innovation_header

The future of innovation does not look like its past. Huge growth in data analysis requires a new path, to apply machine learning and accelerate results.  We are that KNUPATH.


The first implementation of this approach is the KNUPATH Hermosa processor.  To demonstrate the performance superiority of a system running with KNUPATH Hermosa processors on LambdaFabric, we have 2 videos running an image recognition algorithm.  The left hand video is demonstrating GPU performance based upon technology from an established industry vendor, while the right hand video is running on a system using KNUPATH Hermosa processors with LambdaFabric technology.  As demonstrated below, the KNUPATH solution will perform approximately 2.5 times faster than the alternative in real world environments.

If you’d like to speak with us about how you could get your hands on this technology for yourself, drop us a line at info@knupath.com.

Video on the left shows GPU performance while the video on the right is running with KNUPATH Hermosa processors with LambdaFabric technology


KNUPATH, and its parent company KnuEdge, are in a market that is undergoing dramatic upheaval.  For example, worldwide data growth continues to accelerate: data production is forecast to grow by 4300% by 2020. Mobile data traffic alone is forecast to reach 173 million terabytes by 2018 (IDC).  That much data will require a faster way to make it useful; with a tidal wave of data coming in, how do you quickly find and understand new trends, connect it to other trends, and know what actions to take?

The super scalability and advanced computation of our KNUPATH Hermosa processors with LambdaFabric™ speeds evaluation of expanded scenarios within limited time windows. This enables accelerated insight and highly accurate, information-based decisions.

neurons

Nature’s Design Principles

  • Communication occurs via synapses
  • The world’s first multi-tasking parallel machine
  • Context is contained within the neuron body
  • Axonal connections define geometry
  • The most power-efficient computational engine
purple_circuits

KNUPATH’s Design Principles

  • Communications-driven router-based architecture
  • Scalable massive parallelism
  • Proximity of memory and processing
  • Processing adapts to application not vice versa
  • Outstanding performance per watt

Steve Cumings, KNUPATH’s Senior Director of Market Development discusses KNUPATH’s innovation approach to machine learning architectures:

 

KNUPATH products are optimized for the a variety of highly parallel environments and workloads, including but not limited to the following types of processing:


High Performance Computing
High Performance Computing has been a staple in research for years, and more recently it has entered into the commercial world as businesses rely more on specialized applications to make their decisions and lead their business. The extreme efficiency of our products are perfectly matched with these workloads, enabling greater performance within data center power constraints.


Speech Processing
While many tools exist today that are built on voice recognition, the next advance is voice authentication. Authentication doesn’t focus on what is said, but on who said it. This verification requires far more specific computing in order to deliver the accuracy in critical situations.


Pattern Recognition
Some problems are the classic needle in a haystack – if you know what you are looking for, it is easier to solve and deliver results. But what about finding the needle in a stack of needles? This requires a different approach and a new technology.


Problem Solving
In our changing world, the hardest problem to solve is the one that we don’t fully understand. Through our research on the human brain and exclusive IP, we’re helping create the machine learning systems to bring the world new solutions and solve the complex problems of today and tomorrow.