mikkeyboi

I'd like to know more about what kind of tweaks can be used, but some that worked for me were: - Kernel (5.7? vs 5.4 shipped with Pop! OS vs low latency) - Driver `ubuntu-drivers install` / proprietary / free / unix - Vulkan shader caching (massive fps drops when you first run the game, after awhile it gets better) - Runner options for Proton/Lutris (GLSL, prefer system libraries) but choosing Linux rather than Windows when competing with a Nvidia GPU is pretty rough, if it was an AMD GPU I think we'd have a pretty clear winner.

Any of the mentioned laptops above are good if you want the battery life. You won't be training models locally (be better using AWS), and even if you do, you leave simulations running overnight on a plugged-in computer. Making use of the GPU in deep learning libraries like tensorflow, keras, or pytorch at a budget of $1500 wouldn't make much of a difference. So I guess my recommendation will be Lenovo P series (P1 for Quadro P2000 or P52s for Quadro P500) - the quadros for the extra push for data preprocessing, not necessarily for deep learning. Excellent battery life too. I would also recommend the Razer blade, they'll likely have previous gen for around that price for Black Friday or random refurb sales. You can also check here to see which GPUs matter for deep learning. http://timdettmers.com/2018/11/05/which-gpu-for-deep-learning/

Agreed, G2 is a good Tier 1 PSU http://www.besthostingpro.com/understanding-psu-tiers-updated-list/ I think the motherboard is fine, but the X470 Aorus Gaming 7 is the more popular pick with lots of cooling and RGB. Both are decent at CPU/RAM overclocking if you're into that kinda thing.