The user seeks a retail GPU with high VRAM at a reasonable price to run local LLMs after their RTX 4090 failed, suspecting that recent price increases may be intended to hinder open‑source AI. They ask for recommendations on the best current GPUs that maximize VRAM without excessive cost.