X
Tech

Academics develop super-fast 1,000 core FPGA chip

Academics have created a field programmable gate array chip with the functionality of 1,000 distinct cores for super-fast image processing.On Tuesday Glasgow University announced that academics led by Glasgow's Dr Wim Vanderbauwhede and colleagues at the University of Massachusetts, Lowell had created a field programmable gate array (FPGA) with the functionality of 1,000 separate cores.
Written by Jack Clark, Contributor

Academics have created a field programmable gate array chip with the functionality of 1,000 distinct cores for super-fast image processing.

On Tuesday Glasgow University announced that academics led by Glasgow's Dr Wim Vanderbauwhede and colleagues at the University of Massachusetts, Lowell had created a field programmable gate array (FPGA) with the functionality of 1,000 separate cores.

"FPGAs are not used within standard computers because they are fairly difficult to program, but their processing power is huge while their energy consumption is very small because they are so much quicker — so they are also a greener option," Vanderbauwhede said in a statement accompanying the release.

A FPGA is a chip consisting of an array of individual hardware logic cells that can be configured in software by the end user. FPGAs have a parallel architecture and their usage areas include code breaking, such as in the Cost-Optimised Parallel Code Breaker (COPACOBANA) device and data encoding and decoding.

The chip processed an algorithm — a Discrete Cosine Transform — central to the encoding of MPEG video at 5Gbps, 20 times faster than top end desktop computers, Vanderbauwhede told ZDNet UK on Wednesday.

"What I'm doing is finding more efficient ways of programming FPGAs, because for a lot of applications they are more efficient — they can do more work for the same clock cycles," Vanderbauwhede said.

The 1,000 cores were made by creating a network of 1,000 custom processor cores out of a single FPGA. Each core had its own dedicated batch of memory — known as processor-in-memory (PIM) — to speed calculations, Vanderbauwhede said.

"My driver for this is to show that you can use these devices in datacentres", Vanderbauwhede told ZDNet UK. "If you would deploy them in the near future they would be as an extension of an existing system to offload computationally intensive parts," he said.

"Basically, the FPGA used in this work consumes only something like 5 watts, whereas a typical high-end processor consumes around 100-200 watt, so you get a huge gain in power consumption" he said.

Intel demonstrated a variant of its Atom E600 processor named Stellarton that included an on-chip FPGA at the Intel Developer Forum in September.

Intel has said that a flexible 1,000 core x86-based processor is possible, but the challenge is developing programming frameworks for 1,000 cores that will be usable by a general-purpose programmer.

Editorial standards