On November 22, 2026, XDA reported that Abhinav examined how ChatGPT, Claude, and Gemini handled a severe storage bug caused by faulty graphics drivers, finding that most AI models struggled to deliver a reliable solution.
- Nvidia drivers 621.14 and 621.24 generated 750MB of rogue disk writes in ten minutes on an idle desktop.
- The AI test evaluated GPT-5.6 Luna, Gemini 4 Hydrogen, and Claude Sonnet 5.5 using identical prompts.
- Consumer solid-state drive prices more than doubled over the past year, amplifying the cost of wasted endurance.
The tech enthusiast used a desktop machine where NVDisplay.Container.exe hammered a Crucial T500 drive. Resource Monitor tracked the process rewriting a single log file at up to 6.5MB per second while the system sat completely idle. CrystalDiskInfo logged about 750MB of host writes in ten minutes, wasting roughly 4.5GB of endurance every hour.
To test the major AI assistants, the reviewer fed each model the exact same vague symptom: an SSD wearing out too fast with 750MB of written data accumulated in ten minutes on an empty desktop. All three models ran at default settings, requiring up to five replies each via screenshots and context to reach a conclusion.
This discovery proved particularly disappointing because the correct fix was one web search away all this time.

The search engine outperformed Google’s own Gemini when identifying the correct solution. Out of the tested AI models, only one provided a fix the reviewer would trust.
ChatGPT began its troubleshooting process by sending the user straight to Resource Monitor. It specifically advised not to kill any process during the examination.
Windows Search, Defender, and SysMain were listed as early suspects by the chatbot. Nvidia was not included on that initial list of potential culprits.

ChatGPT quickly identified the Nvidia process in its second reply but sent the user down a complicated diagnostic rabbit hole involving Task Manager write counters and Process Monitor without ever explaining how to stop the writes.
Gemini located the rogue process within two replies and suggested stopping the Nvidia service, but then advised locking all accounts, including SYSTEM, out of the Nvidia folder to prevent future updates from reversing the fix.
Claude took the longest route, asking for the driver version and correctly identifying the file. It ultimately triggered a web search to spot the driver bug and recommended stopping the container service and deleting the log file. However, it completely missed Nvidia’s 621.35 hotfix that had been released just a day prior.
Abhinav Raj is an XDA contributor and editor-at-large who has written about consumer technology for more than 7 years, bringing extensive hands-on experience tracking semiconductor hardware and artificial intelligence developments. Holding a Master of Science in HRM and AI—with a dissertation exploring how artificial intelligence reshapes the future of work—his expertise spans GPUs, PC systems, and AI evaluation methodologies built through years of hardware enthusiast testing.
“For problems like this, it’s best to stick with a search engine and do your own research for now,” Abhinav concluded regarding the overall efficacy of AI troubleshooting.