@wgm247
This type of video demonstrates exactly why the justjosh channel is on another level to other tech/laptop channels on youtube
@RetroGunzy
As a writer who trains our company's in-house AI models, I make it a point to watch each of these videos. Even when the content doesn't directly relate to my work, I find that there's always valuable insight for my laptop search.
@jeffrey5602
DS here who managed to get an M2 Max with 96GB at work. Love the device. But don't kid yourself. You will not do any significant LLM training on this thing either. Its fine if you stay below that 1B weights range I would say. If you think about buying something for personal use just get a cheaper device + setup a local workstation and ssh into that. Will also help you build up your skills
@PiracyAgreement
An underrated feature for Excel heavy users is the full-size arrow keys. For me, it's a non-negotiable.
@Mayur7Garg
I am a professional Data Scientist and here are the things that I prefer in a laptop: Minimum requirements: 15+ inches and 16GB RAM Preferred: 32 GB RAM Case dependent: - If you use something like Excel (or other Office Suite) or in memory analytical libraries like pandas a lot, prefer laptops with a high single core performance. Also, choose the RAM based on how much data you process at once. - If you perform distributed analysis or build CPU based models (like scikit-learn), prefer high multi-core performance and high thread count. - For deep learning models for tabular data, I would prefer faster GPUs over VRAM. But depends on the scale of the data. - For deep learning models for images or LLMs, I would prefer GPUs with high VRAM over raw GPU performance. My suggestion would be 8GB+. You could get a lot more done with a higher batch size. - A good keyboard is good to have but should not be the main buying decision imo unless you are a super Excel user or something. This is since most programmers, including myself, spend most of the time reading code and data and a lot less typing. - If you use a laptop for professional use, most likely you would be using a desk style setup. In that case, a long battery life and a nice trackpad is nice to have but should not be the focus. In a desk setup, it is better to plug in the laptop and use a mouse. - For deep learning models, I prefer Nvidia GPUs since I feel like it is more supported and easier to set up. - Screens are subjective. Choose the screen type based on your budget and how likely you are to use the laptop for other stuff (gaming, Netflix, YT). Choosing the screen specifically for coding is a bit overrated imo. Bonus: - If you use Excel a lot or any IDE (VS Code, PyCharm, etc.) a lot, buy a wide screen monitor. It is a game changer for development. These are just my opinions and preferences.
@franciscowilhelm1083
Using personas to calibrate the recommendations to prototypical use cases. Excellent quality video!
@copacialex
Wait what?! I don't believe it is possible to use a laptop for something different than content creation. It should not be allowed! 'Sad' thing is that who needs a laptop for other stuff then 'EXPORTING MASSIVE 4K VIDEOS" they already know what they need or want. However, this is a great educational video and very professional. Thank you!
@corey7219
3:06 having a high refresh rate display lessens the strain on your eyes when you see moving content on screen. Which makes a big difference while I'm at work
@akin242002
As a former Insurance underwriter turned IT analyst, most of the time you are in an office or working from home. 300 nits of brightness are perfectly fine. Lenovo ThinkPad T14 GX, Dell Latitude 7XXX, and HP EliteBook 840 GX will be the laptops assigned to most data analyst workers by their company.
@jonathantran7102
7.5 minutes in.. just wow, amazing data, love how you present the different processors, and the tier chart for fan noise/heat is incredibly helpful
@Mike-jb3xe
Never think a laptop is good for AI / ML, you don't want a long running task living in your laptop, and make everything else awkward (annoyed by the heat but can't do a pause / resume, accidentally close the lid, etc). Would rather build and ITX pc and ssh into it.
@Ty.mauricee
I just started studying data science so this is the perfect video for me, thanks josh👍🏾
@Hundredthldiot
I'm really not convinced about the case for LLMs on laptops, or even desktops. I don't work in AI, but conventional software engineering went through this over a decade ago with the transition from local and on-premise CI builds (e.g. Jenkins) to cloud CI (Travis, GHA and so on). I think it's beyond the scope of this channel, but if I was making this decision I would start with "what is it going to cost me in terms of cloud compute, and so what's the ROI period for a more expensive laptop capable of running the same work locally?". There are other benefits of running stuff on cloud, like greater flexibility in terms of instance sizing, and simply being able to close your laptop and let stuff run while you do other things.
@pewpewpower
This is one of your best videos yet! Excellent content organization, pacing, charts📊, the detailed recs and scenario analyses. Just wow. Keep it up!👍
@jonathanofrivia4192
today i was thinking that it would be good to see this subject on your channel and boom! here it is! thanks as always ❤
@lhl
Since I just spent quite a bit of time doing research and running tests, I just want to note that the M4 Max 40CU (top of the line), sadly only has 34.08 TFLOPS of FP16. This is roughly equal to a desktop RTX 4060 in terms of compute. A mobile RTX 4090 (which is pretty cut down from the desktop version) will still have twice the Peak FP16 Tensor TFLOPS w/ FP32 Accumulate (also, 264 INT8 TOPS for quantized inference). Based on 40 RDNA3.5 CUs, Strix Halo should have just shy of 60 FP16 TFLOPS (but only 256GB/s of MBW vs the 576 GB/s that both the Mobile 4090 and M4 Max have). For reference, a desktop 4090 will have 165.2 Tensor FP16 TFLOPS (FP32 Accumulate) and 1008 GB/s MBW. Bottom line: line, if you're doing local training, don't use a laptop unless you really have to.
@nayyarrazakazmi8386
Excellent as always Josh. Please also review laptops for heavy AI based video editing workflows like using AI based functions in Davinci Resolve Studio
@LunahLiu
Here’s the thing, as a DS, I’d like my work laptop to be as powerful as possible as work pays for it. But again we do most of the training on the cloud (burn cash) so it’s not necessarily to max out the hardware. As for my personal laptop, I don’t do much heavy lifting on it. hardware evolve really fast, instead of future-proofing my setup, I’d rather just focus on what I need for now.
@oremad7055
I have to say this method of describing different individuals and their needs regarding their tech is a pretty straightforward and efficient method of helping put perspective in new buyers. You and your team nailed it.
@JustJoshTech