Upgrading to an M1 MacBook Air

This summer I ended the year with a little extra cash on hand than I expected from my yearly expenses. As I am prone to do, I got the desire to splurge. My mind went rolling. A drone? A new PS5 (I still want that one, despite my inconsistency using my PS4)? A new iPad? My last iPad is from 2013, and it doesn’t even get updates anymore. That’s when my eyes got set on an iPad Pro.

Upon a bit of research, I learned that it can’t do the things I need it to do (coding, modeling, etc.), but it isn’t for lack of power. The iPad Pro has Apple’s newest self designed chip, M1. Realistically, the chip wasn’t that revolutionary. It was really just the next generation of Apple’s iPhones and iPads that had their own Apple made chip. Over the years, its performance has gotten progressively better, even outpacing Intel’s own chips. The revolution is that it is now the chip in Apple’s Mac computers. Now, it’s marketed as the power of the Mac inside of the iPad (not that you can do anything with it), but really, it’s the power of the iPad inside of the Mac.

A graph of Apple’s in-house chip’s performance change with time compared to Intel.

Objectively, the M1 chip isn’t some power house of core’s and ram. It’s actually rather tame by today’s standards. It’s true magic comes from it’s efficiency that it gets from integrating the CPU, GPU, and RAM into one chip. Everything is faster and more efficient. This has lead to Apple’s lowest Mac’s matching (and sometimes besting) it’s highest end intel alternatives. Don’t get me wrong, it isn’t invincible. It can’t handle huge graphics needs or extreme computing, but anything you could do on Apple’s MacBook Pro 16 inch you can do on the M1 (even on the MacBook Air, without a fan).

That said, the M1 isn’t meant to replace Apple’s highest end Macs. In fact, it’s highly rumored that a much more powered M1X chip, with double or even quadruple graphics power, is imminent. This is what I want. It was highly speculated that these would be announced at Apple’s June event, but it never happened. We may be looking at an October or November release. I didn’t want to wait (out of excitement and promising my little brother my gaming laptop to use with his new VR headset). I decided to settle on M1, and when the next one comes out, I’ll assess my finances and decide if I want to upgrade (either with a second device or selling this one).

I decided to buy the MacBook Air, which may be a surprise to anyone familiar with the Air. It isn’t known for its power. Historically, it is slow and way too hot because of how compact it is. With this new chip, it easily rivals the Pro and even the desktop alternatives. The trick is it doesn’t have a fan, so when it comes to heavy prolonged usage you will see minimal performance loss due to thermal throttling. That’s harder to achieve than you think, and it’s likely minimal power loss anyway. Everything you can do on the Pro you can do on the Air nearly perfectly. If something crashes on the Air, it will probably crash in the Pro. The only thing I haven’t managed to do is play Call of Duty because the CPU gets too hot. I can easily run my research modeling and even do 4k video editing. There are plenty of benchmarks showing that on YouTube.

The purpose of this post is to compare the performance of this computer to my other computers. I’ve taken 4 of the most intense tasks I’ve done in the last year or so and reproduced them on each computer with timestamps to measure each computer’s progress. Before we get into those, let’s stop and consider which computers I have to compare.

Computer	CPU	GPU	RAM (GB)
Predator Helios 300	Intel Core i7 (6 Cores)	NVIDIA GeForce RTX 2060 (6GB RAM)	16
Surface Book (First Generation)	Intel Core i5 (4 Cores)	NVIDIA GeForce dGPU (1GB)	8
Mac Mini 2014	Intel Core i5 (2 Cores)	Intel Iris (1.5 GB)	8
M1 MacBook Air	M1 (8c)	M1 (8c*)	16

The computers I work on (or have worked on)

My Predator Helios 300 computer was a gaming laptop I bought back in 2018. I don’t do a lot of gaming on PC (I’m a playstation player), but I wanted to give it a try. Plus, I thought it would be nice power to have, especially for my TV. Sadly, it wasn’t as great as I hoped. Despite its spec-ed out graphics, it blipped a lot connected to my 4k TV. It struggled with Photoshop and Premiere pro. You might ask, why use both at the same time? Well, I processed a video while making clip art for YouTube. Little did I know, the hours it took to process a video was not normal, or at least not anymore with the M1. This computer is still great, as you will see in the benchmarks; it comes the closest to the M1 in many aspects. Nevertheless, I was itching for an upgrade and my little brother wanted a gaming PC for the Oculus Rift he got for Xmas. That’s why I decided not to hold off for the M1X.

