Best hardware configuration for running DeepSeek locally?

Question

My old laptop just died and I really need to run DeepSeek locally for some coding projects by next week. Im stuck deciding between a Mac Studio with 64GB RAM or building a PC with a RTX 4090. Got about 3k to spend. Which one is actually gonna handle the bigger models better?

sxmxuzlsdk · Accepted Answer

Re: TL;DR: Go with the Mac if you need... honestly, just look at NVIDIA GeForce RTX 3090 24GB GDDR6X cards used to save cash. Check out the Ollama docs too, super helpful.

pisvyuhfiv · Answer

Agree with the others, but consider a NVIDIA GeForce RTX 4090 24GB GDDR6X build if you want speed. Mac is great, but that raw CUDA performance is still hard to beat.

Bobthito · Answer

TL;DR: Go with the Mac if you need VRAM for huge models, but build the PC for raw speed. I spent months fighting with a local LLM setup on my old machine before finally switching to a unified memory config. The ease of loading those massive coding models without constant memory errors changed everything for me. Honestly, if you want it working by next week without tinkering for hours, stick to the Mac.