HARDWARE ACCELERATION
Over the past decade, big data has been adopted by several industries in different fields (to name a few, biology, finance, high-tech, physics, etc.) as the key to better understanding their environment. But big data alone is not sufficient, it must first be transformed to extract relevant and decision critical information. This is done through a process called data processing.
​
In many data processing applications, getting the information as fast as possible is a critical criterion. For those applications where time is of the essence, hardware acceleration using FPGAs becomes of great interest!
WHAT IS AN FPGA?
A field-programmable-gate-array (FPGA) is a device made of different hardware components but most notably it contains logic blocks, block rams, interconnects, DSP slices, etc. All these components can be programmed using a hardware description language (HDL) to perform a specific and dedicated processing task, yielding unparalleled processing capabilities. In fact, because they process data at the hardware level, FPGAs are extremely fast when compared to other processing engines.
Hardware acceleration is the process of using an FPGA to design an extremely fast data processing pipeline. The motivation being that most mathematical models can be executed much faster directly on hardware. This is typically done by converting parts of the algorithm into a parallel architecture. Over the years, FPGAs systems have become as important for real-time embedded systems AT THE EDGE as they are for intense cloud computing applications.
​
Thi figure in this section shows how a hardware accelerator can be integrated into your application using a common shared memory architecture.
​
WHAT IS HARDWARE ACCELERATION?
CASES STUDY
BIOLOGY
--
MACHINE LEARNING WITH RANDOM FOREST CLASSIFIER
FPGA is
Up to 10 times* faster than CPU
Up to 63 times* faster than GPU
*IEICE published study
TECH
--
LIVE VIDEO COMPRESSION & TRANSCODING
FPGA is
Up to 32 times* superior to CPU implementation
*Twitch case study
FINANCE & INSURANCE
--
MONTE CARLO METHODS
At same CPU throughput, FPGA saves you money by reducing power consumption by up to 96%*
​
​
*IEEE published study
PHYSICS
--
REAL-TIME IMPLEMENTATION OF EXTENDED KALMAN FILTER
FPGA is
Up to 20 times* faster than solo real-time microcontroller
implementation
​
* IEEE published study
HOW CAN
SIX METRIX
HELP YOU?
Designing a hardware accelerator requires both deep knowledges of FPGA systems architectures/design tools and deep knowledge of how processing algorithms translate into hardware. This is exactly the expertise brought by Six Metrix's professional engineers!
Contact us to learn more about how we can help you!