The Surface Book was the first generation tablet-laptop combo from Microsoft that I got in 2016. It is a great computer in a lot of ways. The screen was beautiful, and the sound was amazing. It was great for note taking and had a good processor for it’s time. Considering it’s compact size, it had decent graphics (1GB RAM) too. I paid for the dedicated graphics in the keyboard. That was still 5 years ago, and as you will see, it has aged. It also has (and had) it’s share of issues. The screen always detached from it’s keyboard base. It got very hot, and that caused the lithium batteries to expand and wrap the screen. I used it that way for 2 or more years because I assumed it was the screen that was thermally warped. I realized while searching for a new laptop that it was the batteries that was the issue. The computer wasn’t safe to keep using. I ended up ripping the screen off and taking the batteries out. It still works (with a broken but visible screen). Still, it doesn’t work great for my everyday use. It’s slow for research and virtually useless for Youtube. YouTube was another reason I wanted the M1, but I will get to that.

Lastly, I have access to a Mac Mini (2014 edition) in our research lab. It’s been a decent computer for the five years I’ve been here. It might be time for Catherine to consider upgrading (and the M1 isn’t a bad upgrade 😉). In fact, the M1 seems to have double the performance of the 2018 Mac Mini based on most benchmarks. It has its own dedicated graphics (1.5GB RAM), but it has only two cores in its processor. Those cores are easily used. For these tests, I made sure to close out over as much as I could (even OneDrive and google remote desktop). Of course, I don’t think it’s wrong to expect some level of multitasking on the work computer–something it isn’t very good at. One of my goals in this post is to demonstrate that an upgrade is warranted. I won’t lie (I can’t without fudging the results); it isn’t the worst computer in the mix. Then again, that’s a low bar. The Mac Mini is the absolute best value Mac. It has the same specs as the MacBooks and other M1 Macs, but it’s the most inexpensive and with better cooling than the laptops. Sure, the boost in performance is minimal, but it’s the best way to optimize performance. You can also get up to 16GB of RAM to help future proof the laptop. It’s worth noting, while the M1 is the first generation of it’s kind, the kind of jump in performance we and efficiency is largely due to the transition to the integrated chip. The new chips will always be better, but this is going to maintain its performance edge for quite a while. If you’re willing to invest a bit more, there will likely be a Mac Mini Pro with the M1X by early next year (if not late this year), but you don’t need to spend that much to see a huge boost in performance.

I wend the M1 MacBook Air 8c CPU and 7c GPU. The tests I have here are for the same specs with 8c GPU. I bought a spec-ed out Air but decided it wasn’t worth spending so much more for a minimal performance boost. 7c GPU isn’t going to be out-down in any significant way with one more core. I started to get an 8GB RAM option; literally bought it and returned it. I figured it was worth testing it, but I decided it wasn’t worth the effort. Everything I’ve heard suggests the M1 is really good at optimizing with less to give the performance of what normally requires more. Still, I didn’t want to risk it. I don’t plan to stick with this as my main laptop for an extended period of time (either get a secondary device or sell this one), but I also want to be prepared if that ends up needing to happen. I hate the idea of upgrading to lower specs (which I’ve done with the GPU). I’d like to be able to do more as time progresses. I’d also like room to have a solid virtual Windows Machine if need be). I ended up getting a refurbished base MacBook Air (i.e. 7c GPU, 256GB SSD) but with double the base RAM (i.e. 16GB vs 8GB) and I did it for only $140 USD more. Normally double RAM costs $200, but I got 10% discount in the Apple Education Store and a larger discount because is refurbished. I could run these specs and see if the single core makes a difference, but I don’t think it’s worth it. These results still prove my point that an upgrade was warranted.

My first test wasn’t an intense task (despite what I said before). I wanted to compare performance on basic tasks, so I timed how long each computer took to plot my most recent results for my Pluto research. The table below shows the M1 doing the best followed by my Predator, the Mini, then the Surface Book. This one was easy enough to redo, so I also show the results for my final MacBook Air (with 7c GPU) and it was about 10-20ms slower. These were all done with no multitasking. That said, I did have a statistics program running on the Macs to see what this (and the other tests) used CPU, GPU and Memory-wise. All in all, this may seem small, but small things add up. The M1 chip makes everyday tasks way quicker. This is one of it’s best features because was much as we love power, we aren’t running high end tasks 24-7.

