90K+ Macbook M4 line up worth the wait for a fresher in CS?

jakehifi

Discoverer
Hi Forum Members,
I am looking to give my niece a macbook. She is enrolling for CS so the scope of usage would be a minimum of 4 years.
I want to get technical inputs if it would worth the wait to give the latest line up instead of choosing current m3 ones since Apple is infamous for slowing their old gen devices with OS updates to market their newer gen "faster". Would an M3 powered Macbook pro survive 4 years without lagging or lacking of doing anything a CS student required to do?
 
Yes for atleast 6 years it's good. Using an M1 as my work daily driver since last 2.5 years and it still feels the same as it was when i got it, the performance and the battery.
 
It really doesn't get that "slow". Still I think you should wait and buy the M3 version as it will be available for cheaper when the newer version releases and that way you can snag a good deal. I am sure some will be available on the fourm too. Otherwise you can always buy the newer version, just don't buy the M3 at retail or regular discounted price right now.
 
  • Like
Reactions: meizul
There will be massive changes w.r.t Neural engine on M4 as Apple is now investing heavily in AI. If it not urgent, as new laptops are just months away, wait!
Also, M4 would be more efficient than M3. M3 MBPs used to heat up more and there were instances where heavy workloads would cause laptop to heat up. Dont know if Apple released any microcode updates to throttle a bit. They did it to A17 Pro chips on iPhone 15 Pro series.
Do not buy one with non-Prom M* chip.
Do not buy the base config.
You need at least 16GB memory and 512GB storage.
Factor in Apple Care + for Mac cost when you are buying. This is mandatory. Repairs without Apple Care plus is not easy on purse. I have seen people move to WIndows laptops from Mac when the motherboard of Mx Mac had issues and they did not take Apple Care plus. Yes, it costs less to buy a Windows laptops in case where you have to do major repair or replacement to MBP motherboard.

As for lag, I have an Intel Mac from 2018 and still, it just works. My pesonal macbook is powered by M2 Pro and it is insanely fast. When it comes to a Mac, you do not have to worry about speed/lag.
 
  • Like
Reactions: jakehifi
Do not buy one with non-Prom M* chip.
why avoid the non pro? are the differences that drastic? I have used a base M2 and it seemed fine to me granted I havent dived into the Apple ecosystem at all. OP would have to easily shell out 2L+ for the config you are specifying and considering its for a student, it does seem a little drastic
 
why avoid the non pro? are the differences that drastic? I have used a base M2 and it seemed fine to me granted I havent dived into the Apple ecosystem at all. OP would have to easily shell out 2L+ for the config you are specifying and considering its for a student, it does seem a little drastic
M3 MBP with 16GB/512GB config has 8-core CPU, 10-core GPU, 16-core Neural Engine. This costs 1.9L rupees.
M3 pro MBP with 18GB/512GB base config has 11‑core CPU, 14‑core GPU and 16‑core Neural Engine. This costs 1.99L.

For just 10k more, you are getting better CPU, GPU and little bit more memory. Other changes include an extra Thunderbolt 4 port. Another major difference is that M3 chip supports single external display while M3 Pro supports two external displays. This is without doing jugaad to support more displays. All this for just 10k more.

Here is the shocker though. A 15" M3 MBA that costs less than the M3 MBP has bigger display and has same performance as MBP. The only major differences between the two are HDMI port, SD slot (a dongle/dock can rectify this) and 120Hz display. It is thinner, lighter though it has 1" bigger display.
 
Last edited:
M3 MBP with 16GB/512GB config has 8-core CPU, 10-core GPU, 16-core Neural Engine. This costs 1.9L rupees.
M3 pro MBP with 18GB/512GB base config has 11‑core CPU, 14‑core GPU and 16‑core Neural Engine. This costs 1.99L.

For just 10k more, you are getting better CPU, GPU and little bit more memory. Other changes include an extra Thunderbolt 4 port. Another major difference is that M3 chip supports single external display while M3 Pro supports two external displays. This is without doing jugaad to support more displays. All this for just 10k more.
ah the Apple ladder, say no more lol
 
First I would try to see where her interests lie in, if it includes heavy processing tasks like ML or maybe video editing only then I would suggest the pro lineup otherwise even a base model would do great for at least 5-6 years sure for future proofing get the 16gb/512gb version but beyond that diminishing returns. Most people won't even notice the difference between M3 and M4. Also I would highly suggest using student ID to get one as it gives extra discount and free airpods, which if not wanted you could easily sell.
Once you get in the trap of spending a little more for extra features , it's a spiral better to set a hard budget limit like 1.3-1.5L.
 
There's always something better around the corner. No point in waiting if you need it now. Base models are good enough for casual or semi pro stuff (just get 16GB ram), unless she is going to work with a lot of ML/AI, for which wait for M4.
 
