- Some of you felt that I
was being dismissive of Apple during our Apple Silicon reaction video, and I can understand why. My skepticism of Apple's
vague and optimistic performance claims didn't
exactly scream "enthusiastic." But make no mistake: Now that it's in my hands,
I am super enthusiastic, and the Mac Mini represents not
just the least expensive way to bring an Apple Silicon
M1 SoC into your life, but also the most performant, thanks to its form factor and its cooling. So let's take the Apple Silicon
M1, or as I like to call it, the ASM1, for a cruise
and see if it's seaworthy. Today's video is brought
to you by Ridge Wallet. Ridge Wallet wants to redefine the wallet with its compact frame
and RFID-blocking plates. Check out how they can
keep your wallet bulge down and use offer code LINUS to get 10% off and free worldwide shipping. (electronic dance music) Apple's not known for
reinventing the wheel often, and in the Mac Mini's case, the wheel is very much the same
as it's always been: square. At least, on the outside. There is one telltale
feature that lets you know that you're looking at
an M1-equipped Mini, and that's the fact
that there are only two Type C ports on the rear which
are capable of Thunderbolt 3, or because Thunderbolt is
actually an Intel trademark, more accurately, USB4. So the functionality is
all the same, mostly, but, it's got a different name, now that Intel released
the spec to the USB IF. That's totally not confusing at all. Thanks, Intel! Whatever we call them, both
Thunderbolt and USB Type-C hubs worked just fine on the new
ports, including devices like 10 gigabit network
adapters and display hubs, although the latter has a limitation where you can only
expand past two displays via display link adapters,
which are basically software-driven displays running
over USB with performance that's exactly as good as
software-driven sounds. Now, if you were hoping to run an external GPU to fix that, no luck. We hooked up a Razer Core X just to see, but it wouldn't detect the GPU inside, and Anthony thinks this is
probably to do with differences in how X86 and ARM address displays. So it's possible that this
will change in the future with either Mac-optimized GPU's, or some tweaks to next generation M Series SoCs, but for now, nada. So that's bad news for
Battlestation enthusiasts who love walls of monitors,
but the silver lining is that at least it's possible
to go beyond two displays, even if it will take up more of your precious USB ports to do so. There are some other workarounds for multiple monitor setups, of course, like running with an iPad inside car mode, or by using Airplay, and a DisplayPort Multi-Stream Transport hub may also work, but those haven't been
relevant for a long time, and we didn't one on hand to test with. This lack of external expansion, presumably because of the
limited I/O on M1 SoC, probably won't be a big
deal for most people, but because only one of the USB4 ports can be used for display port at a time, and you're stuck with gigabit
ethernet out of the box, until the rumored 10-gig versions show up, it's pretty easy to see how a
power user could overload it. If you want two displays, a
connection to a high-speed network storage device, a card reader, and a USB keyboard or
something, you've exhausted your expansion options,
unless you resort to hubs. Thankfully, the Mac Mini
does have Bluetooth 5.0, as well as WiFi 6, making
its wireless connectivity at least as good as wired
for the typical desktop user. Now, I've already taken apart our Mac Mini on our ShortCircuit unboxing channel, but in case you missed it, there's no internal expansion
either, despite the presence of the same access hatch
as previous generations. And it really is a shame, because the PCB, and especially the SoC, are comically small for the
enclosure that they're in, and compatibility permitting,
Apple could have easily thrown in a storage bay or made the thing quite a lot smaller without
sacrificing any more usability, which brings us neatly into what it's like to use the Mac Mini. Fundamentally, nothing has
changed over the last generation. Plug in power, keyboard,
mouse, monitor, press go. However, there are some things
that are very different. One of the big ones is the ability to run not just Intel applications via Rosetta 2, but also iPad and iPhone apps. Installing iOS apps is as simple
as firing up the App Store, searching for the app you
want, then choosing iPhone and iPad apps on the subheader. iOS apps installed this way just work, but, there's a big asterisk. While you may seemingly get
full mouse control over them, a resizable window, and
the ability for files to import and export to regular folders, using these apps is a
little bit kludgy for now. Because your mouse is emulating a finger, you need to long press
to get context menus, click and drag isn't quite the same, resizing window panes just
isn't a thing, and gestures are unfortunately awkwardly
handled by keystrokes, or by pretending the
touch pad on a MacBook or a Magic Trackpad are touchscreens. This clunkiness makes
LumaFusion, for example, much less intuitive to
use than it appears. You wanna be able to click
and drag stuff around on the timeline, but you
have to click and hold. None of the tools have tool tips, and you can't rearrange the workspace beyond what the app's presets allow. To be clear, this is certainly way better than not having access to a mobile app that you really like on the desktop, but it clearly has a
generation or two to go before it has that
characteristic Apple polish. And anyway, why use LumaFusion
when Adobe exists, right? This is where Rosetta 2 comes in, and shockingly, it just works. A window will pop up and
ask to install Rosetta the first time you run an
Intel app, and after that, it's totally invisible
to the user, that is, assuming that they don't run
any comparative benchmarks, which is surprisingly easy to do. You can choose to run
any universal binary app, including even built-in apps like Safari, by going to Get Info and then
ticking "Open using Rosetta." This is a really useful feature, not just for performance comparisons, but also for testing or accessing features that are only present in the
Intel versions of some apps. You can also run Terminal apps in Rosetta by adding this prefix to whatever command you're trying to run, something
that you'll have to do for home brew for now, but MacPorts has a native version already,
if that works for you. If you're ever curious what
architecture an app is using, Activity Monitor has a
new "Architecture" column showing you exactly that. Neat! But you'll quickly notice
that just having an Intel app running can have a performance penalty. Now, we're looking at only
one to two percent CPU wasted by having Creative Cloud
Background apps running, but if you're running a MacBook and battery life is a concern, it's just something that
you should keep an eye on. Rosetta provides us with some interesting opportunities here. Not only can we run macOS applications written for Intel CPU's; we can also run Windows applications by
using CrossOver Office. It does have the same limitations
as Wine does on Linux, meaning that things
like the Windows version of Microsoft Office won't run
because of its always-on DRM. But if your application runs,
this might save your butt now that running Windows
directly in Boot Camp isn't really a thing anymore. Ah, Boot Camp. It's unclear as at the time of writing whether Parallels' upcoming support for M1 will enable X86 Windows
somehow through Rosetta 2, or if we're just gonna
be stuck with macOS, Linux, and Windows for
ARM, the last of which, assuming that you
actually wanted to use it, you still can't even officially download or license for anything but a prebuilt Windows machine running an ARM CPU. We also found some Mac
programs that failed to run, of course, but in our experience so far, this has been for obvious reasons. Macs Fan Control, for example, relies on the System Management
Controller found in Intel Macs, and while system monitor apps might run, they might not know
what sensors to look at until they've been updated
specifically for Apple Silicon. Also, don't expect an
Apple Silicon version of Intel Power Gadget anytime soon. As for the system itself, at first blush, it is just ridiculously responsive. All those videos of people
launching every app on the dock and the thing just doing it,
(Linus snaps) that's the real deal. The sheer level of
optimization in macOS Big Sur is such that even if you
were to fire up Prime95 and let it rip, you would
barely even notice it in the UI. In fact, doing so doesn't
even make the fan ramp up. It keeps cruising along at
its 1700 RPM idle state, and good luck hearing it
over the ambient noise. But wait a minute, if this
thing doesn't break a sweat, how will we sell all our
sweat bands on lttstore.com? Yeah, we snuck that in there. Temperature sensors
weren't in the usual spots on the M1-based Mac Mini, so
it wasn't until third-party utility iStatistica was updated for the M1 that we were able to get actual readings. We're glad to have them, but