Computer	Time (ms)
Predator Helios 300	172
Surface Book (First Generation)	313
Mac Mini 2014	287
M1 MacBook Air	121/~140

Time to plot some figures I made to present my most recent results.

Now let’s take a step back from my Pluto Results and see what it was like doing the research. In this first tasks, I timed how long it took to load the DEM data from .txt files into MATLAB, saving them as sorted .mat files to use to get results. Every computer has an SSD main drive (which I made a point to run the models on for optimal performance), except for the Mac Mini which has a dated “Fusion Drive”. I’m not sure what that is exactly. In addition to hard drive speed, this is just a matter of loading and processing large amounts of data. As before, the M1 did the best, being more than 4x faster than my Surface Book and 15x to 25x faster than the Predator and Mini respectively. Note, “Step” is just a random location that I placed a time stamp. I didn’t have an easy metric to use for this comparison.

Once the .txt files were loaded into MATLAB, I loaded my crater list from ArcGIS. This was just a .txt file with crater sizes and locations. Then I use the size and location to pull DEM data from that crater (and it’s terrain). Then I turned the 2D DEM into 8 DEM profiles, saving these for each crater. I did this by crater, marking how long it took to process each crater. The craters are sorted from smallest (800) to largest (1), so you can see the computers speed through the first craters and slow as it gets to the larger craters that have large DEM files with longer processing times. The Surface Book is by far the worst. The others are more closely packed, but the M1 still out did the others at ~3.5hrs vs ~4.25hrs and ~5.25 hrs for the Predator and Mini respectively. This kind of speed is not trivial.

Loading DEM data for 800 craters (#1-800)

Lastly, we take a step away from my Pluto work to take another look at my Titan melt pond modeling. While the other tests look at data processing, this looks at modeling. These models never push the CPU to make processing, but the differences in performance are abundantly clear nevertheless. It was a surprise to see the 2 core CPU in the Mac Mini out perform the other computers, but it still pales in comparison to the M1 which completed a 5m model in less than half an hour. That is an hour shorter than the Mini, an 1.5hrs quicker than my gaming laptop, and more than 2 hours faster than my Surface Book. This was particularly interesting for a few reasons. I always wondered which computer was quickest with this model because I had to do it so many times. While my first round is finished, I will be doing this again which is why it is nice to have a faster computer to do it. Although, I wouldn’t mind that M1x for it too. For a bit more context, this model is an ice model that tracks a layer of water as it is freezing. The flattening of the lines isn’t a performance issue. As the ice thickness it takes more time steps to freeze at deeper depths. These results were by far the most shocking. I just didn’t expect the M1 to perform that much better. Honestly, I thought I made a mistake, but I don’t think it is. The computer is just that good.

Modeling an ice sheet over a liquid water layer as it freezes from 5m thick to 10m thick.

In conclusion, the M1 works wonders for with my research. My research isn’t the most performance heavy, but clearly it had room to improve. Part of me wonders if I even need a computer with M1X processor. I really want one in hopes it will help, but the MacBook Air already works wonders and without a fan to keep it cool. The beautiful, compact, quiet, and cool (temperature) MacBook Air is quite the computer. You don’t know just how annoying heat and fan is until you live life without it. Another reason this computer is so great, as I alluded to earlier, is because of how much easier it is for my YouTube work. My Predator worked. It was easy enough to edit, but it shuddered often and took a very long time to export the project into a final video file. Now, I am experimenting with filming and editing in 4k on a MacBook Air with no problem! It is so seamless. I have been in such a slump with YouTube and editing on the M1 MacBook Air was a genuine pleasure. I don’t see the M1X improving my YouTube needs, but it would expand my abilities for research. That may mean its faster with my current tasks or allow for even more intense task to consider exploring. My dream is to have a Mac powerful enough to do all my work and Windows based tasks on a virtual platform (e.g. ArcGIS). All we can do is wait and see. Will I upgrade or will I decide my MacBook Air is all I really need? Although, that doesn’t do much for you now does it? Maybe you ought to upgrade too.