From the Graphical Processing Unit (GPU) to computing power: exploring the situated practices of liquid nitrogen overclocking

Drawing from a fieldwork conducted at COMPUTEX Taipei, one of the largest computer expo in the world, this contribution proposes to zoom-in at the level of Graphical Processing Units (GPU) manufacturers and their interactions with computer hardware hobbyists. More specifically, the paper focuses here on hobbyists competing during the event over liquid nitrogen overclocking: a computer hacking practice consisting of building custom GPU cooling rigs in order to increase computing power and benchmark graphically demanding games and simulations. From this empirical analysis, the contribution aims here to situate the crucial role of these hardware enthusiasts in collaborating with GPU manufacturers and corporations, “testing the limits” of such devices before these are produced and commercialised worldwide. This contribution is divided in three parts, shedding light through these hobbyists on the design, optimisation and commercialisation process of our GPU hardware chips. The first section inquires the community’s underlying discourses (or lack of discourses) about computing power and the digital. The section maps their motivations, objectives, goals and desires behind the development of such situated practices. Second, we further dive into the analysis of their strategies and techniques for computing power optimization. In other words, we shed light, here, on their modes of knowledge-building - from data collect, inscription to process optimization - for addressing the possibilities and drawbacks of liquid nitrogen GPU cooling. Finally, we expand on their partnerships with manufacturers and corporations through the process of feedback implementation. Using concrete examples, we show how competition and exploration practices of computer hobbyists impact the design of our GPU chips.