FPGA technology is having a renaissance. New designs like the recently released Intel Stratix 10 offer better performance, 70 percent lower power consumption, up to 1 TB/sec. of memory bandwidth, and up to 10 tera floating point operations per second (FLOPS) of single-precision floating-point performance. It even has its own onboard quad-core ARM processor.
What does FPGA do?
FPGA stands for field-programmable gate array. What makes this technology increasingly relevant in today’s market is its ability to be programmed — and re-programmed — after being deployed in the field. Applications include data centers, the Internet of Things (IoT), high performance computing (HPC), autonomous vehicles and even machine learning.
Here’s an example, courtesy of the FPGA For Dummies eBook. It involves a car’s rear-view camera. Imagine that a manufacturer has put onto the market a car with a rear-view camera that delivers its image to the dashboard screen in 250 milliseconds. Now imagine that the government has changed its regulation to require a maximum delay of only 100 milliseconds. What's the automaker to do?
If the company used a standard processor to control its rear-view camera, it would need to either upgrade or replace the camera. But if it used a connected FPGA, the company could simply send the FPGA new instructions anytime, anywhere. In fact, the FPGA could run a brand-new command set every time it starts up. New instructions could potentially be delivered to the FPGA-enabled camera hardware through a wireless update. This could save millions of dollars in recall costs, while also increasing customer safety and satisfaction.
The flexibility offered by an FPGA isn’t limited to automakers, of course. The technology can also benefit servers, IoT networks and more.
Servers and HPCs can work faster and more efficiently by offloading software-based algorithms and analytics onto FPGA hardware. This can be especially useful for large server arrays tasked with complex video streaming, such as those deployed by Netflix and Amazon. FPGA-equipped servers, when combined with high-performance Xeon processors, can achieve extremely low latency and very precise frame-level synchronization.
FPGA’s bright future
FPGA is one of those older technologies that has made a comeback, similar to the way computer timesharing has been reborn as the cloud. The first FPGAs date back to the 1980s. But those early devices offered only a fraction of the power and functionality available now, and they soon fell out of favor. It was FPGA market leader Altera (acquired by Intel in late 2015) that made the technology relevant again.
In the ensuing years, FPGA architecture has certainly come a long way. In fact, there are now so many possible implementations, engineers are still just scratching the technology’s surface. We can only guess what the future of this technology will offer. But one thing is sure: With a platform this flexible, the possibilities are nearly limitless.
You may also enjoy: