PulseAugur
EN
LIVE 09:22:53

Used RTX 3090 beats new GPUs on LLM value, analysis finds

By PulseAugur Editorial · [1 sources] ·

A new analysis of 117 local LLM rigs reveals that used RTX 3090 GPUs offer significantly better value for money than newer models like the RTX 5090. Despite the RTX 5090 being faster, the older RTX 3090 provides approximately three times more tokens per euro. For models fitting within 16GB of VRAM, cards like the RTX 5070 Ti and 5060 Ti offer competitive performance per euro. The study also highlights that speculative decoding techniques can provide a substantial speedup without requiring new hardware. AI

IMPACT Suggests that older, used GPUs may offer superior cost-efficiency for running local LLMs compared to newer, more expensive hardware.

RANK_REASON Analysis of hardware performance and value for local LLMs based on a dataset of user builds. [lever_c_demoted from research: ic=1 ai=0.7]

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Used RTX 3090 beats new GPUs on LLM value, analysis finds

How we ranked this

Signal score
22 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Analysis of hardware performance and value for local LLMs based on a dataset of user builds. [lever_c_demoted from research: ic=1 ai=0.7]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · RigBase ·

    What 117 real local-LLM rigs tell us about tok/s per euro in 2026

    <p>"Which GPU should I buy for local LLMs?" usually gets answered with spec sheets. I wanted an answer from <strong>measured</strong> numbers, so I took the open <a href="https://getrigbase.com/data/" rel="noopener noreferrer">RigBase dataset</a> (117 home AI builds, 298 measured…

RELATED ENTITIES

RELATED TOPICS