Flex Logix has launched their new NMAX Neural Inferencing Engine, delivering 1 to >100 TOPS of neural inferencing capacity in a modular, scalable architecture requiring a fraction of the DRAM bandwidth of existing neural inferencing solutions.
The new NMAX neural inferencing engine employs the same eFPGA interconnect technology, but is mainly targeted toward the neural inferencing portion of the AI industry. It can control simple NNs to fully connected DNN to RNN to CNN and can manage many NNs simultaneously. NMAX is programmed using Tensorflow and will be able support other model description languages in the future.
Flex Logix has developed this technology for inferencing because eFPGA permits re-configurable, fast control logic for each phase of the network.
“The difficult challenge in neural network inferencing is minimizing data movement and energy consumption, which is something our interconnect technology can do amazingly well,” said Geoff Tate. “While performance is key in inferencing, what sets NMAX apart is how it handles this data movement while using a fraction of the DRAM bandwidth that other inferencing solutions require. This dramatically cuts the power and cost of these solutions, which is something customers in edge applications require for volume deployment.”
NMAX is currently in development and will be available in the second half of 2019. Prospective buyers may go to www.flex-logix.com to view the slideshow presented at the Linley Processor Conference and/or contact email@example.com for further details of NMAX under NDA.
For more information, visit http://www.flex-logix.com.