Apple M1 Pro

Yeah… I was shocked when I realized my M1 Pro was twice as fast as my desktop from a few years ago. Minus the gpu

I think a lot of that was just the Steve Jobs’s presence and presentation style. It’s pretty bananas they’ve crammed an M2 into a headset with all the cameras, sensors, ML engines, and high res displays. The M2 chip alone (I have an M1 Pro) is still blowing my mind with the low wattage performance. But much like the first iPhone, it is yet to be seen if it will stick. I do think it’s far more capable than existing VR headsets.

192GB of memory on the Mac Studio is enough to run Llama 65B in full FP16. And at 800GB/s bandwidth, it will do so pretty quickly. I think my M1 Pro memory bandwidth is 200GB/s and I was running quantized 13B Alpaca relatively quickly, I'd say useable for a personal chatbot, and I think it was swapping every now and then causing pauses. So having 4x the memory bandwidth should allow large models to run pretty damn fast. Maybe not H100 GPGPU speeds but enough for people to do some development on.

It always depends on the use case. The casual tests I did for my own use case found a M1 Pro with 10 cores to be comparable to a Ryzen 3700X, which is pretty impressive for a machine that basically makes zero noise when it's not under sustained load.

I have a mac m1 pro at work and it's hands down the best laptop in terms of hardware I've ever had if you forget extensability. By a far margin. It makes no noise at all, to the extent that I even checked a dissasembly video to check if they had fans or just passive refrigeration. The display is the best I've ever seen in any laptop. The trackpad is excellent, the keyboard think is very good, the speakers, microphone and webcam are also very good. I've owned before a lenovo T530, T460s and T580 and quite frankly if we're talking about just hardware this is even better and my last experience with the T580 was just bad. My perception of the lenovo brand was that they made the best laptops until then. In fact I asked to get a macbook because the T580 was so so problematic and not just for me. I have to say I personally dislike macOS. In fact because now I'm working on 100% FOSS software I'm only using the laptop for corporate calls or situations where I have to deal with customer data, but the rest I do on my personal computer with ubuntu.

For me it's more the hardware. I don't think AMD or Intel can compete at all with M1 Pro. The performance per watt is insane, and given the ML/AI thing has blown up recently, I'm happy to have the neural engine. The Apple screen is likely still much better ( colors, variable refresh rate ), the trackpad and it's integration.

It depends on your workload and codebase. I have a Ryzen 5600X in my desktop and for C++ work, my M1 Pro is quite a bit faster for clean or incremental builds. The desktop is still useful/required for some of my work (using x86_64 windows with an nvidia gpu) but I default to the Mac for anything that could be done in either place. It also helps that I prefer the Mac tools so it’s not just about the CPU speed. That said, I’d rather find a new job than trade either system for a cloud desktop. I count myself fortunate that I’ve always been in a position to choose my computer and tools.

I have older XPS 13. I bought a M1 PRO recently, because I was tired of issues with my XPS 15 (which eventually died - won't power on). I wish I didn't have the mindset that XPS was enough for me. For instance when I am running tests on M1, they finish in about a minute, whereas on XPS I could go on a quick coffee break. Having ADHD it was a productivity killer. On M1 I can fire off tests, read something quickly on interwebs and then go back to developing in no time. Best of all, no distracting fans (it has fans, and sometimes they do activate, but I wouldn't know if I hadn't installed a fan monitor - they are that silent). > and setting up the dual-boot myself. Only good thing about XPS is that you can replace the NVMe drive. So when I bought mine, I also bought a separate NVMe drive and simply removed the one laptop came with and just installed Linux on a new drive. When my XPS died, I could buy a cheap used laptop (got a ThinkPad!) and put my XPS's drive in it and recover any data I needed. With M1 you don't have that option sadly.

Not using full-time but have a partition for it. I went no DE on my install, figured I'd hold off until we got GPU support. I'm on an M1 Pro and all the essential hardware works so it's fine for my purposes (C++ development).

Small correction; it’s DisplayLink, not DisplayPort. You have to use DisplayLink because they do some part in software as a part of their driver (which you must download and install) whereas the native Apple stuff has some internal hardware limitation. I run three 1080p monitors off of my M1Pro, and I don’t notice any heavy CPU load from whatever software load the driver may be adding.

