The Robservatory

Robservations on everything…

 

My impressions of the M1 MacBook Pro

I recently received my Apple M1-powered 13" MacBook Pro, which is primarily going to be used for testing our apps on Apple silicon, and supporting customers using these machines. But that doesn't mean this is a work machine; it's a personal purchase as I'll use it for my own needs as well. (Thankfully, it only had a net cost of $33 after I sold my 16" MacBook Pro.)

By now, you've probably read a slew of stuff about both the MacBook Pro and its slightly-lighter MacBook Air cousin. Between unboxing videos, extensive benchmark suites, and multi-thousand-word reviews, there is no lack of coverage of these machines. (However, I will add that I did make a video of my MacBook Pro—with its 16GB of RAM—opening 75 apps in just over a minute. Not bad for an entry-level machine!)

I'm not going to try to replicate those reviews, because they do an excellent job of covering the new M1-powered Macs in a level of detail that I just don't have time to get into. Instead, here's what I'll be discussing…

  1. Why I chose the 13" MacBook Pro
  2. A few benchmark results of interest
  3. Rosetta and non-native apps
  4. Using iOS apps on macOS
  5. General discussion on performance
  6. The future of Apple silicon Macs

So why a MacBook Pro and not an Air?

Why I chose the 13" MacBook Pro

Basically, I chose the MacBook Pro because it has a Touch Bar—despite the fact that I hate the Touch Bar. But some of our apps make use of the Touch Bar, so I need it to support our users and test our apps.

Given I had to get the MacBook Pro, that really only left choices for RAM and storage size. On the RAM side, I went with 16GB, thinking that more is better in a machine where you can't ever add any. On the storage side, I stuck to the minimum 256GB SSD, because I don't store a lot of stuff on my laptops. That brought the total to $1499.

If I didn't have the "must have Touch Bar" constraint, I think I still would have chosen the MacBook Pro. At the base level, there's a $300 price gap, which buys you a slightly nicer display, the Touch Bar, an 8-core GPU (versus 7-core in the Air), supposedly better speakers (HDR) and microphone (studio quality).

If you want an 8-core GPU in the Air, though, the price delta drops to $200 (if you equalize both with 512GB of storage). And if you don't care about storage and are fine with a 256GB SSD in the MacBook Pro, then the price delta for an 8-core GPU machine is only $50. At that amount, it's a no-brainer to get the MacBook Pro.

I think the Air will be more than sufficient for almost anyone, but me being me, even if I were free to do so, I'd find it hard to buy a machine that was lacking a GPU core, regardless of what marginal performance difference it might make.

A few benchmark results of interest

When I got my machine, I still had my 16" MacBook Pro, along with my 2018 MacBook Air, and of course, my desktop iMac. I ran a slew of benchmark results, and they all—well, almost all*See openssl benchmarks—showed what you've read on other sites: These new entry-level M1 Macs may be the fastest single-core processors in existence, and they hold up well in multi-core tests as well.

General benchmarking

I'm not going to bother reproducing every test I ran here; it's just a whole slew of numbers, and you can find similar on many other sites. You can, however, download my full Excel spreadsheet if you wish. Instead, I'd like to focus on some of the more interesting results.

There is no doubt the M1 is a fast CPU. Using Geekbench 5's single core test, my M1 MBP scored 1731, which absolutely crushed the 2018 Air (790), 2019 MBP (1029), and even the 3.6GHz Core i9 in my 2019 iMac (1177). The results were the same in multi-core, where the 8-core M1 MBP (7524) expectedly crushed the dual-core Air (1603), but also the six-core (12-thread) MBP (5522). It also nearly matched my 8-core/16-thread iMac, which score 7661. Impressive stuff.

In my disk tests, the SSD in the M1 MBP was much faster than the one in my Air and my iMac, and slightly faster than the one in the 16" MBP. The GPU kept pace with the discrete Radeon on the 16" MBP, and absolutely destroyed the integrated GPUs in the Air and MBP. (The iMac's dedicated Radeon card was faster.) In short, these machines are wicked fast. With one apparent exception that may not even matter much…

openssl benchmarks

There's a simple CPU benchmark built into the openssl tool in macOS (which is really LibreSSL, not OpenSSL, but that's another story). Just type openssl speed in Terminal, press Return, and then wait. Eventually you'll get a summary showing the rate at which openssl can sign and verify keys using various cryptographic algorithms.

I've long used this test in my benchmarking, as it's stable over time and gives a good sense of progress in CPU speeds. And it's this test—and only this test—where the M1 MBP shows true weakness*True benchmark weakness, at least against its Intel-powered relatives. Here's a look at a few of the results from my testing:

2019 iMac2019 MBP2018 Air2020 M1 MBP
CPU Information3.6GHz Intel Core i9
8 cores/16 threads
95 watts
2.6GHz Intel Core i7
6 cores/12 threads
45 watts
1.6GHz Intel Core i5
2 cores/4 threads
7 watts
3.2GHz Apple M1
8 cores
10 watts
Single core openssl tests
DSA 2048 sign/sec4,1433,3962,5802,038
DSA 2048 verify/sec3,8603,2182,4731,873
RSA 4096 sign/sec17915612490
RSA 4096 verify/sec6,8275,8214,4403,645
Multi core openssl tests
DSA 2048 sign/sec28,45118,9184,8999,994
DSA 2048 verify/sec26,78418,3084,6279,127
RSA 4096 sign/sec1,256880219442
RSA 4096 verify/sec49,05333,3658,20517,803

As you can see, the M1 MBP is nowhere near the Intel chips' performance on these openssl benchmarks. Why is that? My theory is that it's due to a set of instructions specific to Intel's chips, the Intel® Advanced Encryption Standard (AES) New Instructions, or AES-NI for short. I dug into this about a year ago, as I was looking to explain a huge increase in openssl benchmark scores on some of my existing Macs. At that time, my theory was that Apple had enabled AES-NI in macOS 10.14.5.

And I think it's this same set of instructions—or more accurately, the lack of such instructions—that accounts for the M1 CPU's poor performance here relative to the Intel chips. From my prior article, here's how AES-NI works:

The new AES-NI instruction set is comprised of six new instructions that perform several compute intensive parts of the AES algorithm. These instructions can execute using significantly less clock cycles than a software solution.

I think it's pretty obvious that the M1 chip isn't doing anything special to accelerate the AES algorithm in openssl. In the single-core tests, my M1 MBP is running at about half the rate as that of my iMac, and it's even slower than the MacBook Air. In the multi-core tests, it's also quite slow, though its eight cores are enough to defeat the MacBook Air. But does it matter?

Does it matter?

Despite the poor results, I'm not sure they matter at all to the vast majority of macOS users. Why not? Consider this quote from Security Week:

In terms of security features, Apple says the M1 chip includes the latest generation Secure Enclave, a high-performance storage controller with AES encryption hardware, and hardware-verified secure boot. The company says it has built new security protections "deep into the code execution architecture of M1."

The Secure Enclave has AES encryption hardware, so I would expect it's very fast. It appears to me that openssl doesn't leverage the Secure Enclave (honestly, I don't even know if Apple allows such use), so there's no hardware acceleration for the encryption and decryption calculations in the openssl tests.

If you use openssl to, for example, encrypt huge archives, you'd be affected by this issue. Unless you do that, though, I don't think you'll notice these speed differences in day-to-day use—it's not like it takes longer to connect to an encrypted web site.

Rosetta and non-native apps

A hidden star in the Intel-to-Apple silicon transition is Apple's Rosetta technology, which allows you to run Intel-native code on the M1 chip. In my role with Many Tricks, Rosetta makes a huge difference for us: While we've already updated some of our apps to run natively on the M1 chip, we don't have to rush to get them all done, as Rosetta handles the hard work for us—the user simply sees an app that works as it did on Intel-powered Macs. But our apps are generally low in CPU and GPU usage, so Rosetta doesn't have to work very hard for us.

What kind of impact does Rosetta have on other apps?

Web browsers

As I write this, both Safari and Chrome have Apple silicon-native apps available; Firefox, Microsoft Edge, and Opera all run via Rosetta. (Chrome also has an Intel-specific build available; they don't distribute a universal binary.) I thought it'd be interesting to see how all of these browsers handle some web benchmarks, so I put them through their paces using four different tests.

Update: Firefox 84 Beta supports Apple silicon, so I re-ran my benchmark tests using that version as well—I've added its results to the table.

My benchmarking was pretty simple: I made each browser window the same size, insured no other apps were running, and ran each benchmark exactly once in each browser. So I wouldn't call my results robust, but they do clearly show the impact of browsing using a Rosetta-translated browser.

BrowserArchitectureJetStream 2MotionMarkspeedometerBasemark 3
ChromeApple1855252061,316
Firefox 84βApple121334201998
SafariApple2201,029208844
Safari Tech PreviewApple2151,3822111,355
ChromeIntel95342111631
EdgeIntel106339112805
FirefoxIntel7714196740
OperaIntel97316103647

Higher scores are better, and the best are highlighted in green. No surprise that Apple's browsers fair best, with the Safari Tech Preview scoring best in each test (except for JetStream 2, where it's only a couple of percentage points behind regular Safari.) Native Chrome easily bested non-native Chrome, roughly doubling most of the non-native app's scores.

Of the four Intel browsers, Firefox feels the slowest in actual use—perhaps reflective of its very low score in the MotionMark benchmark? The Apple-native beta version of Firefox is better, though it's not competitive in the JetStream 2 or MotionMark tests, it feels much faster than the non-native version in actual use.

In any event, if you're using an M1 Mac, you really should be using one of the native browsers; the Rosetta tax on web browsers is quite apparent.

X-Plane

How what about X-Plane, a very complex flight simulator with realistic weather and scenery? In case you haven't seen it, here's what X-Plane looks like:

But how would that run on an M1-powered Mac through Rosetta? As you may have suspected, the above video is X-Plane running in Rosetta. It was recorded on my M1 MacBook Pro, using X-Plane's internal movie recorder, though it doesn't capture sound. To get the audio, I used QuickTime Player to record the microphone (Apple no longer allows QuickTime to capture computer audio, which is a shame).

X-Plane was running at 1440x900 resolution, and averaging something at or above 60fps the whole time. While it was recording the screen and QuickTime was recording the audio. Amazing. (X-Plane will eventually be native on Apple's silicon.)

What you can't see—or rather, what you can't hear—in the video is what makes the M1 MacBook Pro probably the best Mac I've ever used for X-Plane: Fan noise. There isn't any. On any other Mac I've ever used, including my Core i9 iMac with the Radeon Pro Vega 48 video card, the fans starting going really loudly whenever I fly X-Plane. If I'm wearing headphones it's not so bad, but if I'm not, the fan noise is incredibly distracting.

And yet, on the lowest-end M1-chipped Mac, the fan simply doesn't come on while flying (at least, not that I've heard yet). The machine's chassis will get warm after a while, but never super hot like Intel-powered Mac laptops sometimes do. But I've yet to hear the fan come on during X-Plane, and overall, I've only heard it a couple times during various benchmark tests.

When the fan does come on, it's a much quieter fan than on any other Mac laptop—you can tell it's on, but if there's much ambient noise at all, it's quite hard to hear. When the fans are going on my 16" MacBook Pro, drowning it out would require extremely loud sounds. I am frankly amazed at how rarely the fan comes on, and how quietly it does so when needed.

I mostly live in the Apple ecosystem—I use Safari for web browsing, Mail for email, Messages, Music for music, etc. As such, I've found the M1-chipped MacBook Pro with Rosetta to be an excellent machine. However, if you rely on non-native software based on Electron or Java, or use peripherals that have their own drivers, you may have a much worse experience:

So far, I have experienced application crashes in Microsoft Edge, Outlook, WinZip and Logitech Camera Control. I got installation errors with Adobe Reader XI, Adobe Acrobat Reader DC, a Samsung SSD backup application, and Xbox 360 Controller for Mac.

The author has some valid criticisms, though things will improve over time. Apps based on Electron, for instance, will benefit when they include Electron 11, as it now supports Apple silicon. Personally, the big hole in my toolset right now is Homebrew, the package manager that lets you easily install a huge array of Unix apps and tools. Just a couple days ago, though, a new release works on Big Sur (which was one issue) and can function on M1 Macs. (I rely on Homebrew so much that I've made a donation to help them out.)

It's early in the Apple silicon transition, and generally speaking, I'm thrilled with how well everything "just works." But I also have an Intel-powered iMac running Mojave that's my main work machine, so I can usually work around the bits that don't work well. If I could have only one Mac right now, it probably wouldn't be an ARM-powered Mac, just because it is so early in the transition.

Using iOS apps on macOS

Don't. It's a horrid experience. Controls are strange, windows are strange, the apps don't behave anything like a Mac app. They're just…wrong.

Ok, so that's a bit of an overstatement. If you have some iOS game or app that you really want to use on your Mac, and that app's control interface translates well to a mouse, then it can be kind of nice to be able to use that app on your Mac. As an example of an app I found that works well, here's a screenshot from Mini Motorways, one of the "keeper" games from my recent deep dive into the Apple Arcade:

In some ways, it actually works better on macOS than it does on iOS, as the mouse-drag interface works well for the dragging required in this game. On the iPad, you have to tap the screen to toggle between overview mode and road building mode; on the Mac, click-and-drag builds a road without requiring a click to switch modes.

But things I take for granted in most Mac apps, like a Preferences screen for settings, are notably missing. Instead of knowing I can find an app's settings in one spot, each time I installed an iOS app, I had to look around to see how to configure it. Mini Motorways also initially launched in full screen mode, but it (thankfully) remembers my non-fullscreen window size preference between relaunches.

I haven't tried any productivity apps, and I don't think I will: Given how frustrated I became trying to interact with simple games, I can't imagine my frustration level with more complex apps. For now, I view iOS support as a nice way to run some simple apps—probably games—on your Mac that you enjoy using on your iPad. Beyond that, it really feels like Apple's, well, mixing apples and oranges with iOS support on ARM Macs.

General discussion on performance

I've had my M1 MacBook Pro for a few weeks now, and I can confidently say this is the best-performing Mac laptop I've ever owned—and not just because of its benchmark scores. It's that the entire system has been designed to take advantage of Apple's homebuilt powerful but low-power CPU. As I use the machine, I have to keep reminding myself that it's the basically the lowest-end ARM chipped Mac you'll ever be able to buy.

One of the more amazing aspects of the machine is its battery life. Apple claims up to 17 hours of wireless web work (and 20 hours of movie watching), and I have no reason to doubt their claims based on my early use of the machine. While I was setting it up, I downloaded and installed probably 60GB worth of apps and data, including Xcode, which is a notoriously-bad battery killer. I did this all on battery power, and when it was done, hours later, I'd only used about 25% of the battery. There's never been an Apple laptop with battery life like this.

It's been so good, in fact, that I've moved the battery indicator out of the menu bar (where I'm in need of more space) and into the Control Center.

There are some caveats to battery life, however. Apps running via Rosetta seem to use more power, which makes some sense. And if you run apps that are more intense than web browsing, editing text files, and emailing, you will probably see lower battery life. But if that's the case, you were seeing those same reductions (Rosetta apps excluded) on Intel-powered Macs, too. And here, you're starting with a battery life that's much longer than that of any previous Mac laptop.

The GPU does a great job with graphics, especially given this machine is a lower-end laptop. Not only does it push good frame rates, but they're silent frame rates—the fan rarely came on, and when it did, it's a very quiet fan compared to my 16" MacBook Pro.

Overall, it's amazing what Apple's been able to deliver in a chipset that only draws 10 watts—barely three watts more than my MacBook Air, and less than one-quarter of the 45 watts the 16" MacBook Pro uses. And it simply destroys the performance of either of those Macs.

The M1 machines combine incredible battery life with CPU and GPU performance that's way above anything Apple's offered in Intel-powered laptops, and do so in a design that either eliminates the fan completely (MacBook Air), or makes it so quiet that it's no longer a distraction. Unless I had some task that absolutely demanded Intel power (virtualizing Windows and/or Boot Camp is about all that comes to mind), I can't see buying an Intel-based Mac laptop going forward.

