Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Aging-aware CPU Core Management for Embodied Carbon Amortization in Cloud LLM Inference

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Aging-aware core management—deep-idling unused CPU cores and routing tasks to least-aged ones—cuts yearly embodied carbon in simulated LLM clusters by 37.67%.

desk verdict The scheduling idea is plausible and the implementation is transparent, but the headline 37.67% carbon saving rests on an unvalidated linear lifespan model that the paper never derives or justifies. read the letter →

arxiv 2501.15829 v1 pith:Y5QJGOJS submitted 2025-01-27 cs.DC

Tharindu B. Hewage , Shashikant Ilager , Maria Rodriguez Read , Rajkumar Buyya This is my paper · ORCID
classification cs.DC
keywords embodiedcarbonCPUagingNBTILLMinferencecoreidlingtaskschedulingamortizationcloudsustainability
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that cloud LLM inference clusters can cut embodied carbon without new hardware by managing CPU aging in software. Using production traces, the authors find inference CPUs are mostly idle with bursts, so they design a core manager that deep-idles unused cores—halting NBTI-induced aging—and routes tasks to the least-aged active cores. In simulations, the technique reduces estimated yearly CPU embodied carbon by 37.67% (p99) and 49.01% (p50), slashes CPU underutilization by 77%, and keeps oversubscription below 10%. If true, it gives operators a software-only way to lengthen the CPU refresh cycle, converting wasted idle capacity into carbon savings.

What carries the argument

The central mechanism is a two-part core manager built on an NBTI aging model. The aging model computes each core's frequency as $f(t)=f_0(1-\Delta V_{th}/(V_{dd}-V_{th}))$ with $\Delta V_{th}$ updated recursively, using process-variation-sampled initial frequencies. Algorithm 1 (Task-to-Core Mapping) maintains the last eight idle durations per core, scores each free core by the sum of those durations, and assigns the new inference task to the highest-scoring (least-aged) core. Algorithm 2 (Selective Core Idling) computes a normalized error between active tasks and available cores, passes it through an asymmetric reaction function—slower for underutilization, faster for oversubscription—and moves that many cores between deep idle and active, choosing most-aged cores to idle and least-aged cores to wake. Deep idle (C6-style power gating) is what halts aging; the reaction function is what keeps the working set sized to demand.

What would settle it

Measure actual threshold-voltage drift ($\Delta V_{th}$) or frequency degradation on a server CPU whose cores have been deep-idled (C6) versus continuously active over a one-year accelerated-aging run; if the deep-idled cores degrade at more than a small fraction of the active cores' rate, the age-halting premise fails. Alternatively, check published NBTI lifetime data to see whether, say, a 37.67% reduction in mean frequency degradation over the first years corresponds to a 37.67% extension of usable life, or a different percentage.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that CPU aging in LLM inference servers can be actively slowed by a pair of online algorithms. A selective-core-idling algorithm dynamically shrinks the active working set to match current inference demand, putting surplus cores into deep idle so their transistors stop switching and NBTI stress stops accumulating. A task-to-core-mapping algorithm then estimates each core's age from a rolling window of its idle durations and assigns new inference tasks to the least-aged idle core. Together these even out aging across cores and delay mean frequency degradation. The estimated consequence is a longer usable CPU life, which amortizes the CPU's embodied carbon over more years: in the authors' simulated 22-server cluster, 37.67% lower yearly embodied carbon at p99 mean-frequency performance, rising to 49.01% at p50.

Load-bearing premise

The headline carbon saving rests on a linear model that turns reduced mean frequency degradation into an equal percentage of extra CPU lifespan; if frequency loss does not translate linearly into hardware lifetime, the 37.67% saving is not supported.

