2004, we realised that, to be successful in this arena, you not only need COTS hardware, but also the software and particularly middleware to provide a compelling solution. A new office was opened in Billerica, MA to develop a suite of software products called AXIS to address this.

In order to develop the middleware, we need to understand much more about our customers’ applications. Then, in order to test our software, we found that we needed to create our own HPEC sub-systems. We also found, in the process of helping our customers use the middleware, that we became very tightly coupled during their application integration process - and ended up doing a lot of integration work as a support service.

It became obvious we should offer this value to our customers by providing integrated development systems, hence the HPEC CoE was formed.


How do you see the role of GE’s newly-opened HPEC CoE?

Our key role is to provide our customers with integrated HPEC solutions that greatly reduce their time to deployment. We can help customers define and size their HPEC system to meet their application needs.  Along with this we will offer training and application integration support as well as services such as signal processing algorithm optimization.


How can customers benefit from the HPEC CoE?

Customers can leverage our expertise in our COTS board level products and application knowledge and buy an application-orientated solution. Rather than selling a set of components that then need be integrated and made to work together, we can provide a fully lab-integrated development system, with a path to a fully rugged deployable solution. These systems are pre-configured and tested from a hardware and software perspective. The customer can turn the system on, and start developing their application within minutes.


Are we seeing growing interest from customers in our HPEC offering? For what kind of applications are they considering it?

Yes, since starting this initiative in March this year we have had a number of system design wins, some of which have a roadmap to volume production within the next five years.


What are the key technologies and products that would typically go to make up an HPEC solution?

From a hardware perspective, high performance processing technologies such as the second and third generation Intel Core i7 and Freescale QorIQ processors as well as the massively parallel CUDA enabled GPUs from NVIDIA. These are usually coupled with a high speed fabric interconnect. These are all bound together by middleware such as the OFED software stack, AXIS and open standard DSP and math libraries such as VSIPL and VSIPL++.


What are the technical/technology challenges of HPEC?

The biggest challenge is the rapidly expanding landscape of hardware and software offerings that are relevant to HPEC. When we started developing middleware, the landscape was much narrower;  the processor technology was PowerPC,  the interconnects were PCI and VME and the operating system was VxWorks. Now we have added x86 and GPU processor technologies (and others are emerging), a plethora of interconnects, Windows support and Linux with its many different distributions.

Keeping on top of all this from a software support and test perspective is very challenging.


What do you believe differentiates GE’s approach to HPEC from the approach of other manufacturers?

GE is unique in the breadth of its product offering, with the majority of those products having the potential to play a key role in HPEC systems. For example: many of today’s HPEC systems are designed to capture, process and distribute sensor-derived data, whether radar, sonar or video. GE has families of products designed to capture this sensor data – video tracking systems, for example – to process this data (applying relevant codecs, for example, to minimize bandwidth consumption while maximizing quality), and to deliver the back end processing of the data in the form of powerful multiprocessors that leverage both ‘traditional’ CPU technology as well as leading edge multicore technology from NVIDIA. Add to these capabilities a wide range of single board computers – in 3U and 6U form factors, using VPX, VME and CompactPCI architectures, with a range of processing options from Intel and Freescale; a rapidly-growing range of preconfigured, prevalidated, ready-to-run COTS Rugged Subsystems; and supporting infrastructure solutions such as GE’s battlefield routers, and it’s possible to see how a comprehensive system can be developed and deployed with the minimum of integration problems and support issues that arise from solutions built from the product offerings of a large multiple of vendors. GE’s HPEC CoE is designed to help customers take the most advantage from that product line breadth.