Jan Gray on Twitter: "#FPGA @IntelFPGA Stratix 10 NX: https://t.co/EgHBU1em3t https://t.co/AZBfLYlOSg (30 mul + 30 add) per AI Tensor Block x 3960 blocks per NX 2100 device = "up to 286 INT4/block
![A Block-Floating-Point Arithmetic Based FPGA Accelerator for Convolutional Neural Networks | Semantic Scholar A Block-Floating-Point Arithmetic Based FPGA Accelerator for Convolutional Neural Networks | Semantic Scholar](https://d3i71xaburhd42.cloudfront.net/e2cc24b5a97db672baf3ebfce30d93f8d2c15223/3-Figure2-1.png)
A Block-Floating-Point Arithmetic Based FPGA Accelerator for Convolutional Neural Networks | Semantic Scholar
![US20130054661A1 - Block floating point compression with exponent difference and mantissa coding - Google Patents US20130054661A1 - Block floating point compression with exponent difference and mantissa coding - Google Patents](https://patentimages.storage.googleapis.com/e2/4c/64/f5bd60ca43b984/US20130054661A1-20130228-D00000.png)
US20130054661A1 - Block floating point compression with exponent difference and mantissa coding - Google Patents
![Figure 3 from A Block Floating Point Implementation for an N-Point FFT on the TMS 320 C 55 x DSP | Semantic Scholar Figure 3 from A Block Floating Point Implementation for an N-Point FFT on the TMS 320 C 55 x DSP | Semantic Scholar](https://d3i71xaburhd42.cloudfront.net/701ca0b5e9b7dced07ce6142f820c8692ff3a815/4-Figure3-1.png)
Figure 3 from A Block Floating Point Implementation for an N-Point FFT on the TMS 320 C 55 x DSP | Semantic Scholar
![Figure 2 from Computation Error Analysis of Block Floating Point Arithmetic Oriented Convolution Neural Network Accelerator Design | Semantic Scholar Figure 2 from Computation Error Analysis of Block Floating Point Arithmetic Oriented Convolution Neural Network Accelerator Design | Semantic Scholar](https://d3i71xaburhd42.cloudfront.net/2edce0cfc10032601e9db688a64ec31dad1b0897/5-Figure2-1.png)
Figure 2 from Computation Error Analysis of Block Floating Point Arithmetic Oriented Convolution Neural Network Accelerator Design | Semantic Scholar
![Learn About the New Block Floating Point Arithmetic Unit for Processing AI/ML Workloads in Speedster7t FPGA | Achronix Semiconductor Corporation Learn About the New Block Floating Point Arithmetic Unit for Processing AI/ML Workloads in Speedster7t FPGA | Achronix Semiconductor Corporation](https://i.ytimg.com/vi/pjYtyN4eVfQ/maxresdefault.jpg)