Balancing speed energy and accuracy for AI models on edge devices
A Measurement Study of LLM Inference Trade-offs Across Edge Continuum Hardware
Distributed, Parallel, and Cluster ComputingArtificial IntelligenceMachine Learning
Summary
Running large language AI models on devices near users, like phones or local servers, means deciding how to balance speed, energy use, and answer accuracy. This study measured how different hardware setups like a small GPU device, a CPU-only server, and a GPU-enabled server perform when answering questions. The results show that GPUs are fastest but can use more power, while the small device uses less energy but is slower. Just looking at model size or parameter count does not fully explain performance. The authors also found that focusing only on how fast the AI runs can lead to bad choices when responses are sent in small chunks over the network.
large language modelsedge computinginference latencymodel quantizationGPUCPUenergy consumptionPareto frontiertensor decodingmodel footprint
Authors
Maysam Khatib, Moysis Symeonides, Demetris Trihinas, George Pallis, Marios D. Dikaiakos
Abstract
Large language models (LLMs) are increasingly used as backends for intelligent web services, but serving them across the edge continuum requires balancing quality, latency, model footprint, and energy. This paper presents a controlled measurement study of self-hosted LLM inference across edge and near-edge deployment nodes: an NVIDIA Jetson AGX Orin and a near-edge server with CPU-only and GPU-enabled inference modes. We evaluate multiple open-weight LLMs and quantization variants using a fixed question-answering workload, and compare them against GPT-4o as a cloud-hosted accuracy and latency reference. Our benchmarking pipeline reports accuracy, model footprint, per-token decoding latency, prefill latency, and overall execution energy. The results show that GPU-enabled server execution provides the lowest compute-side latency, while Jetson Orin shows lower measured energy, consistent with its lower platform power under our setup. CPU-only execution is consistently dominated in latency for our workload and shows higher measured energy. We also show that parameter count and downloaded weight-file size alone do not reliably predict observed accuracy or latency. Finally, using Pareto-frontier analysis, we study how deployment decisions may change under possible streamed-token delivery overheads, highlighting that compute-side inference metrics alone can lead to suboptimal placement for latency-sensitive interactive web services.