# Precision-Scalable Microscaling Datapaths with Optimized Reduction Tree for Efficient NPU Integration

arXiv: https://arxiv.org/abs/2511.06313 · HTML: https://mlsanity.com/paper/2511.06313/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 38 references that are in the map.
- First posted on arXiv 9 Nov 2025.

## Builds on (most cited first, 18 of 18)

- [Microscaling Data Formats for Deep Learning](https://mlsanity.com/paper/2310.10537/) (2023)
- [Ekya: Continuous Learning of Video Analytics Models on Edge Compute Servers](https://mlsanity.com/paper/2012.10557/) (2020)
- [Training LLMs with MXFP4](https://mlsanity.com/paper/2502.20586/) (2025)
- [8-bit Numerical Formats for Deep Neural Networks](https://mlsanity.com/paper/2206.02915/) (2022)
- [Oscillation-Reduced MXFP4 Training for Vision Transformers](https://mlsanity.com/paper/2502.20853/) (2025)
- [On-device Training: A First Overview on Existing Systems](https://mlsanity.com/paper/2212.00824/) (2022)
- [Snitch: A tiny Pseudo Dual-Issue Processor for Area and Energy Efficient Execution of Floating-Point Intensive Workloads](https://mlsanity.com/paper/2002.10143/) (2020)
- [A Precision-Scalable RISC-V DNN Processor with On-Device Learning Capability at the Extreme Edge](https://mlsanity.com/paper/2309.08186/) (2023)
- [Anda: Unlocking Efficient LLM Inference with a Variable-Length Grouped Activation Data Format](https://mlsanity.com/paper/2411.15982/) (2024)
- [Edge Computing and its Application in Robotics: A Survey](https://mlsanity.com/paper/2507.00523/) (2025)
- [DaCapo: Accelerating Continuous Learning in Autonomous Systems for Video Analytics](https://mlsanity.com/paper/2403.14353/) (2024)
- [Edge-Assisted On-Device Model Update for Video Analytics in Adverse Environments](https://mlsanity.com/paper/2308.16413/) (2023)
- [M4BRAM: Mixed-Precision Matrix-Matrix Multiplication in FPGA Block RAMs](https://mlsanity.com/paper/2311.02758/) (2023)
- [Efficient Precision-Scalable Hardware for Microscaling (MX) Processing in Robotics Learning](https://mlsanity.com/paper/2505.22404/) (2025)
- [TaxoNN: A Light-Weight Accelerator for Deep Neural Network Training](https://mlsanity.com/paper/2010.05197/) (2020)
- [An Open-Source HW-SW Co-Development Framework Enabling Efficient Multi-Accelerator Systems](https://mlsanity.com/paper/2508.14582/) (2025)
- [DataMaestro: A Versatile and Efficient Data Streaming Engine Bringing Decoupled Memory Access To Dataflow Accelerators](https://mlsanity.com/paper/2504.14091/) (2025)
- [MXDOTP: A RISC-V ISA Extension for Enabling Microscaling (MX) Floating-Point Dot Products](https://mlsanity.com/paper/2505.13159/) (2025)

## Abstract (arXiv)

Emerging continual learning applications necessitate next-generation neural processing unit (NPU) platforms to support both training and inference operations. The promising Microscaling (MX) standard enables narrow bit-widths for inference and large dynamic ranges for training. However, existing MX multiply-accumulate (MAC) designs face a critical trade-off: integer accumulation requires expensive conversions from narrow floating-point products, while FP32 accumulation suffers from quantization losses and costly normalization. To address these limitations, we propose a hybrid precision-scalable reduction tree for MX MACs that combines the benefits of both approaches, enabling efficient mixed-precision accumulation with controlled accuracy relaxation. Moreover, we integrate an 8x8 array of these MACs into the state-of-the-art (SotA) NPU integration platform, SNAX, to provide efficient control and data transfer to our optimized precision-scalable MX datapath. We evaluate our design both on MAC and system level and compare it to the SotA. Our integrated system achieves an energy efficiency of 657, 1438-1675, and 4065 GOPS/W, respectively, for MXINT8, MXFP8/6, and MXFP4, with a throughput of 64, 256, and 512 GOPS.