SEARCH · Search NASA

Results for “analysis of variance”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 181 records · Page 10

All sources NASA NTRS DOE OSTI

DOE OSTI2022

Electrical Substation Configuration Effect on Substation Reliability

The nation’s electrical grid must be reliable because of its critical role in the economy, defense, and security of the nation. Many common tasks and industrial operations rely on proper electrical power distribution. In the design of new bus systems or evaluation of existing bus systems, the type of bus configuration affects the risk, reliability, maintenance, and cost of the system. This work utilizes Systems Analysis Programs for Hands-on Integrated Reliability Evaluations (SAPHIRE) to perform a probabilistic risk assessment (PRA) based sensitivity study on the reliability of substations with differing bus configurations. This study examines single bus, main and transfer, breaker and a half, double bus/double breaker, and ring bus configurations applied to busses of varying number of lines. The case variances, methods, and rankings presented can provide a guide for bus configuration for future electrical grid design and analysis.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2025

Polynomial Chaos Surrogate Construction for Random Fields with Parametric Uncertainty

Engineering and applied science rely on computational experiments to rigorously study physical systems. The mathematical models used to probe these systems are highly complex, and sampling-intensive studies often require prohibitively many simulations for acceptable accuracy. Surrogate models provide a means of circumventing the high computational expense of sampling such complex models. In particular, polynomial chaos expansions (PCEs) have been successfully used for uncertainty quantification studies of deterministic models where the dominant source of uncertainty is parametric. We discuss an extension to conventional PCE surrogate modeling to enable surrogate construction for stochastic computational models that have intrinsic noise in addition to parametric uncertainty. We develop a PCE surrogate on a joint space of intrinsic and parametric uncertainty, enabled by Rosenblatt transformations, which are evaluated via kernel density estimation of the associated conditional cumulative distributions. Furthermore, we extend the construction to random field data via the Karhunen–Loève expansion. We then take advantage of closed-form solutions for computing PCE Sobol indices to perform a global sensitivity analysis of the model which quantifies the intrinsic noise contribution to the overall model output variance. Additionally, the resulting joint PCE is generative in the sense that it allows generating random realizations at any input parameter setting that are statistically approximately equivalent to realizations from the underlying stochastic model. The method is demonstrated on a chemical catalysis example model and a synthetic example controlled by a parameter that enables a switch from unimodal to bimodal response distributions.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Harnessing graph convolutional neural networks for identification of glassy states in metallic glasses

Graph Convolutional Neural Networks (GCNNs) have emerged as powerful tools for analyzing materials. In this study, we employ GCNNs to examine structural characteristics of CuZr metallic glasses (MGs) and identify their states. We use molecular dynamics to simulate the quenching process of CuZr, using cooling rates ranging from 10 9 to 10 15 K/s, to produce six unique glassy states. For each state, we create a dataset comprising 1,800 distinct samples. We evaluate the effectiveness of various GCNNs, including Graph Attention Neural Network (GANN), Graph Sample and AggreGatE (GraphSAGE), Graph Isomorphism Network (GIN), and Relational Graph Convolutional Neural Network (RGCN). GANN and GraphSAGE demonstrate comparable performance, achieving an overall accuracy of 81% in classifying the MG states. Furthermore, these results underscore the potential of GCNNs to detect subtle structural variances in disordered materials and point to broader application of deep learning in the analysis of MGs and other amorphous substances.

36 MATERIALS SCIENCE↗

DOE OSTI2025

Monte Carlo Dropout Uncertainty Quantification of Long Short-Term Memory Autoencoder Anomaly Detection in a Liquid Sodium Cold Trap

Advanced high-temperature fluid reactors, such as sodium-cooled fast reactors (SFRs) and molten salt–cooled reactors (MSCRs), require coolant purification systems to prevent fluid contamination and local freezing that can lead to plugging. Liquid sodium purification can be achieved with a cold trap, where the sodium temperature is reduced to a near-freezing point to precipitate out impurities. Automation of monitoring of the cold trap performance with machine learning algorithms can aid in early detection of incipient anomalies. An efficient approach to loss-of-coolant–type anomaly detection in a cold trap monitored with more than two dozen thermal-hydraulic sensors consists of a long short-term memory (LSTM) autoencoder. This work develops the uncertainty quantification of the LSTM autoencoder performance for cold trap anomaly detection using the Monte Carlo (MC) dropout method. The MC dropout methodology creates a distribution of sister distributions that all slightly differ from each other because of random neurons being turned off for testing. The variances of the sister network distributions are used to make an uncertainty interval. Our analysis shows that the uncertainty in the autoencoder performance is largest near the peak of the anomaly signal. Using the MC dropout method, we investigate the uncertainty in the anomaly detection with missing sensor inputs. This capability allows the reactor operator to evaluate resilience of the anomaly detection system and to make informed decisions about continuity of operation in the event of sensor failure.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2026

Turbulence and Cloudiness in Cumulus-Topped Marine Boundary Layers

Turbulence and cloudiness in cumulus-topped marine boundary layers are studied using data from the Atmospheric Radiation Measurement (ARM) Eastern North Atlantic (ENA) site. The analysis includes eight periods of nonprecipitating shallow cumulus clouds, spanning 141 h and encompassing 603 individual clouds. On average, the cumulus had bases at 558 m, were 99 m thick, had a chord length of 500 m, and exhibited an hourly base-layer cloudiness of 12%. Changes in cloud fraction were primarily driven by variations in cloud number not chord length. High-resolution Doppler lidar and cloud radar observations were used to estimate updraft and downdraft mass fluxes in the cumulus base layer under both clear and cloudy regions. Cloudy updrafts contributed only ∼23% of the total updraft mass flux at cloud base, indicating that most upward mass flux originated from clear-air eddies. Updraft strength, rather than updraft fraction, was found to predominantly control both clear and cloudy updraft mass fluxes. Cloud-base cloudiness strongly correlated with cloudy updraft mass flux but showed negligible correlation with clear-air updraft mass flux. Clear-air downdraft mass flux exhibited a strong relationship with the ratio of surface buoyancy to inversion strength. Mesoscale analysis revealed that moist patches had greater cloudiness, updraft mass flux, and vertical velocity variance compared to dry patches. Additionally, mesoscale base-layer cloud fraction was highly and significantly correlated with cloudy updraft mass flux in both dry and moist environments. Results presented herein have implications for cumulus parameterization development along with climatological and model evaluation studies conducted at the ENA site.

cumulus clouds↗

DOE OSTI2025

Measurement of Parity-Violating Modes of the Dark Energy Spectroscopic Instrument (DESI) Year 1 Luminous Red Galaxies' 4-Point Correlation Function

Here we report the first measurement of the parity-violating (PV) 4-Point Correlation Function (4PCF) of the Dark Energy Spectroscopic Instrument's Year 1 Luminous Red Galaxy (DESI Y1 LRG) sample, motivated by the potential detection of the PV 4PCF in the Sloan Digital Sky Survey Baryon Oscillation Spectroscopic Survey (SDSS BOSS) galaxies. In our auto-correlation ("auto") analysis, we find a statistically significant excess of the PV signal compared to mocks without any PV, at 4-10$σ$ depending on details of the analysis. This could arise either from genuine PV or from an underestimation of the variance in the mocks; it is unlikely to arise, at the signal level, from a systematic. We then cross-correlate ("cross") the putative PV signal between different, independent patches of sky, and there find no detection of parity violation. The two measurements are in significant tension: while the cross has somewhat larger error bars than the auto, this is not sufficient to explain the discrepancy. We thus present the current work as an intriguing addition to the PV work on BOSS and as motivation for exploring further the relationship between the auto and cross PV 4PCF analyses.

Slepian, Zachary [Florida U.]↗

DOE OSTI2021

Accurate determination of genotypic variance of cell wall characteristics of a Populus trichocarpa pedigree using high-throughput pyrolysis-molecular beam mass spectrometry

