A Cray-1 supercomputer replica from 30 "obsolete" Mac Minis
First reported by Ars Technica ·
Old consumer hardware can be repurposed to achieve significant computational power for niche applications and historical demonstrations.
The Museo de Historia de la Computación in Spain has unveiled a 1:1 scale visual replica of the 1976 Cray-1 supercomputer. This tribute, developed over 18 months for the Cray-1's 50th anniversary, is powered by 30 "obsolete" 2012-era Mac Minis. When networked, these Minis provide approximately 1.3 teraflops of performance, utilizing 64 active cores. The project highlights the evolution of computer architecture, contrasting the Cray-1's vector processing with the replica's parallel clustering. The museum aims to reach the TOP500 supercomputers list by acquiring more Mac Minis through crowdfunding, with the replica serving as an educational tool for visitors to observe real-time performance mapping.
This project demonstrates a novel approach to building functional, albeit not cutting-edge, supercomputing clusters by leveraging readily available, older consumer hardware. The use of macOS Mojave and MPI showcases how existing operating system features can be adapted for distributed computing, bypassing the need for more specialized or complex setups. Furthermore, the focus on passive convection cooling highlights an innovative solution for noise reduction in dense computing environments, a common challenge in traditional data centers.
The ambition to enter the TOP500 list with repurposed hardware signals a potential shift in how computing power is assessed and achieved, emphasizing resourcefulness and adaptability over raw, high-cost investment. This endeavor affects hobbyists, educational institutions, and museums by proving that significant computational feats are attainable with minimal budget. The next steps to watch involve the success of their crowdfunding campaign and whether they can effectively scale their system to meet the rigorous performance benchmarks required for the TOP500 ranking.
AI-written summary. May contain errors.