I just saw the GamersNexus benchmarks and I wondered to myself, why do they have so many problems with Nvidia on Bazite? I’ve used PopOS with my RTX 30 graphics card and I’ve essentially had performance parity with Windows.

Do you guys think they misconfigured something in the background or do you think that the driver has just gotten worse? What are your experiences?

  • billwashere@lemmy.world
    link
    fedilink
    English
    arrow-up
    7
    ·
    15 hours ago

    Just take a look at their workflow and you tell me. NVIDIA uses an AI-powered code editor called Cursor. It is a fork of VS Code that integrates Large Language Models (LLMs) directly into the programming environment.

    Here a quote from this article:

    A high-profile endorsement from NVIDIA’s CEO Jensen Huang in late 2025 underscored Cursor’s rapid ascent – calling Cursor his “favorite enterprise AI service” and noting that 100% of NVIDIA’s engineers now use AI assistance with a remarkable boost in productivity.

    https://digitalstrategy-ai.com/2025/11/07/cursor-ai-business-model/#%3A~%3Atext=A+high-profile+endorsement+from%2Ca+remarkable+boost+in+productivity.

    In late 2024 and early 2025, CEO Jensen Huang explicitly stated that “every single software engineer at NVIDIA uses Cursor.”

    https://dataconomy.com/2025/10/15/jensen-huang-says-every-nvidia-engineer-now-codes-with-cursor/#%3A~%3Atext=Replit%2C+Cursor%2C+and+Lovable+offer%2Cchip+designers+utilize+the+tool.

    And then there is this article:

    https://fortune.com/2025/11/25/nvidia-jensen-huang-insane-to-not-use-ai-for-every-task-possible/

    So it’s kinda obvious they use a lot of AI in their code. Now there is no direct proof they use it in their drivers but they obviously do given the CEO’s stance.

    • utopiah@lemmy.ml
      link
      fedilink
      arrow-up
      2
      ·
      5 hours ago

      No doubt NVIDIA is peddling AI as they are financially depending on it now.

      Now from claiming something is powerful and even used to actually shipping code on something low level and benchmarkable like (GPU) drivers I have doubt. I imagine they can say they use AI there to rephrase comment and it would “technically correct” but beyond that I’m still skeptical.

      Regarding chip design, AI has been used for decades … if you consider routing to be AI. It’s not generative in the modern sense, it’s not using LLM, but it’s automated a process.

      To me it’s the typical Harvard Business School playbook. C-suite repeat keywords they read in their peer most popular magazine, they aggregate in a document they call “strategy” they lower down the chain of commands people “execute” that because they must, thanks to KPIs.

      I’d love to hear it from an actual engineer working on drivers but I imagine it’d be hard to get a honest opinion with NDAs and all.

      Thanks for providing all the sources!