Pimba: A Processing-in-Memory Acceleration for Post-Transformer Large Language Model Serving

• URL: http://arxiv.org/abs/2507.10178v1
• Date: Mon, 14 Jul 2025 11:40:17 GMT
• Title: Pimba: A Processing-in-Memory Acceleration for Post-Transformer Large Language Model Serving
• Authors: Wonung Kim, Yubin Lee, Yoonsung Kim, Jinwoo Hwang, Seongryong Oh, Jiyong Jung, Aziz Huseynov, Woong Gyu Park, Chang Hyun Park, Divya Mahajan, Jongse Park,
• Abstract summary: Transformers are the driving force behind today's Large Language Models (LLMs), serving as the foundation for their performance and versatility.<n>In response, the algorithm community is exploring alternative architectures, such as state space models (SSMs), linear attention, and recurrent neural networks (RNNs)<n>This paper presents a serving system that efficiently supports both transformer and post-transformer LLMs within a unified framework.<n>Our evaluation shows that, compared to LLM-optimized GPU and GPU+PIM systems, Pimba achieves up to 3.2x and 2.1x higher token generation throughput.
• Score: 1.9508863993381267
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Transformers are the driving force behind today's Large Language Models (LLMs), serving as the foundation for their performance and versatility. Yet, their compute and memory costs grow with sequence length, posing scalability challenges for long-context inferencing. In response, the algorithm community is exploring alternative architectures, such as state space models (SSMs), linear attention, and recurrent neural networks (RNNs), which we refer to as post-transformers. This shift presents a key challenge: building a serving system that efficiently supports both transformer and post-transformer LLMs within a unified framework. To address this challenge, we analyze the performance characteristics of transformer and post-transformer LLMs. Despite their algorithmic differences, both are fundamentally limited by memory bandwidth under batched inference due to attention in transformers and state updates in post-transformers. Further analyses suggest two additional insights: (1) state update operations, unlike attention, incur high hardware cost, making per-bank PIM acceleration inefficient, and (2) different low-precision arithmetic methods offer varying accuracy-area tradeoffs, while we identify Microsoft's MX as the Pareto-optimal choice. Building on these insights, we design Pimba as an array of State-update Processing Units (SPUs), each shared between two banks to enable interleaved access to PIM. Each SPU includes a State-update Processing Engine (SPE) that comprises element-wise multipliers and adders using MX-based quantized arithmetic, enabling efficient execution of state update and attention operations. Our evaluation shows that, compared to LLM-optimized GPU and GPU+PIM systems, Pimba achieves up to 3.2x and 2.1x higher token generation throughput, respectively.

Related papers

• Scaling Linear Attention with Sparse State Expansion [58.161410995744596]
  Transformer architecture struggles with long-context scenarios due to quadratic computation and linear memory growth.<n>We introduce a row-sparse update formulation for linear attention by conceptualizing state updating as information classification.<n>Second, we present Sparse State Expansion (SSE) within the sparse framework, which expands the contextual state into multiple partitions.
  arXiv  Detail & Related papers  (2025-07-22T13:27:31Z)
• Echo State Transformer: When chaos brings memory [2.07180164747172]
  We introduce Echo State Transformers (EST), a hybrid architecture for sequential data processing.<n>EST integrates the Transformer attention mechanisms with Reservoir Computing principles to create a fixedsize window distributed memory system.<n>EST achieves constant computational complexity at each processing step, effectively breaking the quadratic scaling problem of standard Transformers.
  arXiv  Detail & Related papers  (2025-06-25T09:56:25Z)
• PaTH Attention: Position Encoding via Accumulating Householder Transformations [56.32365080761523]
  PaTH is a flexible data-dependent position encoding scheme based on accumulated products of Householder transformations.<n>We derive an efficient parallel algorithm for training through exploiting a compact representation of products of Householder matrices.
  arXiv  Detail & Related papers  (2025-05-22T08:36:09Z)
• TransMamba: Flexibly Switching between Transformer and Mamba [43.20757187382281]
  This paper proposes TransMamba, a framework that unifies Transformer and Mamba.<n>We show that TransMamba achieves superior training efficiency and performance compared to baselines.
  arXiv  Detail & Related papers  (2025-03-31T13:26:24Z)
• Sliding Window Attention Training for Efficient Large Language Models [55.56483740523027]
  We introduce SWAT, which enables efficient long-context handling via Sliding Window Attention Training.<n>This paper first attributes the inefficiency of Transformers to the attention sink phenomenon.<n>We replace softmax with the sigmoid function and utilize a balanced ALiBi and Rotary Position Embedding for efficient information compression and retention.
  arXiv  Detail & Related papers  (2025-02-26T05:31:44Z)
• PointMT: Efficient Point Cloud Analysis with Hybrid MLP-Transformer Architecture [46.266960248570086]
  This study tackles the quadratic complexity of the self-attention mechanism by introducing a complexity local attention mechanism for effective feature aggregation. We also introduce a parameter-free channel temperature adaptation mechanism that adaptively adjusts the attention weight distribution in each channel. We show that PointMT achieves performance comparable to state-of-the-art methods while maintaining an optimal balance between performance and accuracy.
  arXiv  Detail & Related papers  (2024-08-10T10:16:03Z)
• Mamba: Linear-Time Sequence Modeling with Selective State Spaces [31.985243136674146]
  Foundation models are almost universally based on the Transformer architecture and its core attention module. We identify that a key weakness of such models is their inability to perform content-based reasoning. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even blocks (Mamba) As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics.
  arXiv  Detail & Related papers  (2023-12-01T18:01:34Z)
• Blockwise Parallel Transformer for Large Context Models [70.97386897478238]
  Blockwise Parallel Transformer (BPT) is a blockwise computation of self-attention and feedforward network fusion to minimize memory costs. By processing longer input sequences while maintaining memory efficiency, BPT enables training sequences 32 times longer than vanilla Transformers and up to 4 times longer than previous memory-efficient methods.
  arXiv  Detail & Related papers  (2023-05-30T19:25:51Z)
• Stable, Fast and Accurate: Kernelized Attention with Relative Positional Encoding [63.539333383965726]
  We propose a novel way to accelerate attention calculation for Transformers with relative positional encoding (RPE) Based upon the observation that relative positional encoding forms a Toeplitz matrix, we mathematically show that kernelized attention with RPE can be calculated efficiently using Fast Fourier Transform (FFT)
  arXiv  Detail & Related papers  (2021-06-23T17:51:26Z)
• Bayesian Transformer Language Models for Speech Recognition [59.235405107295655]
  State-of-the-art neural language models (LMs) represented by Transformers are highly complex. This paper proposes a full Bayesian learning framework for Transformer LM estimation.
  arXiv  Detail & Related papers  (2021-02-09T10:55:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.