[2605.19775] Understanding Inference Scaling for LLMs: Bottlenecks, Trade-offs, and Performance Principles

-->

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2605.19775 (cs)

[Submitted on 19 May 2026]

Title:Understanding Inference Scaling for LLMs: Bottlenecks, Trade-offs, and Performance Principles

Authors:Moiz Arif, Avinash Maurya, Sudharshan Vazhkudai, Bogdan Nicolae<br>View a PDF of the paper titled Understanding Inference Scaling for LLMs: Bottlenecks, Trade-offs, and Performance Principles, by Moiz Arif and 3 other authors

View PDF<br>HTML (experimental)

Abstract:The transition from standard generative AI to \emph{reasoning-centric architectures}, exemplified by models capable of extensive Chain-of-Thought~(CoT) processing, marks a fundamental paradigm shift in system requirements. Unlike traditional workloads dominated by compute-bound prefill, reasoning workloads generate long chains of reasoning tokens that shift inference into a \emph{Capacity-Bound regime}. This paper presents a comprehensive system characterization, evaluating models ranging from 8B to 671B parameters on GPUs clusters. By systematically exploring the interplay between Data, Tensor, and Pipeline parallelism, we identify critical bottlenecks that defy standard scaling heuristics. Our analysis reveals that data parallelism is throughput efficient for small models but hits a capacity trap on reasoning workloads as KV-cache fragmentation forces early throttling resulting in sub-optimal compute utilization. Tensor parallelism unlocks stranded memory and delivers sublinear gains near the 32B crossover. At frontier scale, dense models (e.g., Llama-405B) are interconnect and memory-bandwidth bound and favor high-degree TP, while sparse Mixture-of-Experts (MoE) models (e.g., DeepSeek-R1) are limited by routing and synchronization latency and benefit from hybrid strategies. These insights provide a rigorous decision framework for navigating the reasoning cliff, establishing new architectural imperatives for the next generation of inference infrastructure.

Comments:<br>ISCA'26: The 53rd International Symposium on Computer Architecture, Industry Track

Subjects:

Distributed, Parallel, and Cluster Computing (cs.DC); Performance (cs.PF)

Cite as:<br>arXiv:2605.19775 [cs.DC]

(or<br>arXiv:2605.19775v1 [cs.DC] for this version)

https://doi.org/10.48550/arXiv.2605.19775

Focus to learn more

arXiv-issued DOI via DataCite (pending registration)

Submission history<br>From: Avinash Maurya [view email]<br>[v1]<br>Tue, 19 May 2026 12:43:51 UTC (10,743 KB)

Full-text links:<br>Access Paper:

View a PDF of the paper titled Understanding Inference Scaling for LLMs: Bottlenecks, Trade-offs, and Performance Principles, by Moiz Arif and 3 other authors<br>View PDF<br>HTML (experimental)<br>TeX Source

view license

Current browse context:

cs.DC

next >

new<br>recent<br>| 2026-05

Change to browse by:

cs<br>cs.PF

References & Citations

NASA ADS<br>Google Scholar

Semantic Scholar

export BibTeX citation<br>Loading...

BibTeX formatted citation

&times;

loading...

Data provided by:

Bookmark

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

Which authors of this paper are endorsers? |<br>Disable MathJax (What is MathJax?)