Abstract Background Pyrolysis-molecular beam mass spectrometry (py-MBMS) analysis of a pedigree of Populus trichocarpa was performed to study the phenotypic plasticity and heritability of lignin content and lignin monomer composition. Instrumental and microspatial environmental variability were observed in the spectral features and corrected to reveal underlying genetic variance of biomass composition. Results Lignin-derived ions (including m/z 124, 154, 168, 194, 210 and others) were highly impacted by microspatial environmental variation which demonstrates phenotypic plasticity of lignin composition in Populus trichocarpa biomass. Broad-sense heritability of lignin composition after correcting for microspatial and instrumental variation was determined to be H 2 = 0.56 based on py-MBMS ions known to derive from lignin. Heritability of lignin monomeric syringyl/guaiacyl ratio ( S / G ) was H 2 = 0.81. Broad-sense heritability was also high (up to H 2 = 0.79) for ions derived from other components of the biomass including phenolics (e.g., salicylates) and C5 sugars (e.g., xylose). Lignin and phenolic ion abundances were primarily driven by maternal effects, and paternal effects were either similar or stronger for the most heritable carbohydrate-derived ions. Conclusions We have shown that many biopolymer-derived ions from py-MBMS show substantial phenotypic plasticity in response to microenvironmental variation in plantations. Nevertheless, broad-sense heritability for biomass composition can be quite high after correcting for spatial environmental variation. This work outlines the importance in accounting for instrumental and microspatial environmental variation in biomass composition data for applications in heritability measurements and genomic selection for breeding poplar for renewable fuels and materials.

09 BIOMASS FUELS↗

DOE OSTI2022

Decay Dose Shielding Analysis with Hybrid Unstructured Mesh/Constructive Solid Geometry Monte Carlo Calculation and ADVANTG Acceleration

The Oak Ridge National Laboratory (ORNL) Second Target Station (STS) neutron production facility is an accelerator driven pulsed neutron source that is currently being actively developed at ORNL. The neutrons are produced by proton-induced spallation reactions. A proton beam of 700 kW power is delivered to a spallation target in short, less than 1 µs long pulses, with 15 Hz repetition rate. The spallation target of ORNL STS is a rotating water-cooled tungsten target with tantalum cladding housed in a stainless-steel shroud. It is divided into 21 segments. These segments become highly activated due to spallation reactions or nuclei transmutation by the emitted neutrons. The radioactive nuclides continue to decay after ceasing operation. The decay dose rates generated from the target segments once they are removed from their operational location within the core vessel must be accurately quantified to determine the shielding configurations of remote handling tools and transport casks and to aid in planning maintenance events. To determine the shielding configurations needed for an activated target segment after ceasing operation, both the hybrid unstructured mesh (UM)/constructive solid geometry (CSG) approach that was previously utilized for STS analyses [1] and the ADVANTG code [2] were used. Even though the ADVANTG code does not include UM capability, the utilization of its advanced variance reduction technique was crucial to accelerate the extremely difficult final photon transport calculation in this analysis. This paper also describes the procedures taken to mitigate the convergence issues that occur when ADVANTG uses a source definition that does not match the source of the final Monte Carlo (MC) calculation. These convergence issues often occur because of inconsistencies between source and transport biasing parameters.

Ibrahim, Ahmad↗

DOE OSTI2021

Machine Learning-Based Classification of Lignocellulosic Biomass from Pyrolysis-Molecular Beam Mass Spectrometry Data

High-throughput analysis of biomass is necessary to ensure consistent and uniform feedstocks for agricultural and bioenergy applications and is needed to inform genomics and systems biology models. Pyrolysis followed by mass spectrometry such as molecular beam mass spectrometry (py-MBMS) analyses are becoming increasingly popular for the rapid analysis of biomass cell wall composition and typically require the use of different data analysis tools depending on the need and application. Here, the authors report the py-MBMS analysis of several types of lignocellulosic biomass to gain an understanding of spectral patterns and variation with associated biomass composition and use machine learning approaches to classify, differentiate, and predict biomass types on the basis of py-MBMS spectra. Py-MBMS spectra were also corrected for instrumental variance using generalized linear modeling (GLM) based on the use of select ions relative abundances as spike-in controls. Machine learning classification algorithms e.g., random forest, k-nearest neighbor, decision tree, Gaussian Naïve Bayes, gradient boosting, and multilayer perceptron classifiers were used. The k-nearest neighbors (k-NN) classifier generally performed the best for classifications using raw spectral data, and the decision tree classifier performed the worst. After normalization of spectra to account for instrumental variance, all the classifiers had comparable and generally acceptable performance for predicting the biomass types, although the k-NN and decision tree classifiers were not as accurate for prediction of specific sample types. Gaussian Naïve Bayes (GNB) and extreme gradient boosting (XGB) classifiers performed better than the k-NN and the decision tree classifiers for the prediction of biomass mixtures. The data analysis workflow reported here could be applied and extended for comparison of biomass samples of varying types, species, phenotypes, and/or genotypes or subjected to different treatments, environments, etc. to further elucidate the sources of spectral variance, patterns, and to infer compositional information based on spectral analysis, particularly for analysis of data without a priori knowledge of the feedstock composition or identity.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2026

