Hacker News Reader

Obligatory "I switched to Linux" comment.

But no seriously, after having Windows as my main OS for my entire life I finally switched to Linux (Pop!_OS) a few months ago.

Got tired of nonsensical UI decisions, terrible search functionality, user-hostile behaviour, privacy issues, ads... aaaah! That's not what an OS should be like!

View on HN

I would do the same, but Windows is technically superior in one regard: it actually enables and does proper GPU scheduling (Linux only does CPU and I/O). Super important if you want to do anything on your machine while some neural network is training.

I would imagine distracting GPU with interface rendering significantly reduces ML performance.

On Linux you can use integrated video while your RTX 3090 is 100% loaded.

If you have an integrated video that is sufficient. I don't even know if iGPU can handle 3 monitor setups, let alone smooth experience or any kind of gaming.

ML tasks being delayed by a few minutes is generally a drop in a bucket, but the machine would be barely usable for hours if not days. I honestly don't know how ML enthusiasts do their bidding on Linux.

Reply on news.ycombinator.com