According to the company the architecture can effectively be used for DSP acceleration for key functions.
For FIR (finite impulse response) filters, which are widely used in a large number of commercial and aerospace applications, the nnMAX is able to process up to 1 Giga samples per second with hundreds and even thousands of “taps” or coefficients.
Commenting Cheng Wang, Flex Logix’s senior VP engineering and co-founder, disclosed these benchmarks at the Linley Processor Forum, which was being held online, in a presentation titled “DSP Acceleration using nnMAX.”
“Because nnMAX is so good at accelerating AI inference, customers started asking us if it could also be applied to DSP functions,” said Geoff Tate, CEO and co-founder of Flex Logix. “When we started evaluating their models, we found that it can deliver similar performance to the most expensive Xilinx FPGAs in the same process node (16nm), and is also faster than TI’s highest-performing DSP – but in a much smaller silicon area than both those solutions. nnMAX is available now for 16nm SoC designs and will be available for additional process nodes in 2021.”
NMAX is a general purpose Neural Inferencing Engine that can run any type of NN from simple fully connected DNN to RNN to CNN and can run multiple NNs at a time.
nnMAX is programmed with TensorFlow Lite and ONNX; numerics supported include INT8, INT16 and BFloat16, which can be mixed layer by layer to maximise prediction accuracy.
nnMAX is a tile architecture and any throughput required can be delivered with the right amount of SRAM for a specific model.