Tree-ring evidence marks year 2022 as the driest spring season in nearly four centuries in the Western Himalayas

The Hindukush-Karakoram-Himalayan region is a crucial freshwater source for billions across South Asia, yet its climate remains poorly understood due to limited long-term records. Winter and spring precipitation govern snow accumulation and downstream water availability in dry months particularly across the Western Himalayas (WH), but recent decades show intensifying droughts with unclear long-term context. Here, we have reconstructed a nearly four-century-long spring i.e. February to May (FMAM) precipitation for the Lahaul region of the (WH), an area dominated by the Western Disturbances. This record was developed using moisture-sensitive Cedrus deodara (Deodar) tree-rings from three high-elevation sites. A regional composite tree-ring-width chronology, developed through a Nested Principal Component Analysis and modeled with a nonlinear Generalized Additive Model (GAM) that explains 71 % of the variance during the calibration period. We identified the last two decades as the most precipitation deficit phase and the year 2022 showing the driest FMAM on record. The observed rise in the FMAM dry episodes post 1999 CE in our reconstruction, corresponds to the meteorological records. This recent drying is linked to a northward shift of the subtropical westerly jet and reduced moisture transport, both associated with unusual sea surface temperature patterns in the tropical Indian Ocean and the Western Pacific Ocean. Our results provide compelling evidence of long-term hydroclimatic instability in the WH and emphasize the value of tree-ring records in extending precipitation histories beyond the instrumental observations. Such reconstructions can be benchmarks to validate high-resolution climate models and formulate adaptation policies to mitigate future risks.

Cedrus deodara↗

DOE OSTI2020

Metagenome-assembled genome distribution and key functionality highlight importance of aerobic metabolism in Svalbard permafrost

Permafrost underlies a large portion of the land in the Northern Hemisphere. It is proposed to be an extreme habitat and home for cold-adaptive microbial communities. Upon thaw permafrost is predicted to exacerbate increasing global temperature trend, where awakening microbes decompose millennia old carbon stocks. Yet our knowledge on composition, functional potential and variance of permafrost microbiome remains limited. In this study, we conducted a deep comparative metagenomic analysis through a 2 m permafrost core from Svalbard, Norway to determine key permafrost microbiome in this climate sensitive island ecosystem. To do so, we developed comparative metagenomics methods on metagenomic-assembled genomes (MAG). We found that community composition in Svalbard soil horizons shifted markedly with depth: the dominant phylum switched from Acidobacteria and Proteobacteria in top soils (active layer) to Actinobacteria, Bacteroidetes, Chloroflexi and Proteobacteria in permafrost layers. Key metabolic potential propagated through permafrost depths revealed aerobic respiration and soil organic matter decomposition as key metabolic traits. We also found that Svalbard MAGs were enriched in genes involved in regulation of ammonium, sulfur and phosphate. Here, we provide a new perspective on how permafrost microbiome is shaped to acquire resources in competitive and limited resource conditions of deep Svalbard soils.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2020

Forecasting cosmological constraints from the weak lensing magnification of type Ia supernovae measured by the Nancy Grace Roman Space Telescope