Editorial extensions

If this is right

  • If the results transfer to hardware, an inference cluster can extend its CPU refresh cycle from a typical 3 years to roughly 4–5 years at the same service quality, cutting yearly CPU embodied carbon by about a third to a half.
  • The 77% reduction in CPU underutilization means previously wasted capacity now contributes to longer hardware life rather than to idle power draw.
  • Oversubscription below 10% (p1 at or above −0.1) bounds the service-quality cost, so the technique is compatible with latency-sensitive LLM serving.
  • Because the technique works at both 40- and 80-core VM sizes and across throughput levels, it is a candidate for fleet-wide deployment.
  • As grids decarbonize and operational carbon falls, embodied carbon becomes the dominant share of inference-cluster footprint; extending CPU life directly targets that dominant share.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The 37.67% figure inherits the paper's linear assumption that a 37.67% reduction in mean frequency degradation translates into a 37.67% longer CPU life; real NBTI data may make the relation superlinear or sublinear, so the true carbon saving could differ.
  • Deep idle is assumed to fully halt aging, but leakage current, thermal cycling, or power-gating transitions themselves may continue to stress transistors; a hardware study measuring $\Delta V_{th}$ in cores cycled between C0 and C6 would test this.
  • The same selective-idling pattern could be applied to other underutilized server components (memory controllers, NICs, GPU idle states) to amortize their embodied carbon as well.
  • The idle-score heuristic approximates age with a window of eight idle durations; direct core-aging telemetry, if available, could sharpen the even-out behavior and reduce the oversubscription tail.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes an aging-aware CPU core management technique for cloud LLM inference clusters, combining (i) a least-aged-first task-to-core mapping algorithm and (ii) a selective core idling mechanism that deep-idles underutilized cores while a reaction function controls the active working set. The authors extend the splitwise-sim simulator with an NBTI aging model, process-variation sampling, and a temperature model, and evaluate against Linux-default and least-aged baselines on Azure inference traces. They report an estimated 37.67% reduction in yearly CPU embodied carbon emissions (49.01% at p50), a 77% reduction in CPU underutilization, and CPU oversubscription below 10%.

Significance. The paper addresses a timely problem—embodied carbon in LLM inference clusters—and the algorithmic contribution is well motivated by the observed CPU underutilization patterns. The design is clearly specified, the implementation is open source, and the evaluation uses realistic Azure traces and a widely used simulator. However, the headline quantitative carbon savings rest on an unvalidated linear model that converts observed frequency degradation into lifecycle extension, so the significance of the specific carbon numbers is provisional. If the lifetime-extrapolation step can be justified or replaced with a validated model, the work would constitute a useful step toward sustainable LLM serving.

