Local LLM Setup: Is a Modded 3080 20GB Worth It Next to a 3090?
YourNightmar31 · reddit · 2026-07-29
A developer on Reddit asked for hardware upgrade advice: whether it's worth buying a modded RTX 3080 20GB for around €450 from Alibaba to pair with an existing RTX 3090, achieving a total of 44GB VRAM.
- Target Needs: The goal is to run the Qwen3.6 27B model at Q8 quantization with FP16 K/V cache, aiming for a context length of at least 128k or even 256k.
- Hardware Limits: Dual 3090s (48GB) are ideal but currently too expensive (an extra €1,200). The motherboard has two Gen3 16x slots, so physical compatibility is fine.
- Discussion Focus: The community debated the reliability of modded GPUs, the performance overhead of mixing different VRAM capacities within the same architecture, and the trust risks of cross-border second-hand transactions.
More from Infra
- AI is creating a new digital economy around cloud, compute and agents — Scobleizer · 2026-07-29
- Laguna XS 2.1 claims a 39.8% speedup on Mac after benchmark runs — gajesh · 2026-07-29
- Half of major cloud backlogs may now be tied to OpenAI and Anthropic — GaryMarcus · 2026-07-29
- Global semiconductor revenue climbed 24% in Q2 to $394 billion — Beth_Kindig · 2026-07-29
- llama.cpp Fixes MTP Performance Bug, Boosting Qwen Token Generation by 10% — solyarisoftware · 2026-07-29
- Rapid7 says a SmartConsole bypass can hand attackers full admin access — cyb3rops · 2026-07-29