We report that the weak lensing magnification of type Ia supernovae (SNe Ia) is sensitive to the clustering of matter and provides an independent cosmological probe complementary to SN Ia distance measurements. The Nancy Grace Roman Space Telescope is uniquely sensitive to this measurement as it can discover high redshift SNe Ia and measure them with high precision. We present a methodology for reconstructing the probability distribution of the weak lensing magnification μ of SNe Ia, p(μ), from observational data, and using it to constrain cosmological parameters. We find that the reconstructed p(μ) can be fitted accurately by a stretched Gaussian distribution and used to measure the variance of μ, ξ μ , which can be compared to theoretical predictions in a likelihood analysis. Applying our methodology to a set of realistically simulated SNe Ia expected from the Roman Space Telescope, we find that using the weak lensing magnification of the SNe Ia constrains a combination of matter density Ω m and matter clustering amplitude σ 8 . SN Ia distances alone lead to a better than 1% measurement of Ωm. The combination of SN Ia weak lensing magnification and distance measurements result in a ~10 % measurement on σ 8 . The SNe Ia from Roman will be powerful in constraining the cosmological model.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2022

The DESI N -body Simulation Project – II. Suppressing sample variance with fast simulations

ABSTRACT Dark Energy Spectroscopic Instrument (DESI) will construct a large and precise three-dimensional map of our Universe. The survey effective volume reaches $\sim 20\, h^{-3}\, \mathrm{Gpc}^{3}$. It is a great challenge to prepare high-resolution simulations with a much larger volume for validating the DESI analysis pipelines. AbacusSummit is a suite of high-resolution dark-matter-only simulations designed for this purpose, with $200\, h^{-3}\, \mathrm{Gpc}^{3}$ (10 times DESI volume) for the base cosmology. However, further efforts need to be done to provide a more precise analysis of the data and to cover also other cosmologies. Recently, the CARPool method was proposed to use paired accurate and approximate simulations to achieve high statistical precision with a limited number of high-resolution simulations. Relying on this technique, we propose to use fast quasi-N-body solvers combined with accurate simulations to produce accurate summary statistics. This enables us to obtain 100 times smaller variance than the expected DESI statistical variance at the scales we are interested in, e.g. $k \lt 0.3\, h\, \mathrm{Mpc}^{-1}$ for the halo power spectrum. In addition, it can significantly suppress the sample variance of the halo bispectrum. We further generalize the method for other cosmologies with only one realization in AbacusSummit suite to extend the effective volume ∼20 times. In summary, our proposed strategy of combining high-fidelity simulations with fast approximate gravity solvers and a series of variance suppression techniques sets the path for a robust cosmological analysis of galaxy survey data.

cosmology: theory↗

DOE OSTI2025

TOFHunter—unlocking rapid untargeted screening of inductively coupled plasma–time-of-flight–mass spectrometry data

This study provides an overview of a newly developed open source program written in Python, TOFHunter, which permits the rapid and untargeted screening of inductively coupled plasma (ICP)-time-of-flight (TOF)-mass spectrometry (MS) datasets. ICP-TOF-MS is an analytical tool capable of providing quasi simultaneous detection of all nuclides from Li to Pu. This capability has triggered an increase in studies investigating single-particle analysis in which the TOF-MS provides correlated elemental/isotopic signatures on a particle basis in time. Similarly, laser ablation mapping has seen rapid growth owing to ICP-TOF-MS's capacity to handle fast washout times (<10 ms) while providing a broad nuclide coverage. The caveat to this broad mass coverage and high time resolution comes in the form of large, overwhelming datasets. With datasets typically on the scale of gigabytes, it is easy for a user to only focus on very targeted analytes; however, this focus diminishes the opportunity offered by the TOF-MS detector. TOFHunter applies chemometric methods, principal component analysis (PCA), and interesting features finder (IFF) on ICP-TOF-MS data, allowing for investigation of correlations, major and minor variance sources, and sample screening. The unique spectra identified by the (IFF) are used to generate a list of mass peaks, which are then matched with both nuclides and potential interferences before being exported for the user to investigate. Several case studies are discussed herein, demonstrating TOFHunter's ability to screen aqueous injections, single-particle/single-cell analysis, and probe laser ablation mapping files for unique regions of interest.

47 OTHER INSTRUMENTATION↗

DOE OSTI2025

A Regularized Variance-Reduced Modified Extragradient Method for Stochastic Hierarchical Games

We consider an N -player hierarchical game in which the i th player’s objective comprises of an expectation-valued term, parametrized by rival decisions, and a hierarchical term. Such a framework allows for capturing a broad range of stochastic hierarchical optimization problems, Stackelberg equilibrium problems, and leader-follower games. We develop an iteratively regularized and smoothed variance-reduced modified extragradient framework for iteratively approaching hierarchical equilibria in a stochastic setting. We equip our analysis with rate statements, complexity guarantees, and almost-sure convergence results. We then extend these statements to settings where the lower-level problem is solved inexactly and provide the corresponding rate and complexity statements. Our model framework encompasses many game theoretic equilibrium problems studied in the context of power markets. We present a realistic application to the study of virtual power plants, emphasizing the role of hierarchical decision making and regularization. Preliminary numerics suggest that empirical behavior compares well with theoretical guarantees.

