IdeaWell

Project ideas from Hacker News discussions.

A Cray-1 supercomputer replica from 30 "obsolete" Mac Minis

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Theme 1 – Practical value of the compute resources
Many commenters questioned whether renting out or otherwise using the Mac‑Mini cluster would be worthwhile once costs are considered.
- “I doubt distributed compute in a bunch of old i7s is worth much once you factor in the electricity cost” – eric__cartman
- “Rather than rent it out … maybe they could give compute time to universities for free, with the requirement that they provide some sort of display for museum visitors?” – gs17

Theme 2 – Feasibility of reaching TOP500 performance
Discussion repeatedly turned to the raw performance needed to appear on the TOP500 list and why a few‑hundred Mac Minis fall far short.
- “I don't think so, the lowest from 2026 was still 2.66 PFlop/s, so they'd need thousands of Mac Minis.” – gs17
- “Google tells me you can get 350 GF of DP from one 10‑core M4, so you might need 10,000 to get there.” – mat_epice

Theme 3 – Historical perspective on computing power
Several remarks highlighted how dramatically computing power has grown, comparing the original Cray‑1 to modern devices like smartphones.
- “imagine showing even the first iphone to a Cray engineer in 1976 … (Cray-1 was 160 megaflops, smartphones are teraflops)” – ck2
- “The Cray-1 was measured on 64‑bit FLOPS … A modern iPhone can do 300 GF of 64‑bit, so only 2000x.” – mat_epice

🚀 Project Ideas

MacMiniCluster Manager

Summary

  • Provides a simple job scheduler and real‑time monitoring dashboard for a heterogeneous Mac Mini cluster in a museum setting.
  • Core value: turns idle Mac Minis into a usable compute resource while offering visitors live visualizations of workload and energy usage.

Details

Key Value
Target Audience Museum tech staff, exhibit volunteers, university collaborators
Core Feature Lightweight orchestration (job queue, node health, power/thermal metrics) + web‑based visitor dashboard with live graphs
Tech Stack Python (FastAPI), Docker, Prometheus + Grafana, Redis job queue, osxmetrics for macOS stats
Difficulty Medium
Monetization Hobby

Notes

  • HN commenters lamented electricity cost and lack of visualization; this gives them a way to monitor and showcase usage (gs17: “maybe they could give compute time to universities… with requirement that they provide some sort of display”).
  • Enables educational outreach and potential BOINC‑style projects, addressing the desire to show visitors real supercomputing activity.

EduCompute Grant Platform

Summary

  • Matches institutions with spare compute (e.g., museum Mac Mini cluster) to university researchers/students needing free cycles for Monte Carlo or other workloads.
  • Core value: unlocks otherwise wasted compute for academic use while fulfilling the museum’s public‑engagement requirement.

Details

Key Value
Target Audience University researchers, students, museum exhibit managers
Core Feature Marketplace where compute owners list available node‑hours, researchers apply for grants; platform tracks allocation, requires proof of public display (embeddable widget) and provides usage reports
Tech Stack Node.js/Express backend, PostgreSQL, React frontend, OAuth for academic login, API integration with cluster manager
Difficulty Medium
Monetization Revenue-ready: take a small percentage of allocated compute cost as a service fee (e.g., 5 % of credited node‑hours)

Notes

  • HN users suggested giving compute time to universities for free with a display requirement (gs17). This platform formalizes that exchange.
  • Could spark discussion on fair use of nostalgic hardware and drive practical utility by turning a costly exhibit into a research resource.

Cluster Feasibility & Cost Calculator

Summary

  • Web tool that estimates achievable FP64 FLOPS, power draw, and cooling needs for a commodity‑node cluster (e.g., Mac Minis) and compares it to TOP500 thresholds.
  • Core value: helps enthusiasts and museums quickly gauge whether a dream build (like 100 Mac Minis) can reach supercomputer rankings, saving time and money.

Details

Key Value
Target Audience Hobbyists, museum planners, educators, small‑scale HPC enthusiasts
Core Feature Input node specs (CPU/GPU FLOPS, RAM, network bandwidth, electricity price) → outputs theoretical peak FP64 FLOPS, estimated real‑world efficiency, annual power cost, and required network topology to approach TOP500
Tech Stack Python (Streamlit) or JavaScript (Svelte) for frontend, NumPy for calculations, hosted on Vercel or Render
Difficulty Low
Monetization Hobby

Notes

  • HN thread debated whether 100 Mac Minis could reach TOP500, with users citing numbers (gs17, mat_epice). This calculator would give concrete answers.
  • Encourages informed crowdfunding decisions and discussion about networking bottlenecks, a pain point raised by multiple commenters.

Read Later