  • Like
Reactions: jakehifi
Base models are good enough for casual or semi pro stuff (just get 16GB ram), unless she is going to work with a lot of ML/AI, for which wait for M4.
Would 16GB RAM suffice and if there'd be need to work with ML/AI stuffs later down the line aren't most of these things rendered on cloud?

And if she really is passionate about computers, you should give her something that is more tinkerable, with more ports.
Do you mean windows? We are already passing a surface pro as her backup laptop

Most people won't even notice the difference between M3 and M4.
I agree that a year difference in processor wouldnt be substantial and exaggerated by marketing
Also I would highly suggest using student ID to get one as it gives extra discount and free airpods, which if not wanted you could easily sell.
Once you get in the trap of spending a little more for extra features , it's a spiral better to set a hard budget limit like 1.3-1.5L.
We are already minimizing expense by getting it from Delaware, US which is tax free + 20% Apple employee discount
unless you have the dough for atleast the 16gig version, then dont buy her a mac, specially if she's enrolling for CS
I already have 16gb m2 pro which i was thinking to hand over but then whether 16GB RAM would suffice in this era of ML/AI and evolving "Chrome"; thought kicked in
 
I already have 16gb m2 pro which i was thinking to hand over but then whether 16GB RAM would suffice in this era of ML/AI and evolving "Chrome"; thought kicked in
16 is indeed the bare minimum but considering apple and how you need to sell your left kidney to afford their higher tier models, wasn't sure if it would be better to recommend higher tier models, and if Mac is not fixed, why not get a Intel core 7 ultra or AMD 8540U laptop?
 
For a starter in ML and programming, your M2 Pro is more than enough. I have the M2 Pro MBP with 16GB memory. My suggestion is to give that one to your niece if that is an option. In office, we have data scientists who use laptops to build a model but always rely on cloud or servers to train model. 16GB memory is more than enough to build a model and do test runs with small data set. To train LLMs though or any ML models with huge dataset, no laptop in this world will suffice.

Running an ML model that runs for hours and hours on a laptop is a very bad idea, even if it is MBP. If your niece gets super serious about ML, there are two options that are good. Either go with Collab or get a powerful desktop. My RTX3090 desktop puts M2 Pro to shame in ML workloads.

TL;DR: give your M2 Pro MBP, more than enough till she gets serious about ML.
 
Running an ML model that runs for hours and hours on a laptop is a very bad idea,
I have run them on my 1050 ti for hours, same with folks in my uni with 1060s and what not, training def took hours but its no more intensive than running a game, as for ram, that 16 gigs wont last long before its filled up, my laptop used to hang back them when training my models because of low ram.

Best course would be to get a lenovo or something with a core 7 ultra and a 4060 if ML is the main priority
 
  • Like
Reactions: ibose
I have run them on my 1050 ti for hours, same with folks in my uni with 1060s and what not, training def took hours but its no more intensive than running a game, as for ram, that 16 gigs wont last long before its filled up, my laptop used to hang back them when training my models because of low ram.

Best course would be to get a lenovo or something with a core 7 ultra and a 4060 if ML is the main priority
I think you misunderstood. For simpler models, laptops will do. I had a 1650 desktop card and now am using 3090. Yes, you can run the models for hours but that does not mean you should. Current laptops are terrible at heat sensitivity. The heavier you use, the lesser the life is.

Your view of taking a with rtx4060 is also great but there is a laptop that can be used without paying extra (M2 Pro MBP which is good enough for starter).

My point is to use the MBP at hand and decide on upgrade based on path taken by OPs niece.
 
I think you misunderstood. For simpler models, laptops will do. I had a 1650 desktop card and now am using 3090. Yes, you can run the models for hours but that does not mean you should. Current laptops are terrible at heat sensitivity. The heavier you use, the lesser the life is.

Your view of taking a with rtx4060 is also great but there is a laptop that can be used without paying extra (M2 Pro MBP which is good enough for starter).

My point is to use the MBP at hand and decide on upgrade based on path taken by OPs niece.
Then I'm guessing existing MBP m2 will suffice with a windows 32gb laptop as backup system (would've built desktop but I don't think that'd be practical kn hostels)
 
I think you misunderstood. For simpler models, laptops will do. I had a 1650 desktop card and now am using 3090. Yes, you can run the models for hours but that does not mean you should. Current laptops are terrible at heat sensitivity. The heavier you use, the lesser the life is
Laptops have actually gotten a lot better at handling temps than before and with regular maintenance its not an issue, I completed Cyberpunk on PT on 5i without any issue and trust me no model would ever come near the usage that a game like cyberpunk demands, Laptops are actually viable for ML as long as the type of models you are intending to train can fit on the vram bummer, its not an issue to train a basic RNN/CNN a few layers deep, even basic GANs or LSTMs are not an issue, these I have trained on my laptop for hours experimenting on my datasets, but go anything beyond a few layers deep or increase matrix dimensions drastically and yeah your vram buffer is gonna overflow and you wont have a good time