major comments (4)
  1. [Section 6.2] The central carbon claim (37.67% / 49.01% reduction in yearly embodied carbon) is computed from an unvalidated linear model. The text states: 'We then compare the reduction of the mean core frequency of other techniques to linux and estimate an increase in lifecycle extension using a linear model.' No functional form, derivation, or validation is given for this linear mapping, and the simulated trace duration is not reported. NBTI degradation is known to be sublinear in time (typically a power law with exponent around 1/6 to 1/4), so a linear extrapolation from short simulated frequency differences to a multi-year hardware refresh cycle can substantially inflate the carbon benefit. Please replace this with a physically justified or empirically supported lifetime model, or present the carbon results as a sensitivity analysis over a range of plausible lifetime-extension mappings, and state the simulation duration so the extrapolation factor is transparent.
  2. [Section 6.2 and Section 6.1.3] The metric 'p99 mean frequency performance' is not precisely defined. Section 6.1.3 defines coefficient of variation (CV) percentiles across cluster servers, but the carbon calculation uses mean core frequency degradation. It is unclear whether p99 is computed across servers of each server's mean degradation, across cores, or across repeated experiments. If p99 selects different servers for different techniques, the ratio of p99 values can be biased by server selection rather than reflecting overall aging management. Please define the exact computation of the 'mean frequency performance' percentiles and report the full distribution (e.g., all percentiles) or per-server values rather than only p99 and p50.
  3. [Section 3.2] The aging model is calibrated to a single external datapoint: a 22nm CPU with a worst-case 30% frequency reduction over 10 years, from reference [1]. The value of the fitting parameter K is solved to match this scenario, and all unspecified parameters are inherited from that 22nm prior work. The simulated cluster, however, is intended to represent modern LLM inference servers (H100-class), for which the technology node and aging characteristics may differ substantially. The paper should justify transferring the 22nm calibration to the simulated CPUs and should report sensitivity of the headline carbon savings to the calibration point (30%/10yr) and to the process-variation parameters N_chip and alpha. Without this, the quantitative results are conditional on a single historical datapoint.
  4. [Section 3.1] The model assumes that deep idle completely halts aging: 'Being in the active state gradually ages the CPU cores. In contrast, deep idling halts cores from aging.' This assumption is load-bearing because selective core idling is the main source of the proposed technique's advantage over the least-aged baseline. In real hardware, deep-idled cores may still experience degradation through leakage, thermal cycling, and other mechanisms, and the aging benefit of C6 may be less than total. Please discuss this limitation explicitly and, ideally, provide a robustness test with a partial-aging-in-idle model to show how the carbon savings would change.
minor comments (6)
  1. [Section 3.2] In the recursive NBTI equation, the text refers to 'Vth(t_p)' while the equation uses ΔVth; please clarify the notation and define the exponent n (e.g., the recovery/stress exponent) explicitly.
  2. [Section 6.2] The abstract and Section 6.2 state '77% reduction in CPU underutilization,' but the supporting text in the application-impact paragraph reports that the p90 percentile of the idle-core distribution is 'at least 77.8% better.' This is a p90 improvement, not a reduction in overall underutilization; please qualify the claim accordingly.
  3. [Abstract and Section 6.2] The abstract claims 'less than 10% impact to the inference service quality,' while the actual reported result is that CPU oversubscription is maintained below 10%. CPU oversubscription is an intermediate metric, not a direct measure of service quality (e.g., latency). Please reword to avoid overstating the application impact.
  4. [Section 6.1.2] There is a typo: 'the number of output tokes generated' should read 'output tokens generated.'
  5. [Introduction] The claim that Microsoft reported operational carbon reduced by 6.3% and embodied increased by 30.9% is cited to reference [4], which is a blog about building datacenters with wood. This citation appears mismatched; please cite the original Microsoft sustainability report or a more precise source.
  6. [Section 4.2] The reaction function uses constants 0.785 and 1.55 with no derivation or sensitivity analysis. Since these parameters shape the working-set adjustment dynamics and indirectly affect aging and oversubscription, a brief sensitivity study or a discussion of how they were chosen would strengthen the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the aging simulation uses external NBTI and process-variation models plus Azure traces; the lifecycle-extension conversion is an assumption, not a construction.

full rationale

The derivation chain is not circular. The NBTI aging model (Section 3.2) is calibrated to external 22nm data from prior work and process-variation parameters from [28], and the workload is replayed from Azure traces through the external splitwise-sim simulator. The proposed technique (Algorithms 1 and 2) is evaluated against independent baselines (linux and least-aged [42]) using metrics computed from the simulator state. The only fragile step is the conversion from simulated frequency degradation to CPU lifetime extension, stated as 'We then compare the reduction of the mean core frequency of other techniques to linux and estimate an increase in lifecycle extension using a linear model' (Section 6.2). This is an unvalidated modeling assumption and a correctness risk, but it is not circular: the 37.67% carbon figure does not reduce by construction to a fitted parameter, and no load-bearing self-citation is involved. Under the stated rules, an unsupported linear extrapolation without construction-equivalence to the inputs is a correctness issue rather than circularity.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

No new physical entities are introduced. The central result relies on an NBTI aging model calibrated to a single prior datapoint, a process-variation model with hand-set parameters, a hand-tuned reaction function, and a linear lifespan-extension assumption. The most consequential assumption is the linear mapping from frequency degradation to CPU lifetime, which directly produces the headline embodied carbon reduction.

