The paper introduces PWR, a scheduling policy that selects power‑efficient GPU and CPU combinations to reduce power consumption in GPU datacenters. It complements the existing Fragmentation Gradient Descent (FGD) policy, which mitigates GPU fragmentation by using a fragmentation metric. Experiments in a simulated cluster show that combining PWR with FGD balances power savings and fragmentation reduction.
The paper proposes a simple technique of chunking workloads into smaller parts that alternate between compute-intensive and memory-bound operations to smooth power and temperature spikes in GPU systems. By doing so, it prevents throttling, leading to faster wall-clock times and lower total energy consumption. Experiments on a DGX Spark show up to 2% performance and energy gains, while similar benefits, though smaller, are observed on multi‑GPU servers.
arXiv:2501.10375v3 Announce Type: replace-cross Abstract: Mixture-of-Experts (MoE) models, though highly effective for various machine learning tasks, face significant deployment challenges on memory...
arXiv:2603. 16428v2 Announce Type: replace-cross Abstract: Fine-tuning Large Language Models (LLMs) has become essential for domain adaptation, but its memory-intensive property exceeds the capabilities of most GPUs.
The paper investigates the effectiveness of GPU power capping during large language model (LLM) autoregressive decoding. It finds that power caps have minimal impact on decoding performance because decode kernels are memory‑bound and the GPU idles between launches, drawing less than half its 700 W rating. By instead locking the SM clock, the authors can reduce power consumption by up to 32 % with negligible throughput loss and identify three architecture‑dependent frequency‑scaling classes that influence energy use.
arXiv:2606. 17104v1 Announce Type: cross Abstract: As large language models (LLMs) are increasingly deployed in latency- and cost-sensitive settings, inference efficiency has become a central systems challenge.
LLM-PowerTrace is an open dataset that provides time‑resolved, per‑GPU power traces for large language model fine‑tuning and inference. The traces are sampled every 100 ms across various NVIDIA platforms and include detailed workload and system metadata. The dataset covers a wide range of model families, sizes, and controlled sweeps over batch size, sequence length, precision, parallelism, checkpointing, concurrency, GPU frequency, KV‑cache and weight quantization, prompt and output lengths, and request‑arrival patterns, and it defines power‑grid‑facing metrics and an AI‑compute power & energy scaling law.
arXiv:2606. 00735v1 Announce Type: cross Abstract: In distributed Mixture-of-Experts (MoE) inference, input-dependent token routing interacts with GPU performance variability to create persistent stragglers under synchronized execution, where the slowest GPU determines layer latency.
arXiv:2602. 24044v2 Announce Type: replace-cross Abstract: Large Language Model (LLM) adapters enable low-cost model specialization, but introduce complex caching and scheduling challenges in distributed serving systems where hundreds of adapters must be hosted concurrently.
arXiv:2608. 10506v1 Announce Type: cross Abstract: Accurate pre-deployment estimation of CNN inference cost--energy, latency, and peak memory--is increasingly critical as models are deployed on resource-constrained GPU platforms.
Accurate pre-deployment estimation of CNN inference cost--energy, latency, and peak memory--is increasingly critical as models are deployed on resource-constrained GPU platforms. Existing approaches rely on FLOPs, latency measurements, or single-device profiling as energy proxies, overlooking the non-linear interactions between architectural design and hardware load.
The paper systematically studies decode‑phase energy consumption of open‑source large language models using different attention architectures—Multi‑Head Attention (MHA), Grouped Query Attention (GQA), and GQA with Sliding Window Attention (SWA). It evaluates four models across varying context lengths, batch sizes, and generation workloads, measuring GPU energy via NVIDIA counters. Findings show that the attention mechanism is the main driver of how energy scales with context length, with MHA models growing steeply, GQA models growing less, and GQA+SWA remaining nearly constant; model size mainly sets absolute energy use, while batching can cut energy per token and latency by up to 87%.