Tikhonov regularization↗

DOE OSTI2020

Future projections and uncertainty assessment of precipitation extremes in the Korean peninsula from the CMIP5 ensemble

Abstract Projections of changes in extreme climate are sometimes predicted by multimodel ensemble methods that combine forecasts from individual simulation models using weighted averaging. One method to assign weight to each model is the Bayesian model averaging (BMA) in which posterior probability is used. For the cases of extreme climate, the generalized extreme value distribution (GEVD) is typically used. We applied the approach of GEV‐embedded BMA to a series of 35 years of the annual maximum daily precipitation data (both historical data and data gathered from simulation experiments for future periods) over the Korean peninsula as simulated by the models in the Coupled Model Intercomparison Project Phase 5 (CMIP5). Simulation data under two Representative Concentration Pathway (RCP) scenarios, namely RCP4.5 and RCP8.5, were used. Observed data and 17 CMIP5 models for 12 gird cells in Korea have been examined to predict future changes in precipitation extremes. A simple regional frequency analysis of pooling observations from three stations in each cell was employed to reduce the estimation variance and local fluctuations. A bias correction technique using the regression‐type transfer function was applied to these simulation data. Return levels spanning over 20 and 50 years, as well as the return periods relative to the reference years (1971–2005), were estimated for two future periods, namely Period 1 (2021–2050) and Period 2 (2066–2095). From these analyses, relative increase observed in the spatially averaged 20‐year (50‐year) return level was approximately 23% (16%) and 45% (36%) in the RCP4.5 and RCP8.5 experiments, respectively, by the end of the 21st century. We concluded that extreme rainfalls will likely occur two times and four times more frequently in the RCP4.5 and RCP8.5 scenarios, respectively, as compared to in the reference years by the end of the 21st century.

Lee, Youngsaeng↗

DOE OSTI2022

Real-Time Distribution System State Estimation with Asynchronous Measurements

We report state estimation is a fundamental task in power systems. Although distribution systems are increasingly equipped with sensing devices and smart meters, measurements are typically reported at different rates and asynchronously; these aspects pose severe strains on workhorse state estimation algorithms, which are designed to process batches of data collected in a synchronous manner from all the measurement units. In this paper, we develop a novel state estimation algorithm to continuously update the estimate of the state based on measurements received in an asynchronous manner from measurement units. The synthesis of the algorithm hinges on a proximal-point type method, implemented in an online fashion, and capable of processing measurements received sequentially from sensors. A performance analysis is presented by providing bounds on the estimation error in terms of the mean and variance that hold at each iteration and asymptotically. The scheme is also compared with a more traditional Weighted Least Squares estimator that compensates for the lack of measurement data by using, as pseudo measurements, the measurement retrieved during a certain time window. Numerical simulations on the IEEE 37-bus feeder corroborate the analytical findings.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2026

Did You Win the GPU Cloud Lottery? Benchmarking from TFLOPS to Tokens/$

Cloud GPUs are commonly assumed to deliver consistent performance for a given GPU model. This assumption does not always hold: cloud providers employ diverse system configurations and virtualization mechanisms, and GPUs themselves exhibit non-negligible manufacturing variability (the silicon lottery). In this work, we present a large-scale measurement study of GPU performance variability across 11 cloud providers, covering over 3,500 physical GPUs and 6,800 benchmark runs. Our hierarchical analysis shows that while execution-level variation stays below 9%, performance varies by up to 38% across devices and providers for the same GPU model. Regression analysis indicates that driver- and OS-related software factors contribute less than 1% of the variance; instead, silicon lottery effects dominate observed performance variation, and cloud providers further amplify them through persistent, systematic second-order effects.

Slynko, Platon [Silicon Data, New York, USA] (ORCI↗