Hacker News Reader

Comment by behnamoh | Hacker News Reader

Like I said, most people want something that's the "latest"; M4 is already one year old.

wrt the second part of your comment: Academics care about speed, RAM, battery life, the ability to run the latest AI models at a decent speed (M4 is still relatively slow).

View on HN

Most of academics write papers, they care about none of those things.

Unless you work on AI, which most don't, then you don't care that the M4 is a little slow for that purpose. The academics who are working on large dataset frequently have access to cluster computers or large servers running in the university datacenter... or frequently sits under their desk, because they have trust issues.

> most people want something that's the "latest"

Most people want something affordable and adequate.

> Academics care about speed, RAM, battery life, the ability to run the latest AI models at a decent speed (M4 is still relatively slow).

Most academics do not give a flying fuck about running local LLMs. Academia is more than LLM researchers.

Most academics probably care most about battery life and portability and whether it runs Teams, like every other person.

m4 laptop checks the boxes on ram options (albiet a high price) and battery life.

The m4 macbook is running AI slow compared too..... what competitive alternative?

I'm impressed at how fast gpt-oss-20b ran on my m2 24gb system.

> most people want something that's the "latest"

citation needed. AFAICT, sales data says that "most people" want a bargin and would prefer $200 off than 10% faster newest etc... Most people buy 1 or 2 generations old.

Double reverse Citation needed

I would venture a guess that 99% of academics could run whatever workloads they have on something half as powerful as an M1 Mac. Very few need or want to run cutting edge LLM models on their laptop.

It's tough for me to justify upgrading an M1 device to an M4 device for the AI performance improvement, going from an M4 to an M5 for a small segment of the user base is not moving the needle enough to sway something like this.

Anyone really geeked about AI performance to the level of not picking something because of it (be it price or performance) should not be looking at a laptop anyways.

As someone who hangs around with academics outside the comp sci field regularly they mostly seem to have old MacBooks which are falling to bits. Not sure I’ve even seen an ARM one yet. They mostly don’t need a lot of juice and don’t want anyone changing anything on it. Also I don’t think I’ve seen anything more than a sneer on AI.

> (M4 is still relatively slow).

How does it compare to Windows laptops?

There are Windows laptops with faster GPUs. Being Nvidia certainly helps with software support.

Are they comparable? No way. Laptops with x86 CPUs and discrete GPUs need a lot of power and dissipate a lot of heat. This makes them larger, heavier, louder, and with terrible battery life. If they are gamer laptops, they also look like toys.

Unless you compare base models (entry level MacBooks are very price competitive, but then upgrades are very expensive), they blow Apple devices on performance and price/performance. An HP Omen or Asus ROG Zephyrus with 1-2TB SSD, 32-64GB RAM, powerful CPU, and a proper NVIDIA 4/5070-5080 (which is better at gaming and AI than M chips, or if you need "unified memory" for bigger AI models, AMD APUs are pretty good) for less than a similar memory and disk (let alone performance in most tasks) MacBook.

Battery life is worse though. Depending on the use (3-4h gaming, 7-8h regular use) still plenty though.

How universities have changed, almost 40 years ago we cared to have access at a computer, any at all.

I have had classmates that did the entire 5 year engineering degree in Software Engineering, using the computer labs computers, the modern ones running a mix of Windows 95 and Red-Hat dual booting, or the more ancient MS-DOS/Windows for Workgroups 3.11 and terminals to the DG/UX server.

Anyone was happy to have a computer of some sort.

> M4 is still relatively slow

Don't expect too much from a single generation upgrade. M4 to M5 won't be anything like the move from Intel to their own silicon.

But as you said, 90% use Apple anyway, so why would Apple bother at all?

Reply on news.ycombinator.com