Comment by m_ke

when you're waiting for network or LLM inference the raw performance doesn't matter at all