(c)iStock.com/Jason Doiy
After a relatively quiet transaction that began on June 1 2015, in December Intel officially acquired the Altera Corporation and reformed it as an internal department dubbed “Intel Programmable Solutions Group.”
This acquisition represents a crucial step in the industry-wide transition towards data centre architectures that allow faster processing and lower latency. Altera’s Field Programmable Gate Arrays (FPGAs) provide a critical building block in the link between data intake and the rapid, efficient processing demanded by virtualisation, analytics, Internet of Things (IoT) sensor arrays and other computing-intensive use cases.
Bringing on the Altera team internally means that Intel can now develop sophisticated processing architectures and ICs that embed FPGA capability within to unlock powerful parallel processing capabilities. Applications and systems vendors who wish to stay on the cutting edge of these developments can read on to learn more about what these advancements may bring and how to leverage the full extent of functionality with a value-added system integration partner like UNICOM Engineering.
Taking a tour of an Altera FPGA
As processing loads place increasing demands on CPUs, companies like Intel look for solutions to maximise simultaneous computing loads, with an aim towards near-parallel processing for intensive operation. Since CPUs are technically restricted to working on one linear process at a time, Intel has invested countless time and capital in R&D to make computations as close to parallel as possible. Hyperthreading and NVDIMM DDR4 RAM processing all pursue this goal, as does the development of GPGPU technology.
FPGAs bypass many of the complications associated with these advances precisely because they are programmable for specific tasks versus general workloads. Each FPGA array consists of an IC containing clusters of logic cells. These cells are configurable in a nearly infinite number of gated combinations, allowing for the streamlining of one specific logic process or the flexibility to reconfigure logic pathways on the fly. Perhaps most importantly, FPGAs can sequester specific clusters of logic cells to allow for true parallel computing.
Combined with Intel’s latest CPU architecture, FPGAs can allow for rapid linking between two CPUs or within virtualised machine systems. Eventually, the hope is that CPUs and FPGAs can be co-packaged on the same IC ecosystem, integrating the two seamlessly through intermittent clusters of logic cells spread throughout the CPU and local IC.
When fully realised, this technology has the potential for rapid virtualization of multiple VMs through one CPU while simultaneously running Network Function Virtualization (NFV)which is touted as the future of network security and stability.
Put simply, the dynamic capabilities that FPGA technology contains could help reinvent the processor and, subsequently, the data centre as we know it.
Increasing speed to market while decreasing R&D investment
Making the most out of FPGA/CPU combinations will not be an easy technological hurdle for the average applications developer to clear. Once they have a stable build of their intended product, they will need to build-test and modify it to make best use of all the new levels of computing functionality at their disposal.
Since this secondary development period introduces an excessive amount of time and risk, developers could instead build on Intel through the expertise of UNICOM Engineering. Read our recent blog post to learn more about how UNICOM Engineering can help induct you into the emerging world of ultra-low-latency computing.