I asked ChatGPT, Claude and Gemini what was killing my SSD, and only gave me a fix I'd trust

I asked ChatGPT, Claude and Gemini what was killing my SSD, and only gave me a fix I'd trust

Published Nov 22, 2026, 2:00 PM EST Abhinav pivoted from a career in banking to pursue his first love in writing. Even while working full-time, he continued contributing as an editor-at-large, a role he has held for more than 7 years. A lifelong tech enthusiast who has built three gaming and productivity powerhouse PCs since 2018, his passion for technology keeps him closely following the semiconductor industry, from NVIDIA and AMD to ARM. His MSc dissertation explored how artificial intelligence will reshape the future of work, reflecting his curiosity about the wider social impact of emerging technologies. A few days ago, I covered a notorious bug in Nvidia's 621.14 and 621.24 drivers that has been wreaking havoc on consumer SSDs. On my own desktop PC, NVDisplay.Container.exe was hammering my Crucial T500, and for this test, Resource Monitor found it rewriting a single log file at up to 6.5MB/s. When I launched CrystalDiskInfo, it logged about 750MB of writes in 10 minutes while my PC sat completely idle, which is roughly 4.5GB an hour of endurance wasted at a time when consumer SSDs have more than doubled in price compared to last year. Nvidia has since shipped a hotfix, but plenty of affected users will go to a chatbot before they ever open Nvidia's driver forums. To see where these chatbots would lead a user, I gave Gemini, Claude, and ChatGPT the same vague prompt and followed each one's steps until it gave me a fix. Phrased the way a user would notice the issue All of the three models I tested received the same prompt, which describes the only symptom a casual user would notice, and nothing else. I delivered the following to the three models: "My SSD seems to be wearing out faster than it should. CrystalDiskInfo shows its Total Host Writes went up by about 750MB in 10 minutes while my PC sat at the desktop with nothing open. Can you help me find out what's writing to it? Walk me through it one step at a time and I'll tell you what I see." All the three models ran at their default settings. The models I used included GPT-5.6 Luna, Gemini 4 Hydrogen, and Claude Sonnet 5.5. I answered with screenshots and brief context, and up to five replies each to get me to a fix. And proceeded to never give me a fix ChatGPT began its diagnosis process with the right instincts, and for a short while, it was my favorite. It sent me straight to Resource Monitor, told me not to kill any process, and listed Windows Search, Defender, and SysMain as early suspects. Nvidia wasn't on its list yet, but in its second reply, it spotted NVDisplay.Container.exe writing to a log in Nvidia's nvtopps folder and named it the first place to investigate. That, however, was the beginning of the leaps from one rabbit hole to another. It had me check the log's size, time its growth over the course of five minutes, and pull per-process write counters from Task Manager. Those are all reasonable steps, however, they'd require meticulous patience from a user to stick through all the steps, and none of them stops the barrage of writes. What GPT never did was tell me how to stop it. Its final reply asked me to download Process Monitor for yet another test. It narrowed the problem to Nvidia's display container service, and yet it never asked for my driver version or looked up the bug, so anyone following its advice would still be taking 4.5GB of writes an hour. And then suggested locking Nvidia out of its own folder Similar to ChatGPT, Gemini also went straight for the Resource Monitor, but it asked me to sort by writes instead of total disk activity, which would help by pulling up the processes wearing out the drive. In its next reply, it caught the NVDisplay.Container.exe and its log file, and gave me the standard workaround to stop the NVIDIA Display Container LS service, empty the nvtopps folder, and restart it. Then, however, when I asked for a permanent fix, it proceeded to give me the most counterproductive advice of the entire test. It warned that a driver update would bring the writes back, when a driver update is, in fact, the permanent fix. Its own version of the fix was to lock every account, SYSTEM included, out of writing to the nvtopps folder, promising the Control Panel wouldn't notice. Claude took the longest route to the only fix I'd trust But it still missed Nvidia's hotfix Claude continued the pattern of recommending heading into the Resource Monitor to continue the diagnosis, and after I returned with the screenshots it requested, it named the same Nvidia display container service from the process list alone. Fascinatingly enough, it was the only one to ask me for my driver version out of the three, and correctly read the log as a fixed-size file that was being rewritten in place. Then, however, it took me to a dead end by suspecting the Nvidia overlay, and when that test proved inconclusive, it suggested a plain service restart, which had done nothing for me all week on the same driver. Upon being asked for a permanent fix, though, it finally triggered a web search. It identified the bug in drivers 621.14 and 621.24 and gave me the tried and tested workaround to stop the NVIDIA Display Container LS service, delete the nvlog.nvlgstg, and restart the service. It also flagged rolling back to 621.01, which, in itself, is a fix to the issue at hand, which earns it the point. It did, however, miss the permanent fix, which was Nvidia's 621.35 hotfix from a day earlier, even through many sources online that it had cited had already updated their stories to reflect this. I'd rather stick with a search engine for a diagnosis, for now None of the three models needed long to find Nvidia's problematic display container, but finding it wasn't exactly the challenging part of the problem here. What's particularly disappointing is the fact that the fix was one web search away all this time, and a vague Google Search for "ssd writes increased" pulled up the driver, the process, and the 621.35 hotfix instantly for me, which is more than what Google's own Gemini managed. Yet, only Claude went looking, and only after I specifically had to ask if the problem would come back. For problems like this, it's best to stick with a search engine and do your own research for now.

Original Source

Read the full article at Xda-developers →

KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.