free parameters (6)
  • K (NBTI ADF fitting parameter) = not reported; set so model matches 22nm 10-year 30% worst-case degradation from [1]
    Equation 2; calibrated to prior literature rather than to the present experiment, and it affects all aging trajectories in the simulation.
  • K' (frequency model constant) = 1
    Hand-chosen technology constant in the f0 model of Section 3.2; influences the sampled initial core frequencies.
  • N_chip (process variation grid size) = 10
    Hand-chosen resolution for the spatial process-variation grid in Section 3.2; impacts the variance of initial core frequencies.
  • alpha (spatial correlation decay) = set from [28]
    Borrowed from prior work; controls how correlated process variations are across grid cells in the f0 model.
  • Reaction function coefficients = 0.785 for idle, 1.55 for oversubscription
    Hand-chosen in Algorithm 2 to react slower to underutilization and faster to oversubscription; changes working-set dynamics.
  • Idle history window size = 8
    Rolling window of the last eight idle durations, borrowed from the Linux cpuidle governor; affects age estimates in Algorithm 1.
assumptions (6)
  • domain assumption NBTI reaction-diffusion model (Equations 1 and 2) adequately captures CPU aging and core frequency degradation.
    Borrowed from [1] and [22]; the entire aging-management objective is defined through this model.
  • domain assumption Deep idle (C6) fully halts aging because transistor switching stops.
    Section 3.1 and Algorithm 2 assume deep-idled cores do not age; in real silicon, leakage and thermal effects can continue degradation.
  • ad hoc to paper Reduced mean frequency degradation maps linearly to extended CPU lifespan.
    Section 6.2 introduces this without derivation or reliability data; it directly converts aging results into the headline 37.67% carbon reduction.
  • domain assumption The modeled CPU tasks in Table 2 are representative of real LLM inference CPU load.
    Section 5 models only splitwise-sim executor, instance, and link function calls; real servers also run OS, networking, and monitoring tasks.
  • domain assumption The Azure inference traces used in splitwise-sim are representative production traces.
    Section 6.1.2 takes traces from [26]; the underutilization pattern the technique exploits depends on trace realism.
  • domain assumption CPU hardware refresh provides no meaningful performance benefit for inference clusters because single-core performance has plateaued.
    Motivates lifespan extension as the only reason for CPU refresh; newer CPUs also deliver power and efficiency gains that could alter the cost-benefit.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Aging-aware CPU Core Management for Embodied Carbon Amortization in Cloud LLM Inference." pith.science (2026). https://pith.science/paper/Y5QJGOJS

@misc{pith2026250115829,
  author       = {Pith},
  title        = {Pith review of: Aging-aware CPU Core Management for Embodied Carbon Amortization in Cloud LLM Inference},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y5QJGOJS}},
  note         = {Machine review of arXiv:2501.15829}
}
abstract

Broad adoption of Large Language Models (LLM) demands rapid expansions of cloud LLM inference clusters, leading to accumulation of embodied carbon$-$the emissions from manufacturing and supplying IT assets$-$that mostly concentrate on inference server CPU. This paper delves into the challenges of sustainable growth of cloud LLM inference, emphasizing extended amortization of CPU embodied over an increased lifespan. Given the reliability risks of silicon aging, we propose an aging-aware CPU core management technique to delay CPU aging effects, allowing the cluster operator to safely increase CPU life. Our technique exploits CPU underutilization patterns that we uncover in cloud LLM inference by halting aging in unused cores and even-outing aging in active cores via selective deep idling and aging-aware inference task allocation. Through extensive simulations using real-world Azure inference traces and an extended LLM cluster simulator from Microsoft, we show superior performance of our technique over existing methods with an estimated 37.67\% reduction in yearly embodied carbon emissions through p99 performance of managing CPU aging effects, a 77\% reduction in CPU underutilization, and less than 10\% impact to the inference service quality.

Figures

Figures reproduced from arXiv: 2501.15829 by the authors.

Figure 1
Figure 1. Carbon footprint of A100x4 GPU server running [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Distributions of running inference tasks in an LLM inference cluster of 22 H100 machines. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. High-Level System Diagram of Aging-aware CPU [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Changes in Operating temperature when 6 out of [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Behavior of the piecewise Reaction Function ( [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Comparison of managing aging effects in CPU. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Comparison of estimated yearly CPU embodied carbon reduction in the cluster through management of CPU aging [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Comparison of utilization of available cores for running tasks. X-axis in each plot denotes normalized idle CPU cores, [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Software Engineering for Large Language Models: Research Status, Challenges and the Road Ahead

    cs.SE 2025-06 conditional novelty 4.0 of 10

    A literature review organizes LLM development into a six-phase software engineering lifecycle and identifies challenges and research directions for each phase.

