Emilie Yu

@emxtyu.bsky.social

140 Followers 141 Following

Postdoctoral researcher - Expressive Computation Lab @ UCSB http://em-yu.github.io

A laser cut press fit box with 4 big push buttons
A 6-part figure showing the process of designing a small clay 3D-printed cup. A) A person wearing a hololens holds a virtual triangular rib tool in her hand. In front of her there is a cylindrical object that is a toolpath. B) A close up on the toolpath with the bottom area transformed into a more angled shape, and with a wavy texture pattern. A green triangular shape (the rib) is shown coming closer to that area and affecting it. Icons explain that this is a rib (profile) action, with a texture modifier. C) The top part of the toolpath is transformed by making it slightly curved inward. A green half circle shape (the rib) is shown moving towards that part of the toolpath. An icon explains that this is a rib (profile) action. D) The toolpath is transformed by carving out a small area. A green half circle shape (the rib) is shown moving towards that part of the toolpath. An icon explains that this is a rib action E) The toolpath is transformed by adding a diamond shape embossing with a slight wavy texture. A green square with a diamond image on top (the stamp) is shown making contact with that part of the toolpath where a diamond is added. Icons explain that this is a stamp action, with a texture and repeat x6 modifier. F) The final 3D printed cup showcases the slightly curved shape, the diamond texture, the carved handle, and the slanted foot with wavy texture.
Fusion 360 model of a mushroom hat made of planar slices
RECOMMENDATIONS For Authors: Report hardware used in papers. Consider mid-to-low-end hardware as further opportunities to optimize algorithms. For Institutions: Develop performance metrics that incentivize using low-end hardware. Develop supra-local testing infrastructures with a range of heterogeneous, multi-generational hardware. OPEN DISCUSSION By developing and disseminating methods optimized to work on new hardware, are we actively encouraging premature hardware obsolescence among end-users? By necessitating compute clusters to run our algorithms (the 2nd most-cited GPU is a data center GPU), are we consolidating the dependency of individual users on privatized “cloud” services?
We find that GPUs used in research papers (white dots in graph) are consistently higher-end GPUs. 87% of GPUs reported in research papers (above blue line) are available to less than 20% of the consumer-level user base at publication time. A stacked bar chart shows GPU usage among the general user-base, and white dots represent GPUs used in research papers.
MOTIVATION Graphical Processing Units (GPUs) are at the core of Computer Graphics research. These chips are critical for rendering images, processing geometric data, and training machine learning models. Yet, the production and disposal of GPUs emits CO2 and results in toxic e-waste [1]. METHOD We surveyed 888 papers presented at SIGGRAPH (premier conference for computer graphics research), from 2018 to 2024, and systematically gathered GPU models cited in the text. We then contextualize the hardware reported in papers with publicly available data of consumers’ hardware [2, 3]. REFERENCES [1] CRAWFORD, KATE. The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University Press, 2021. [2] STEAM. Steam Hardware Survey. https://store.steampowered.com/hwsurvey [3] BLENDER. Blender Open Data. https://opendata.blender.org