> I've become a huge proponent of where you can [...] using the same os to develop that you deploy in. In my case that would be some sort of Linux targeting servers. Individual Linuxes are pretty different in important ways, so it probably would have to be the exact same flavor. I've seen my colleagues spend a lot of time fighting even the Linuxes that target laptops to get basic things like USB-C, sleep, battery life, docking stations and projectors to work well, which I've read on HN were all solved about five years ago, but I still see colleagues fighting Linux quirks, so I guess a laptop running RHEL will be a terrible hassle. I would also have had to switch my OS every time I've changed employers (sometimes when changing teams), and I'd lose the accumulated muscle memory of over a decade on macOS, plus all of the platform specific tooling, plus the full Microsoft Office suite to interface with the non-tech part of the company. I just try to deliver everything I build as a docker image, that way I get pretty decent isolation from the particularities of the host OS and it's pretty easy to test locally under conditions very close to prod. I've switched to a M1 Pro notebook earlier this year and apart from having to specify the arch when running some docker images, I've hardly noticed a change, though I don't develop Linux or Windows desktop software (hence, no X or the like needed).

All while my M1 Pro is in the laptop with which I have yet to experience a sound of fans running.

MPS (Metal backend for PyTorch) is pretty poor integrated right now most of my custom models are crashing and pure Swift + MPSGraph versions is working 3-10x faster then PyTorch. So I'm pretty sure there is A LOT of optimizing and bug fixing before we can even consider PyTorch on apple devices Also, even for ops that do not crash, it often returns garbage. I built PyTorch from git this morning: >>> torch.arange(10, device="mps") tensor([0, 0, 0, 0, 0, 0, 0, 0, 0, 0], device='mps:0') >>> torch.ones(10, device="mps").type(torch.int32) tensor([1065353216, 1065353216, 1065353216, 1065353216, 1065353216, 1065353216, 1065353216, 1065353216, 1065353216, 1065353216], device='mps:0', dtype=torch.int32) Many bugs will probably be squashed over the coming weeks, so I am still very exited about MPS support. I have done some preliminary benchmarks with a spaCy transformer model and the speedup was 2.55x on an M1 Pro. Which is quite nice because transformers were already very fast on M1 Macs, thanks to the AMX units (PyTorch links against Accelerate, so the AMX units are used for matrix multiplication). If they can squeeze more performance out of M1 GPUs in the future, it will be a very nice speedup.

This may be true of 50-in-one megaprojects such as those from the likes of Uber and Google, but these machines are also beneficial to developers of more lean, focused, and optimized apps too. The few seconds my M1 Pro shaves off of incremental builds and the general improvements in IDE responsiveness compared to an upper-spec Intel mac pile up and become worth it quickly, not even mentioning quality of life improvements like far better battery life and being able to use the laptop in my lap comfortably. For the projects I work on, compile time has been declining with each hardware gen despite app complexity increasing significantly. I’m sure there are other devs who this is true for too. We’re not all 5 abstractions deep and nonchalant about heavy resource usage and bad engineering.

I very briefly had an M1 Pro laptop (unfortunately it came with a faulty keyboard, so now I'm waiting another month for the replacement to ship), and I was blown away by how fast it is. My daily driver is currently the 2015 MacBook Pro, and I hadn't realised what I was missing. Looking at benchmarks, it turns out that the M1 Pro is actually a full 8-9x faster than the 2015 model at multicore workloads (which most dev workflows will be): twice as fast single core, 4x more cores + 2 efficiency cores!

I had a similar experience to the OP with my previous Intel laptops - they've quickly get too hot and throttle and that would kill the performance. When they weren't throttling, performance was fine. For me the big win with the M1 Pro was that I could work flat out and not worry about it overheating and giving up. It's a lot easier for desktops to have adequate cooling and power.

I’ve seen some benchmarks that put the M1 Pro about on par with a desktop 5700X. Probably depends on the task. That performance level is maintained when the laptop is unplugged too, which is hard to find in a laptop. Most of them will throttle themselves pretty hard the moment you’re no longer tethered to a wall because otherwise they’ll burn through their batteries in a flash.

Unscientifically that puts the M1 Pro GPU at about 25% of the performance of a RTX 3080. Not too shabby... EDIT - this comment implies it's much faster: https://news.ycombinator.com/item?id=32679518 If that's correct then it's close to matching my 3080 (mobile).