Reference graph

Works this paper leans on

42 extracted references · 39 canonical work pages · cited by 1 Pith paper

  1. [1]

    Mohsen Ansari, Sepideh Safari, Amir Yeganeh-Khaksar, Roozbeh Siyadatzadeh, Pourya Gohari-Nazari, Heba Khdr, Muhammad Shafique, Jörg Henkel, and Alireza Ejlali. 2023. ATLAS: Aging-Aware Task Replication for Multicore Safety- Critical Systems. In Proceedings of the 29th IEEE Real-Time and Embedded Tech- nology and Applications Symposium (RTAS). 223–234

    work page 2023

  2. [2]

    AWS. 2025. Deploying Multiple Large Language Models with NVIDIA Triton Server and vLLM. https://awslabs.github.io/data-on-eks/docs/gen-ai/inference/ GPUs/vLLM-NVIDIATritonServer

    work page 2025

  3. [3]

    Azure. 2024. NCads H100 v5-series. https://learn.microsoft.com/en-us/azure/ virtual-machines/ncads-h100-v5

    work page 2024

  4. [4]

    Sally Beatty. 2024. Microsoft builds first datacenters with wood to slash carbon emissions. https://news.microsoft.com/source/features/sustainability/microsoft- builds-first-datacenters-with-wood-to-slash-carbon-emissions/

    work page 2024

  5. [5]

    Ricardo Bianchini, Christian Belady, and Anand Sivasubramaniam. 2024. Data Center Power and Energy Management: Past, Present, and Future. IEEE Micro 10 Aging-aware CPU Core Management for Embodied Carbon Amortization in Cloud LLM Inference 44, 5 (2024), 30–36

    work page 2024

  6. [6]

    Jairus Bowne. 2024. Using Large Language Models in Learning and Teach- ing. https://biomedicalsciences.unimelb.edu.au/study/dlh/assets/documents/ large-language-models-in-education/llms-in-education

    work page 2024

  7. [7]

    Intel Corporation. 2018. CPU Idle Time Management. https://www.kernel.org/ doc/html/v5.4/admin-guide/pm/cpuidle.html

    work page 2018

  8. [8]

    Microsoft Corporation. 2022. The role of embodied carbon in cloud emissions. https://go.microsoft.com/fwlink/p/?linkid=2233506

    work page 2022

Show all 42 references
  1. [9]

    Crusoe. 2023. How Together And Crusoe Are Reducing The Carbon Impact Of Generative AI. https://crusoe.ai/blog/crusoe-together-reducing-carbon-impact- of-generative-ai/

    2023

  2. [10]

    Crusoe. 2024. Crusoe to Build Initial 200 MW AI Data Center With Plans to Expand at 1.2 GW Lancium Clean Campus. https://crusoe.ai/newsroom/crusoe- 200mw-ai-data-center/

    2024

  3. [11]

    Ahmad Faiz, Sotaro Kaneda, Ruhan Wang, Rita Chukwunyere Osi, Prateek Sharma, Fan Chen, and Lei Jiang. 2024. LLMCarbon: Modeling the End-to- End Carbon Footprint of Large Language Models. In Proceedings of the Twelfth International Conference on Learning Representations

    2024

  4. [12]

    Zhenxiao Fu, Fan Chen, Shan Zhou, Haitong Li, and Lei Jiang. 2024. LLMCO2: Advancing Accurate Carbon Footprint Prediction for LLM Inferences. arXiv preprint arXiv:2410.02950 (2024)

    2024 arXiv

  5. [13]

    GitHub. 2021. Introducing GitHub Copilot: your AI pair program- mer. https://github.blog/news-insights/product-news/introducing-github- copilot-ai-pair-programmer/

    2021

  6. [14]

    Dennis Gnad, Muhammad Shafique, Florian Kriebel, Semeen Rehman, Duo Sun, and Jörg Henkel. 2015. Hayat: harnessing dark silicon and variability for aging deceleration and balancing. In Proceedings of the 52nd Annual Design Automation Conference. Article 180, 6 pages

    2015

  7. [15]

    Lee, David Brooks, and Carole-Jean Wu

    Udit Gupta, Mariam Elgamal, Gage Hills, Gu-Yeon Wei, Hsien-Hsin S. Lee, David Brooks, and Carole-Jean Wu. 2022. ACT: designing sustainable computer systems with an architectural carbon modeling tool. In Proceedings of the 49th Annual International Symposium on Computer Archite...

