Can a dual-Xeon HP Z8 with 768GB RAM and one RTX 3090 still serve local LLMs?
libbyt91 · reddit · 2026-10-26
A Redditor asks whether an old (2018) HP Z8 workstation with dual Xeon Gold 6248R CPUs, 768GB of 2933MHz DDR4 RAM, an RTX 3090, and hefty storage still has a use case for local LLMs, or is just a boat anchor given the single 24GB GPU. Community discussion typically covers CPU/RAM-based inference of large quantized models or using it as an auxiliary experiment node.
More from Infra
- Meta's Muse Spark Open goes live on Baseten Loops SDK for reward-based post-training — baseten · 2026-10-26
- RTX 3080 20GB gets ReBAR + P2P unlock, boosting prompt processing up to 115% — No-Refrigerator-1672 · 2026-10-26
- vLLM runs DeepSeek v4.1 10-20% faster than SGLang on RTX Pro 6000 — morgymcg · 2026-10-26
- a16z: AI buildout hits skilled-labor bottleneck, job-switching factory workers get ~10% raises — a16z · 2026-10-26
- $599 Mac Mini used as a cheap desktop hides 10 killer uses, from local AI server to VPN — hey_abusiddik · 2026-10-26
- Compound rewrites Excel, PowerPoint and Word engines in Rust for AI agents — peterjliu · 2026-10-26