Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I did this type of research last year. I came to the conclusion that you're likely over paying for the portability premium of a laptop with gimped hardware that generally does quite poorly on ventilation. That means your hardware is getting abused by heat, meaning wear and increased likelihood of failure so overall shorter life.

If you want a laptop to actually run AI/ML/RL workloads that would benefit from a graphics card, you're better off optimizing your search for decent specs, no dedicated GPU, portability and good battery life, and spending the money on a powerful desktop setup instead. Your workflow can look like this:

1 - design experiments on laptop or use laptop to remote into desktop with GPU 2 - scale experiments on desktop with GPU till you reach a point where it's maxed out 3 - use Google co-lab with GPU or Paperspace or other similar services (usually offer a free tier for experimentation) 4 - scale on cloud if you really want to parallelize

If you have no choice but to buy only a laptop due to budget, I still think you'd be better off purchasing a laptop without a dedicated GPU and going for points 3/4 above. Seriously, the gains from using an RTX on a laptop are just not worth the cost.

The benefit of using colab or other types of cloud services, is that your investment is low to 0 and will allow you to determine if you actually should invest in your own dedicated hardware.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: