The RAM shortage is bringing back DDR4

Original Article ↗ Hacker News Discussion ↗

Surging demand for memory in AI data centers is pushing up prices for modern DDR5 RAM, making older standards like DDR4 and even DDR3 attractive again for consumers, homelabs, and refurbished enterprise gear. Commenters note that for most everyday workloads and many games, older memory and CPUs remain “good enough,” even as high-end GPUs and RAM become unaffordable for hobbyists. The trend raises broader concerns about widening gaps between cloud-scale computing and personal hardware, potential regressions in consumer tech, and both the environmental upside and economic downside of extending the life of old systems.

DDR4 vs DDR5 in Real Use

  • Many report little practical difference after moving to DDR5; perceived gains mostly come from newer CPUs, not RAM.
  • Latency between decent DDR4 and DDR5 is described as similar; DDR5’s main advantage is bandwidth.
  • For office / browsing workloads, “more RAM” is seen as much more valuable than “faster RAM,” due to app bloat and disk caching.
  • Several users say DDR4 (e.g., 32–64 GB at 3200 MHz) is still perfectly fine for most consumer tasks, including gaming, at non-maximal settings.

Price Spikes and Shortages

  • Numerous concrete examples of RAM prices having doubled or worse, especially high‑capacity DDR5 kits and DDR4/DDR3 ECC server RAM.
  • Old DDR3/DDR4 server boxes and memory have appreciated sharply on resale markets; some owners could almost finance new servers by selling RAM alone.
  • Some note DDR4 is no longer the “cheap” option; non‑ECC DDR4 and DDR5 prices are converging, ECC is expensive across the board.

AI Demand and Market Dynamics

  • Strong consensus that AI datacenter demand (HBM and DRAM for GPUs/AI accelerators) is hoovering up supply intended for consumer markets.
  • Debate over whether shortages are “artificial”:
    • One view: vendors and large buyers are strategically squeezing supply and over‑signing contracts.
    • Counterview: demand is genuinely massive; timelines may be hyped but not fabricated.
  • Concerns that AI roll‑out may be inflationary and that demonstrated productivity gains are unclear.

Resurgence and Repurposing of Older Hardware

  • Growing ecosystems around decommissioned servers (Xeon DDR3/DDR4, Chinese X79/X99/EPYC boards) for homelabs, NAS, and local AI.
  • Old devices (e.g., Kindles, e‑ink panels) repurposed as AI frontends, dashboards, or art displays.
  • Some celebrate lower e‑waste; others note that old hardware is also becoming expensive and thus less accessible.

Impact on Consumers and Computing Models

  • Worry that consumer tech is regressing: higher prices for phones and PCs, especially painful in low‑income regions.
  • Fears of a shift toward cloud‑only, DRM‑heavy “trusted devices,” weakening user control and personal computing.
  • Counterarguments frame this as normal market allocation: if people value gaming/DIY, they must pay prevailing prices.

Gaming, GPUs, and Performance

  • Historical GPU shortages from crypto are cited as precedent; many feel GPU price‑performance never recovered.
  • Discussion of how DDR5 can help in CPU‑bound games, but most games are fine on DDR4; VRAM remains the bigger bottleneck.