$550 Dell R820 with 768 GB RAM: is an old monolithic server viable for LLM inference?
BobtheGodGamer · reddit · 2026-10-26
A Redditor asks whether a locally available Dell PowerEdge R820 at $550 — 4x Intel E5-4640 CPUs (2.4-2.8 GHz), 768 GB ECC RAM (48x 16 GB), 2x 146 GB HDDs — makes sense for running large-model inference locally, possibly with an added GPU. The reasoning: with DeepSeek-style models now down to 8B/16B active parameters, big-RAM CPU inference could start to be viable. Looking for anyone actually running big models on such monolithic servers.
More from Infra
- GPU rental prices up ~80% this year as token demand climbs, argues analyst — JOBhakdi · 2026-10-26
- Musk warns of stranded compute; trader counters inference demand is near-infinite — markjeffrey · 2026-10-26
- Open-source Kinoloom turns an 8 GB GPU into a local AI film studio with seamless latent-space long takes — Dry-Acanthisitta-652 · 2026-10-26
- Running DeepSeek Flash V4.1 IQ2_XS on Four R9700 GPUs: Real Numbers — poofph · 2026-10-26
- Kubernetes works for vLLM inference, but agent sandboxes needed Modal's custom stack at 1M+ concurrency — NielsRogge · 2026-10-26
- Anthropic reportedly offers Blackstone 10%+ yield on $9B junior debt for 1.5GW TPU deal — Beth_Kindig · 2026-10-26