MacBook Pro 14" M3

Seriously impressive for the 20b model. Loaded on my 18GB M3 Pro MacBook Pro. ~30 tokens per second which is stupid fast compared to any other model I've used. Even Gemma 3 from Google is only around 17 TPS.

I've only tested it on my M3 Pro MacBook but it's not bad. At full speed (5 min = 1 game day) I've only hit performance issues when running my biggest NYC save with 200 stations and like 15 routes. Even then it runs at 70% of max speed, or 90% if you turn of the map layer showing population movement

Hopefully Apple will show Preview some love though. I can barely open 15k x 15k tif images on a M3 Pro MBP and cannot zoom in at all, while Steve Jobs was able to do the same at a keynote in 2009

I think price/hardware unit (CPU/ram/SSD) is actually lower, especially if we target premium models. Strictly speaking you have to spend more on absolute base, but then you get more hardware capacity. The same 3k now gives me double storage, 33% more ram, and faster chip (M3 pro Vs M5 pro) than the

If the main goal is AI/ML, model training, or anything that relies heavily on GPU compute, then spending that much money on a machine without an NVIDIA GPU doesn’t make sense. With that kind of budget, I would go for a laptop with 32 GB RAM and a dedicated 16 GB NVIDIA GPU, then install Linux on it.

I had an m1 8gb MBP for work and M1 8GB for personal use. It worked great but the 8gb started to limit my workflow heavily as apple pushed out newer and newer updates. My workflow hadn't changed. End up with a MBP M3 16GB for work and MBP M4 16GB. Had I gotten a M1 with 16gb still be using it today.

As one of the developers for the fintech software it's definitely not the software lol. I had to profile it to see where the bottleneck was. An example, a build of one of our stacks takes ~20s on my M3 pro MacBook pro / 32gb ram. Doing the same build (the app is multi arch so amd64 on windows and ar

FWIW - I'm using an M3 Pro MacBook Pro, 18gb, and I never (yes, I'm knocking ALL the wood right now) have the problems people seem to experience as endemic in this latest stuff. ** shrug ** I also avoid CocoaPods and non-SPM dependencies like the plague when possible. I use exactly zero plug ins or othe

I do what you do (modelling/large spatial datasets) with R/RStudio.. have a MacBook Pro M3 Pro chip with 36 GB. It is amazingly fast. Very glad I sprung for extra RAM. Surely new M5 chips even more efficient.

For that specific price range. You have a few options. You can go with the 512GB MacBook Neo using the Apple Education Store which will put you at $600 USD plus tax. I'm a huge tech snob, I won't lie. BUT, I've been daily driving it for grad school this past week and it gets the job done and typing