Transforming AI/ML Inference

Take Advantage of Smarter Approaches to AI Acceleration

Artificial Intelligence of Digital Human Brain

The New Wave of Inference Acceleration

A lot has changed in a few years with AI/ML models and inference. The hardware that’s great at crunching training algorithms can fall behind in latency and utilization for real-time, batch size 1 inference. Deep learning models have gotten more complex, requiring new approaches for real-time applications to keep up.

The good new is, as machine learning has matured, acceleration technology has gotten smarter—and more efficient. These are both on a silicon level, such as using dedicated ASIC devices, and in a design approach—like using 8-bit integers for connections.

BittWare, a brand that’s been trusted for over thirty years to bring the best acceleration technology to market, has assembled an ecosystem FPGA- and ASIC-based AI solutions that are optimized for inference.

Whether it’s scaling up a CPU- or GPU-based system to the latest datacenter-grade tensor processor or getting every last watt of performance using an edge-focused solution, we have what you need to reduce risk and get to market faster.

BittWare + ML/AI Inference

When to Engage with BittWare

  • You have a trained model and you seek help deploying it
  • We can help with your deployment proof-of-concept
  • If PCIe cards won’t work for you, we can discuss something custom
  • We can talk about your options for getting real-world data into your inference model

Deployment Targets

Where do BittWare AI/ML customers deploy?

  • “Deployment at the edge” covers everything from a surveillance camera to a telephone pole to a closet inside an office building
  • We can help deploying special-purpose devices inside data centers (our partners generally have the traditional servers inside data centers well covered)

Development Platform

For some partners we are also supplying development platforms used to evaluate their technology

Solutions Partners

Our Partner Program ecosystem includes a range of AI/ML inference options, from next-gen ASIC-based cards to IP for development on FPGAs.

AI/ML Articles

FPGA Neural Networks

We look at the inference of neural networks on FPGA devices, illustrating their strengths and weaknesses.

Acceleration of BWNNs

Programming Stratix 10 using OpenCL for machine learning. Topics Covered: OpenCL, machine learning, Stratix 10.

CNN thumbnail

CNN Acceleration

Using variable precision in FPGAs to build better machine learning inference networks. Topics Covered: machine learning, application tailoring, Arria 10.