Show HN: RAM Prices

Original Article ↗ Hacker News Discussion ↗

A new site that tracks RAM prices on Amazon has prompted both enthusiasm and scrutiny from hardware enthusiasts. Commenters praise its fast, minimalist interface and useful filters, but request more attributes (ECC, timings, voltage, SMR/CMR, manufacturer tiers) and better geographic coverage, while comparing it to established tools like PCPartPicker and Geizhals. The thread also delves into the ethics and fragility of relying on Amazon affiliate scraping, and uses the price data as a springboard for nostalgia about historically high memory costs and criticism of modern device vendors that ship minimal, non-upgradeable RAM.

Site purpose, features, and UX feedback

  • Tool lists RAM sticks with faceted filters (type, capacity, speed, etc.) and affiliate links to purchase.
  • Users request:
    • “Uncheck all” / invert-selection buttons and easier “select only this” behavior.
    • Column sorting (especially by price/GB).
    • Better capacity and speed controls (numeric entry, sane defaults hiding ancient parts).
    • Manufacturer-tier filters to hide low-trust brands.
  • Bugs and edge cases noted:
    • DDR1 filter not working.
    • Incorrect capacities (e.g., 512 MB labeled as 512 GB) inherited from Amazon product pages.
    • Kits (e.g., 2×16 GB) being treated as 1×32 GB due to Amazon listing format.

Data sources, scraping, and affiliate concerns

  • Site currently pulls only from Amazon; several comments argue this should be stated more prominently.
  • Backend uses shell scripting plus Python/Tcl; HTML is statically generated.
  • Amazon Product Advertising API is mentioned but avoided by some due to qualification hurdles and fear of dependency on Amazon.
  • Scraping approach is intentionally not disclosed to avoid helping bot-detection.
  • Affiliate tagging is present on all outgoing links; some see this as “scummy,” others argue it’s fair compensation for the service as long as it’s disclosed. The site does label affiliate links and notes this in its FAQ.

Comparison to other tools

  • Many similar tools cited: diskprices.com, hdd-prices.com, gpuprices.* variants, listofdisks.com, productchart.com, geizhals, tweakers, PCPartPicker.
  • Some say this recreates PCPartPicker “one part at a time”; others value the focused, minimalist, RAM-specific interface.
  • Suggestions to interlink related tools so users can easily jump between RAM, disks, GPUs, etc.

ECC, DDR5 behavior, and advanced filters

  • Strong demand for ECC vs non‑ECC filters, RDIMM, LRDIMM, buffered/unbuffered, voltage, CAS latency, rank, and timing filters.
  • Discussion clarifies:
    • All DDR5 has on‑die ECC, but this is internal and not visible to OS.
    • “Real” ECC requires a wider (e.g., 72‑bit) bus and platform support.
  • Disagreement over how common ECC-capable systems are in practice and whether ECC RAM is “rare.”
  • Performance impact of external ECC is debated; some claim an extra cycle, others say it can be pipelined.

RAM pricing, vendor upselling, and nostalgia

  • Multiple users marvel that commodity RAM is nearing or below $1/GB.
  • Long nostalgia thread recounts paying hundreds or thousands of dollars for a few megabytes or kilobytes decades ago.
  • Criticism of major vendors (especially laptops and phones) for shipping minimal, non‑upgradeable RAM and charging high upgrade premiums; framed as price discrimination and/or “corporate greed.”
  • Debate over whether Apple’s software stack really uses RAM more efficiently than Windows/Linux/Android, with conflicting anecdotes and no consensus.

Hardware design, performance, and RAM capacity

  • Some argue that “just buy more RAM” is often the most economical way to speed up custom, single‑tenant systems, especially when it keeps all working data in memory.
  • Others caution this mindset encourages sloppy design and can hide large optimization opportunities or inflate cloud costs.
  • Anecdotes about GC-heavy, JIT workloads using tens of GB of RAM to meet latency targets.
  • Frustration that cloud instances with 1–2 TB of RAM are rare and expensive, limiting “everything in RAM” designs.