without logging functionality, we're gonna have to make do
with numbers rather than charts. While idling, we were
sitting at under 30 degrees, and when we hit it with
a full Prime95 load, things get really interesting. Without the fan ramping up whatsoever, this thing is hovering
in the low 70s at worst. In fact, if we look at
the total system power, our Mac Mini hovers around
30 watts at its maximum load, which is ridiculous, given, spoiler alert, just how fast it is. Now, speed is where things
are going to get contentious. I called out Apple's
marketing on performance, and Apple fans came
crawling out of every corner to tell me how wrong I was
and how revolutionary it was before they'd seen any real numbers, but you won't get any apology
from me, and here's why. Apple's marketing was bull
(bleep), and they know it. In the short time since the
announcement, Apple has already softened their position from
"world's fastest CPU core" to "the world's fastest CPU
core in low-power silicon," which is fine, and more in
line with what we expected, and more in line with what we got. So tell you what, don't
take my word for it. Take Apple's word for it. You guys are good at that. Or, if you do want my word
for it, keep on watching. We're gonna be comparing
our M1-equipped Mac Mini to its spec'd out Intel counterpart, along with our 27-inch iMac,
a Ryzen-based desktop PC, and the fanless M1 MacBook Air. Get subscribed, by the
way, so you don't miss our review of the Air alongside the Pro, because that is gonna be (bell dings) a tougher decision than you might think. Compared to the 2020 iMac with
the Core i9, we are sitting at about half of the Cinebench
multi-threaded score, but, the M1 chip is much
faster than the Core i7 in our previous generation Mac Mini, and remember, that's a
configuration that costs $400 more. That's especially
impressive when you consider that this is a CPU with no hyper-threading and with fewer cores. As for single-threaded performance, even the lowly MacBook Air, which is the slowest of
the three new M1 machines, nearly meets the performance
of a desktop Ryzen 5000 and completely embarrasses
the Core i9 iMac. This shows that if
Apple had included eight performance cores in the Mac
Mini instead of just four, the M1 could have met or
beaten the Core i9's 20 threads in multi-threaded performance at the cost of maybe having to ramp
up the fan just a smidge. This is huge. Also huge is compilation performance. At half the compile time
in this Xcode test project, it absolutely crushed
the Intel-based Mac Mini, and once again, the MacBook
Air isn't far off either, primarily because this
is a shorter compile test than we would usually run due
to some early teething issues with compiling Firefox or
Chromium on Apple Silicon. Video transcoding is a bit hit or miss, depending on what you're doing. If you're using your CPU to encode H.264, our M1 Mini handily beats its predecessor, but shows its lack of
threads compared to our iMac. If we use hardware encoding, though, the M1-equipped Mac
Mini smokes our i7 Mini and even draws near the
performance of the iMac. Not too shabby. When we switch over to H.265,
things get even more dramatic. I mean, yes, CPU encoding is much, much slower than the competition, but look at that hardware
accelerated encoding! M1 is significantly faster,
not only than the iMac, but then its own H.264 encoder. And while it's not quite
NVENC-level performance, it's clear that Apple's prioritization of HEVC performance on
their mobile devices is now paying off in
spades here on the desktop. We tried to get some
deep learning tests going to give the Neural Engine a
workout, but unfortunately, the Apple Silicon-optimized
TensorFlow package refused to run on most of the
models that we threw at it. This isn't too surprising,
given that it's version alpha0, but it's still disappointing,
because it could have given us a taste of one of the big advantages that the M1 is supposed
to bring to the party. Performance in Rosetta 2,
as has been widely reported by now, is really solid. In our testing, it sat at
about 2/3 to 3/4 of full speed, with Blender putting the M1-equipped Macs at well under half the
speed of our 27-inch iMac, but within spitting distance
of the Intel Mac Mini. And while Blender isn't yet
available for Apple Silicon, if we assume that it scales
the way Cinebench does, we could expect to
shave nearly two minutes off the BMW render time. GPU focus tests don't show
