Energy consumption of scientific visualisation and data visualisation algorithms

DOI

This data set comprises a series of measurements of GPU power consumption when raycasting spherical glyphs, raycasting scalar fields and when showing web-based data visualisation on Observable HQ.

The data sets for sphere rendering were:

pos_rad_intensity : 500000 : 0 : 10 10 10 : 0.01 0.1 500,000 spheres placed randomly, but usinga fixed seed, in a box of 10 × 10 × 10 units with a radius between 0.01 and 0.1 pos_rad_intensity : 5000000 : 0 : 10 10 10 : 0.01 0.1 5,000,000 spheres placed randomly, but usinga fixed seed, in a box of 10 × 10 × 10 units with a radius between 0.01 and 0.1

The scalar fields used for the volume rendering were:

veiled-chameleon.u8.dat A 1024 × 1024 × 1080 8-bit scalar field of a chameleon foot.dat A 256 × 256 × 256 8-bit scalar field of a human foot

The websites visited for the information visualisation application case are:

https://observablehq.com/@d3/bar-chart Bar chart https://observablehq.com/@d3/brushable-parallel-coordinates Brushable Parallel Coordinates https://observablehq.com/@d3/density-contours/3 Density Contours https://observablehq.com/@d3/chord-dependency-diagram Chord Dependency Diagram https://observablehq.com/@d3/world-choropleth Choropleth, World https://observablehq.com/@robsutcliffe/dirty-planet/2</a Dirty Planet (WebGL) https://observablehq.com/@mbostock/hertzsprung-russell-diagram Hertzsprung-Russell Diagram https://observablehq.com/@d3/bollinger-bands Line Chart, Bollinger Bands https://observablehq.com/@pamacha/platonic-gobstopper Platonic Gobstopper (WebGL) https://observablehq.com/@observablehq/plot-horizon Plot: Horizon Chart https://observablehq.com/@d3/sankey Sankey Diagram https://observablehq.com/@d3/json-treemap Treemap, JSON https://observablehq.com/@observablehq/vispubdata Visualization Publication Dataset (Multiple visualisations on the same page) https://observablehq.com/@mbostock/yarn-lock-visualizer Yarn.lock Visualizer (Node-link diagram)

The power sensors employed include the software sensors by AMD and NVIDIA exposed via ADL and NVML, respectively; Tinkerforge Voltage/Current 2.0 bricklets (for measuring most of the internal power rails of the computer individually) and a power analyser attached between the system and the wall socket. They are designated as follows:

ADL/ASIC/{0-6} Power readings obtained from the GPU via the AMD Display Library. HMC Power measurements between system and wall socket obtained from a Rohde & Schwarz HMC8015 NVML Power readings obtained from the gPU via the NVIDIA Management Library TF_ATX_12 Power draw of the ATX 12V rail obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_ATX_3_3 Power draw of the ATX 3.3V rail obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_ATX_5 Power draw of the ATX 5V rail obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_CPU_P4 Power draw of the P4 power connector to the CPU obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_CPU_P8 Power draw of the P8 power connector to the CPU obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_CPU TF_CPU_P4 + TF_CPU_P8 TF_GPU TF_GPU_PCI_12 + TF_GPU_PCI_3_3 + TF_PCIE_1 + TF_PCIE_2 TF_GPU_PCI_12 Power draw of the GPU via the 12V rail of the PEG slot obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_GPU_PCI_3_3 Power draw of the GPU via the 3.3V rail of the PEG slot obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_PCIE_1 Power draw of the GPU via the first PCIe cable obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_PCIE_2 Power draw of the GPU via the second PCIe cable obtained via a Tinkerforge Voltage/Current 2.0 Bricklet TF_SUM The sum of all Tinkerforge Voltage/Current 2.0 Bricklets TF_ATX TF_ATX_12 + TF_ATX_5 + TF_ATX_3_3 - TF_GPU_PCI_12 - TF_GPU_PCI_3_3

For the two scivis application cases, data are provided in form of CSV files that contain pivot tables, which hold all information about a single test run in a single row. The averaged power measurement was made during the time span when measuring the wall-clock timings. As the browser-based information visualisations exhibit a varying power draw while loading and displaying the page, the raw sensor data over time are provided for four separate test runs. For a better overview over these data, we provide line charts in the PDFs located in the infovis-charts folder for each of the tested web-based applications.

Identifier
DOI https://doi.org/10.18419/darus-3044
Metadata Access https://darus.uni-stuttgart.de/oai?verb=GetRecord&metadataPrefix=oai_datacite&identifier=doi:10.18419/darus-3044
Provenance
Creator Müller, Christoph ORCID logo; Heinemann, Moritz ORCID logo; Weiskopf, Daniel ORCID logo; Ertl, Thomas ORCID logo
Publisher DaRUS
Contributor Müller, Christoph; Heinemann, Moritz
Publication Year 2022
Funding Reference DFG 251654672
Rights CC BY 4.0; info:eu-repo/semantics/openAccess; http://creativecommons.org/licenses/by/4.0
OpenAccess true
Contact Müller, Christoph (Universität Stuttgart)
Representation
Resource Type Dataset
Format image/png; application/pdf; text/tab-separated-values; text/csv
Size 125732; 184209; 1231918; 576470; 193676; 214462; 446043; 258322; 45685; 45862; 47301; 43882; 47365; 46454; 45699; 46370; 46561; 45795; 46536; 45681; 49080; 46477; 48654; 90938304; 38507; 37541; 40326; 49115; 39318; 40593; 42142; 43856; 39975; 42200; 42838; 39689; 47876; 52531; 57018; 91647808; 385744; 281505; 295851; 273401; 257830; 217208; 422159; 22401142; 16955018
Version 1.0
Discipline Other
Spatial Coverage Stuttgart