The future of Apple silicon Macs

Given the M1 Macs are Apple's lowest-end machines, one can't help but be excited about what performance we may see in future higher-end Apple silicon powered Macs. However, there are a lot of unknowns, primarily around cores, memory, and graphics. The M1 chip comes in only a few configurations:

  • 8-core CPU with 7-core GPU and 8GB or 16GB of memory
  • 8-core CPU with 8-core GPU and 8GB or 16GB of memory

There are options for the amount of storage space, but for CPU, GPU, and RAM, that's it. Clearly this won't work for higher-end Macs. In particular, users of higher-end machines will want more cores, additional RAM, and faster dedicated video cards. How Apple will (or maybe won't?) deliver on these requirements remains to be seen. The M1 is an integrated solution, where the CPU, GPU, and memory are built as one, making it impossible to add RAM in the traditional manner.

Will future higher-end Macs have a mix of built-in and add-on memory? Will they support discrete video cards, with the built-in GPU taking the role of Intel's integrated graphics in Intel-powered Macs? How many cores can Apple build into the next generation of their CPU? It will be interesting to see what Apple does for the higher-end Macs, but if the performance of these "low end" machines is any indicator, I think the future of high-end Macs looks pretty good. But only time will tell.

For now, I'm very happy with my "low end" MacBook Pro that handily outpaces the "high end" 16" MacBook Pro it replaced—with better battery life, no fan noise, much less weight, and much better performance across the board (and $800 less expensive), it was a no brainer upgrade.

6 Comments

Add a Comment
  1. You can use SoundFlower if you want to record computer audio in Quicktime. I kind of futzed my way through it without really knowing what I'm doing but I've got it working now.

    1. There wasn't a version of Soundflower that worked on an M1 Mac when I tried. As I finished writing this, though, a beta of Audio Hijack came out, and it works well.

      -rob.

  2. I actually have a bunch of productivity iOS apps that I LOVE running on my Macbook Air M1. Mostly investment charting and banking apps. But also YouTube oddly (I vastly prefer the iOS interface; and ads seem more sane than on the web version).

    Sadly I had to install most of these apps via iMazing because for some reason they cannot be downloaded from the Mac App Store app. If this means developers are blocking us; I really wish they would stop doing that. These apps are great! I have bought them and use them often on my iPad; so downloaded them from my iPad using iMazing. For some financial and shopping apps; it's really nice to have a kind of "silo" around them instead of using Safari.

    For the apps I'm using I find the UI identical to using them on my iPad. A lot of iPad apps basically only implement single taps, and the mouse click works identically (but with more precision). Occasionally I need to scroll or zoom and the trackpad does both perfectly.

    Overall iOS apps are the biggest surprise win for my new M1 Mac. Absolutely love this feature!

  3. Using OpenSSL (LibreSSL) might not be on any significance since Apple only provides it for backwards compatibility. Apple has not used OpenSSL for many years, instead Apple created its own security libraries (coreCrypto and related libraries) to perform encryption functions so it seems most likely that these libraries are the ones that are accelerated in Apple Silicon.

    Some of the related crypto library/services in macOS:

    Common Crypto Library
    Security Transforms API
    CryptoKit
    App Transport Security (ATS)
    Secure Transport API
    Others...

  4. The MacPorts OpenSSL port is much faster than the system-supplied binary, so it's not hardware (single-thread):

    MacPorts: (sign/s verify/s)
    dsa 2048 bits 5193.2 5529.9
    rsa 4096 bits 282.4 19585.0

    System: (sign/s verify/s)
    dsa 2048 bits 2033.0 1855.2
    rsa 4096 bits 90.2 3640.5

  5. Apple benefits greatly from a move to Apple silicon in Mac's. It makes their entire hardware line up more uniform and compatible across platforms. But I don't think it actually will bring a whole lot more users to Mac's over Windows PC's. At least not early on as the Apple silicon only closes off some advantages of using a X86 CPU in Mac's. No this won't bother the core Apple only ecosystem user, but it might turn off those professionals who may see this as closing too many doors. Apple does love control and there are benefits to that control as we have seen in performance gains with M1. But performance isn't the only measure of success and it does remain to be seen if this helps or hurts Mac sales.

Leave a Reply

Your email address will not be published. Required fields are marked *

The Robservatory © 2021 • Privacy Policy Built from the Frontier theme