    2022

  8. [16]

    IEA. 2023. International Energy Agency’s report on Low-emissions sources of electricity. https://www.iea.org/reports/low-emissions-sources-of-electricity

    2023

  9. [17]

    Baolin Li, Yankai Jiang, Vijay Gadepally, and Devesh Tiwari. 2024. Sprout: Green Generative AI with Carbon-Efficient LLM Inference. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing . 21799–21813

    2024

  10. [18]

    Yueying Lisa Li, Omer Graif, and Udit Gupta. 2024. Towards Carbon-efficient LLM Life Cycle. In Proceedings of the 3rd Workshop on Sustainable Computer Systems

    2024

  11. [19]

    Lorenzon, Guilherme Korol, Marcelo Brandalero, and Antonio Car- los Schneider Beck

    Arthur F. Lorenzon, Guilherme Korol, Marcelo Brandalero, and Antonio Car- los Schneider Beck. 2023. Harnessing the Effects of Process Variability to Mitigate Aging in Cloud Servers. In Proceedings of the 2023 IEEE Computer Society Annual Symposium on VLSI (ISVLSI) . 1–6

    2023

  12. [20]

    Sitaraman

    Diptyaroop Maji, Noman Bashir, David Irwin, Prashant Shenoy, and Ramesh K. Sitaraman. 2024. Untangling Carbon-free Energy Attribution and Carbon Inten- sity Estimation for Carbon-aware Computing. In Proceedings of the 15th ACM International Conference on Future and Sustainable...

    2024

  13. [21]

    Rick Merritt. 2023. Why GPUs Are Great for AI. https://blogs.nvidia.com/blog/ why-gpus-are-great-for-ai/

    2023

  14. [22]

    Salehi, and Mehdi Kargahi

    Iraj Moghaddasi, Arash Fouman, Mostafa E. Salehi, and Mehdi Kargahi. 2019. Instruction-Level NBTI Stress Estimation and Its Application in Runtime Aging Prediction for Embedded Processors. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems 38 (2019),...

    2019

  15. [23]

    Sophia Nguyen, Beihao Zhou, Sihang Liu, and Yi Ding. 2024. Towards sustainable large language model serving. In Proceedings of the 3rd Workshop on Sustainable Computer Systems

    2024

  16. [24]

    LA Office of the governor. 2024. Landry Announces Meta Selects North Louisiana as Site of $10 Billion Artificial Intelligence Optimized Data Center. https: //gov.louisiana.gov/news/4697

    2024

  17. [25]

    OpenAI. 2022. Introducing ChatGPT. https://openai.com/index/chatgpt/

    2022

  18. [26]

    Pratyush Patel, Esha Choukse, Chaojie Zhang, Aashaka Shah, Íñigo Goiri, Saeed Maleki, and Ricardo Bianchini. 2024. Splitwise: Efficient Generative LLM Infer- ence Using Phase Splitting. , 118-132 pages

    2024

  19. [27]

    GHG Protocol. 2024. Greenhouse Gas Protocol. https://ghgprotocol.org/about-us

    2024

  20. [28]

    Bharathwaj Raghunathan, Yatish Turakhia, Siddharth Garg, and Diana Mar- culescu. 2013. Cherry-picking: Exploiting process variations in dark-silicon homogeneous chip multi-processors. In Proceedings of the 2013 Design, Automa- tion & Test in Europe Conference & Exhibition (DAT...

