Low-Cost Servers

Google and researchers at the University of California San Diego have turned a familiar problem into a practical experiment: what happens if old smartphones are not thrown away, but repurposed for computing work. Instead of becoming e-waste, aging Pixel devices can be rebuilt into small, low-cost server-like systems. The approach matters for more than technical curiosity. It targets two pressing issues at once, namely the cost of computing infrastructure and the carbon footprint left behind by discarded electronics. A different use for hardware that still has value Google Research noted that smartphones keep a form of embodied carbon long after they stop being useful as daily devices. Extending their lifespan, the team argues, is more environmentally sensible than replacing them with new hardware too quickly. That idea shaped a project built around used HP Pixel phones. Rather than preserving the phones as handsets, the researchers focused on the core components that remain capable of handling compute tasks. Testing suggested that even a smartphone about three years old can still deliver respectable performance. In certain SPEC benchmark tests, single-core performance from the device was reported to exceed that of some data center server processors. The comparison included servers such as the Asus RS720A-E11 paired with Nvidia H200 or Nvidia RTX Pro 6000 GPUs, along with two AMD EPYC processors. Even so, the overall output of conventional servers remained far ahead for heavy computing workloads. How the phones were converted To transform the devices into server-style nodes, the researchers stripped away parts that were no longer needed. The display, battery, camera, speakers, and outer frame were removed first. The only major component kept was the motherboard containing the system-on-chip, which served as the processing core. After that, Android was replaced with Linux so the machines could operate in a server-friendly environment. That change also made it possible to use orchestration software such as Kubernetes. With multiple boards assembled together, the phones could be organized into a compact compute cluster. Small clusters can still do meaningful work The researchers said that about 25 to 50 used smartphones could deliver computing power comparable to one dual-socket server processor. For organizations with limited budgets, that level of output could be enough for basic tasks without the cost of building a new server stack. UCSD also said a cluster of 20 used smartphones was enough to run one learning application for more than 75 students. That makes the concept especially relevant for campuses and small institutions that need local processing rather than constant cloud reliance. Rising prices for components such as memory chips and storage make the idea even more timely. When new hardware becomes more expensive to assemble, reusing older devices becomes a more attractive option. Where the model could be useful The research team is also planning a local data center made up of around 2,000 used smartphones. A system at that scale is said to be capable of supporting hundreds of classrooms at once. That forecast points toward use cases in universities, laboratories, schools, and smaller organizations with tight budgets. In those settings, cost efficiency often matters as much as raw performance. At the same time, the researchers do not expect this model to replace hyperscale infrastructure used by companies such as Google, Microsoft, and Nvidia. Those operators still need specialized hardware, high reliability, and fewer individual components to manage. The smartphone-based system also still needs further testing to confirm how well its components hold up over long-term server use. The team at UCSD said the full system is expected to begin operation this year while testing continues. If the project succeeds, old Pixel phones that might otherwise sit unused in drawers could become useful building blocks for affordable computing in education and research. Source: tekno.kompas.com

Old Pixel Phones Get a Second Life as Low-Cost Servers, Cutting Lab Costs