Sounds a bit too simplistic. Apple is about the only company that allows you to run any decently sized LLMs on a laptop. They also have all day battery life, amazing single-threaded performance and great multicore performance. And they didn’t stop there, they also threw in a truly fantastic display and speakers that are so far ahead of any competitors that I wonder what kind of magic they have that no other laptop brand is able to put better speakers in their laptops.
> I wonder what kind of magic they have that no other laptop brand is able to put better speakers in their laptops.
The answer is integration.
With Windows laptops, you have one company doing the sound card, one company doing the speakers and one company doing the enclosure. What you actually want is your sound card's equalizer curve being perfectly tuned to your speakers' frequency response, taking into account how their positioning in your specific laptop enclosure affects the sound. If your sound card maker doesn't know what laptop the card will go into, they just straight up can't do this.
There's also the fact that sound cards have to be conservative about how much power they output, to avoid blowing up the speakers. Apple knows the exact tolerances of the ones in their laptops. They even have special temperature sensors in them, so that they can increase the power even more, and go back to a safer level in software if the temperature ever crosses a safety threshold.
Apple is the only company that makes integrated GPUs that are actually any good, and even their CPUs are really well-optimized for these sorts of workloads.
They also build their RAM in a way that makes it suitable for both CPU and GPU uses. This means you don't have separate CPU RAM and GPU VRAM, it's all just memory. If you get more RAM for your Macbook, that automatically means more memory you can use for LLM inference.
Yes, you can technically buy Nvidia, but you'd pay just as much for that if not more, and I don't think you can get as much VRAM in a consumer GPU anyway.
> Apple is the only company that makes integrated GPUs that are actually any good
I will reiterate; if you have used AMD or Intel iGPUs recently you would know this is just plain wrong. AMD's iGPUs often outperform Apple's for the price, and come with a featureset that isn't intentionally gimped and supports "professional" features like Vulkan conformance. People condemn Apple's GPUs for having a featureset too simple for anything besides video transcoding and compute shaders. If Intel didn't exist, Apple would probably be the single least-efficient and lowest performance GPU designer shipping products today. And they had to cut corners to get there.
> If you get more RAM for your Macbook, that automatically means more memory you can use for LLM inference.
And then it's your paltry GPU that's the bottleneck. You can get comparable results on pretty much any sub-120b parameter model with PCIe offloading on Nvidia hardware that costs a fraction of a Mac Studio with enough memory to compete.
If you do not care about speed, pretty much everywhere.
On any other computers with integrated GPU you can have more memory than on Apple computers, where memory is abnormally expensive.
The main advantage of the current Apple computers is a wider interface with memory, which provides a better memory throughput for their integrated GPUs.
In 2025 it is expected that others will catch up with Apple even from this point of view (e.g. AMD Strix Halo).
> Are there any laptops that support 256 GB of RAM that cost less than a fully speced out Macbook Pro?
No, but there are absolutely laptops with more than 128gb of addressable RAM. If you did your research you might have found them, at a literal fraction of the Macbook's price:
Really, this is what I mean by wondering whether anyone even cares anymore. When you work long enough in the tech industry, you end up sitting next to people that just do not care about anything they interact with. People that use a paid GUI to write git commits, people that pay Pakastani developers to do their work for them, people that brag about computers and smartphones like jewelry. Do you actually research the domain that you claim to care about, or do you just own things and then find reasons to get defensive when people attack your insecurities online?
If you have not used a non-Apple laptop since Apple Silicon released, do not pretend you're in a position of authority to make comparisons. I had an M1 Pro for work, and while it was nice for day-to-day browsing I would have never bought one for myself. It's GPU is truly one of the least appealing parts, both from a performance perspective and considering what little it actually supports.
I did eventually stumble upon the Thinkpad one which supports 192 GB of ram. So that is better than the MacBook. I still can’t see where it says that the Dell machines support that much.
The Lenovo one comes with a decent GPU, but you can’t get it with more than 16 GB of memory. But the 5000 ADA should be a lot faster at running smaller models compared to the M4 Max. It does cost a lot. The one with 128 GB of RAM costs 5519 USD compared to the MacBook at 5999 USD. So it’s not that much cheaper.
GPU is great, for smaller LLMs, but the CPU is not that great. The i9-13950HX is quite behind the M4 Max.
Battery life isn’t half bad either. PCmag was getting close to 10 hours, but then again they got 27 hours out of my current laptop and I can say that I get much closer to 10 hours with my workload, so then I would be getting maybe 4 hours out of the Lenovo, which is not exactly great.
I’m glad that you’ve found a machine that works for you.
I take that to mean that it is basically so slow as to be useless and that you would in fact need a Macbook to be able to run this model locally.
This whole thread has become a bit frustrating with comments about how other laptops are competitive with a Macbook. But no one can actually name a single laptop that can do all the things that a Macbook can.
I just replied to someone saying it couldn't run on a Windows laptop with how it can. It's below reading rate. Both are below skimming rate that is more useful for LLMs. I would rather have a 96GB macbook than a windows laptop if my only case was local LLMs and I didn't care about all the hoops around sealed system volume, signed app troubles, etc. that are bringing Macbooks closer to ipads/chromebooks, or the many other things that are better with nvidia GPUs.
For desktop I'd still rather threadripper etc. where it has more competitive CPU memory bandwidth and is upgradable and can run multiple GPUs.
idk about the LLM part, but the rest of this is not true. there's multiple laptops that compete or even beat MacBooks in display, battery life, performance etc.
and the MacBook displays have really slow response times
I guess we all have different requirements in a screen. I don't care so much about response time. For me it's more important that the screen can do 1000 nits continuously, support 120 hz and is close enough to 4K that I can't see the pixels.
Although I can't find a windows laptop that can beat the M4 Max in single threaded or multi threaded. Got any tips for any laptops that can compete with the M4 Max?