    2013

  21. [29]

    Reuters. 2024. Musk’s xAI plans massive expansion of AI supercomputer in Memphis. https://www.reuters.com/technology/artificial-intelligence/musks- xai-plans-massive-expansion-ai-supercomputer-memphis-2024-12-04/

    2024

  22. [30]

    Faezeh Sadat Saadatmand, Nezam Rohbani, Farshad Baharvand, and Hamed Farbeh. 2021. TAMER: an adaptive task allocation method for aging reduction in multi-core embedded real-time systems. The Journal of Supercomputing 77 (2021), 1939–1957

    2021

  23. [31]

    Philipp Schmid, Omar Sanseviero, Pedro Cuenca, and Lewis Tunstall. 2023. Llama 2 is here - get it on Hugging Face. https://huggingface.co/blog/llama2

    2023

  24. [32]

    Tianyao Shi, Yanran Wu, Sihang Liu, and Yi Ding. 2024. GreenLLM: Disaggregat- ing Large Language Model Serving on Heterogeneous GPUs for Lower Carbon Emissions. arXiv preprint arXiv:2412.20322 (2024)

    2024 arXiv

  25. [33]

    The Financial Times. 2024. OpenAI targets 1bn users in next phase of growth. https://www.ft.com/content/e91cb018-873c-4388-84c0-46e9f82146b4

    2024

  26. [34]

    Abhishek Tiwari and Josep Torrellas. 2008. Facelift: Hiding and slowing down aging in multicores. In 2008 41st IEEE/ACM International Symposium on Microar- chitecture. 129–140

    2008

  27. [35]

    Amanda Tomlinson and George Porter. 2023. Something Old, Something New: Extending the Life of CPUs in Datacenters. SIGENERGY Energy Inform. Rev. 3, 3 (2023), 59–63

    2023

  28. [36]

    Shashank Verma and Neal Vaidya. 2023. Mastering LLM Techniques: Inference Optimization. https://developer.nvidia.com/blog/mastering-llm-techniques- inference-optimization/

    2023

  29. [37]

    Grant Wilkins, Srinivasan Keshav, and Richard Mortier. 2024. Hybrid Heteroge- neous Clusters Can Lower the Energy Consumption of LLM Inference Workloads. In Proceedings of the 15th ACM International Conference on Future and Sustainable Energy Systems. 506–513

    2024

  30. [38]

    Jawad Haj Yahya, Jeremie S. Kim, A. Giray Yağlıkçı, Jisung Park, Efraim Rotem, Yanos Sazeides, and Onur Mutlu. 2022. DarkGates: A Hybrid Power-Gating Architecture to Mitigate the Performance Impact of Dark-Silicon in High Perfor- mance Processors. In Proceedings of the 2022 IE...

    2022

  31. [39]

    Bartolini, Georgia Antoniou, Jeremie S

    Jawad Haj Yahya, Haris Volos, Davide B. Bartolini, Georgia Antoniou, Jeremie S. Kim, Zhe Wang, Kleovoulos Kalaitzidis, Tom Rollet, Zhirui Chen, Ye Geng, Onur Mutlu, and Yiannakis Sazeides. 2022. AgileWatts: An Energy-Efficient CPU Core Idle-State Architecture for Latency-Sensi...

    2022

  32. [40]

    Gyeong-In Yu, Joo Seong Jeong, Geon-Woo Kim, Soojeong Kim, and Byung- Gon Chun. 2022. Orca: A Distributed Serving System for Transformer-Based Generative Models. In Proceedings of the 16th USENIX Symposium on Operating Systems Design and Implementation (OSDI 22) . 521–538

    2022

  33. [41]

    Xiaoyang Zhang, Yijie Yang, and Dan Wang. 2024. Spatial-Temporal Embodied Carbon Models for the Embodied Carbon Accounting of Computer Systems. In Proceedings of the 15th ACM International Conference on Future and Sustainable Energy Systems. 464–471

    2024

  34. [42]

    Jiechen Zhao, Katie Lim, Thomas Anderson, and Natalie Enright Jerger. 2023. The Case of Unsustainable CPU Affinity. In Proceedings of the 2nd Workshop on Sustainable Computer Systems. Article 1, 7 pages. 11

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.