VRAM Upgrade Paths for Qwen 27B Multi-Agent Setup
Electronic_Back1502 · reddit · 2026-07-17
Running Qwen 3.6 27B at Q4 on an existing 5080/CPU yields about 10 tok/s. The author's goal is to run 2-3 subagents simultaneously: two sharing context, and one with independent context, for code generation, docs/testing, and code review. This requires an estimated 35–40GB of VRAM.
They listed several upgrade paths and asked for advice:
- 4090 + 5080: Plenty of space, but mixing different GPU generations might cause compatibility/scheduling issues
- 5090 + 5080: Clearly the best option, but too expensive
- 2×3090: Seems the safest bet, but concerns about falling behind newer cards in performance
- 4×5060 Ti: Feasible, but might require a massive system overhaul
More from coding & agent
- Supabase ships official plugins for Kimi Code and Kimi Web — dshukertjr · 2026-07-21
- Claude Code often abandons the plan when a code change gets too large — roske_e · 2026-07-21
- Belgie: Embeds a TypeScript Sandbox for Python AI Agents Without Node.js — TheRealMrMatt · 2026-07-21
- Anthropic's AI Submits 65% of Engineering PRs, Slashes Prompts by 80% — Simon Willison · 2026-07-21
- OxDeAI adds signed, fail-closed authorization before AI agents can act — docybo · 2026-07-21
- Multiagent v2 playbook calls for 64 agents, diverse proof routes and adversarial checks — danshipper · 2026-07-21