Same AI hype. Same AI garbage code.
NVIDIA and AMD are two of the loudest corporate voices hyping AI and claiming it will do amazing things for productivity. But it still can’t find bugs like this in their garbage code. It’s bad enough their business practices have created massive component shortages in SSDs, DRAM, and consumer GPUs, but now their drive towards always-on telemetry is wearing out the devices that you own today.
NVDisplay.Container.exe. Instead of appending to its log, it rewrites the same file over and over in a way that reduces the lifespan of your SSD cells. Other users measured writes of ~3GB/hour on driver version 621.14 and, worse, up to 30GB/hour on 621.24, with every Alt+Tab dumping another 500MB to 1GB of writes to the same file. Gigabytes of useless data being continuously overwritten. I knew I was running 621.14, and a few md5sum commands in a loop showed me the contents of nvlog.nvlgstg being rewritten as the machine idled. I suspect I was in the ~3GB/hour bracket, I did not spend the time pinning down an actual rate, as that would have needed the drive’s host-writes counter. Instead, I hit the rollback button on the driver and the behaviour just went away.
At time of writing, it has been 7 days since AMD fixed a similar issue. Is the same AI model writing their drivers? Has anyone at either company even noticed?