much disadvantage at all with Rosetta, which is pretty
sick, and the integrated M1 GPU cores reach nearly
half the performance of the dedicated Radeon
5700 XT in our 27-inch iMac. That is much faster
(Linus laughs) than the UHD graphics
in our Intel Mac Mini. The Adobe Suite,
particularly Premiere Pro, looks like a poor fit for
Apple Silicon at first glance, but there's more to the story here. Considering the Core i7-based
Mac Mini scored worse and very close in Photoshop means that once these apps are updated for M1, the new Macs should
completely dominate anything in their power class, and
probably price class, too. That's a new one. So then, wait, why doesn't Rosetta suck as much as other X86 emulators
that we've seen in the past, like Microsoft's, for example. Apparently, this is mostly
down to hardware tweaks made in Silicon to accelerate common
X86 load/store instructions. Essentially, Apple Silicon
can take X86-like instructions and process them directly rather than having to
fully translate to ARM. As of right now, the only
ARM CPU that has the ability to run X86 programs like this is the M1. So, sorry, Microsoft. Windows for ARM devices
are still gonna suck. Of course, that's not all. It wouldn't be an LTT video without trying to play
games on it, would it? Most games available on macOS still only have Intel versions
running under Rosetta, and we're getting between 30 to 60 FPS in the Mac-ported Tomb Raiders. CS:GO, unfortunately,
wouldn't load for us. It just hung on the title screen. And hey, if you're wondering
whether Rocket League will run through CrossOver, you're in
luck, as long as you're okay with a low frame rate and some flickering. But, while those games
are running in Rosetta, there is actually one game that has been updated for Apple Silicon, giving us a glimpse into the future, World of Warcraft Shadow Lands. Now, I'll admit that this is not a game that we've even considered using
for our benchmarking suite, because there's such a
variety of locales and assets that there's just no way
to do it consistently. Also, there's no canned benchmark, so take this with a grain of salt. And a wait there's, no, where's the graph? All righty, no graph. So at 1080P using the recommended
quality level of five, the game appears to run at a solid 60 FPS. Yup, 60 FPS, and no more. I guess DSync is locked on,
'cause the toggle is broken. Oh well, it's still really
cool to see that it runs pretty okay, aside from some
minor graphical glitches due to upscaling from
1080P to native resolution. So then, Mac Mini, or more accurately, M1. Does it live up to the
expectations that Apple set for us? I guess that really depends on whether you're more
of a spirit of the law or letter of the law kinda character. For my part, I still take issue with the overly vague
promises that Apple made about the performance of
their first "desktop silicon," and the most frustrating
thing about the situation is how completely unnecessary it was. If Apple had just come
out on stage and said, "Hey, we've got X86 emulation "that's between 50 to 75%
of native performance," I would have been absolutely blown away, and I probably still
would have been skeptical. X86 emulation on Windows is balls slow, not to mention that it still
doesn't support 64-bit apps, which Apple has handily
taken care of here. The problem for me is that they decided to hold back on specifics,
and from my experience, the less concrete data a
company is willing to give, the less confident they
appear in their product. And so, that quite
unfairly painted a picture that was simultaneously
unrealistically rosy for the Apple fans who take
their claims at face value and far too bleak for
those whose BS detectors wouldn't stop going off throughout
the entire announcement. At the end of the day, though, what Apple has with the 2020
Mac Mini is a good product. It's affordable, it's powerful,
it's quiet, it sips power, and it is a great way
for, let's say a developer who missed out on the transition kit to get into the ecosystem
and start coding. It's not what I want. It lacks far too much in expansion, and especially in high-speed connectivity to be very useful for me, but that's not to say
that it can't ever be. As I already said in my
coverage of the announcement, the M1 is just the first
step into a bold new world where Apple is, for the first time ever, fully vertically integrated on the Mac. They have complete control
over their own destiny with only commodity parts
like RAM, NAND flash, and display panels sourced
from third parties. I absolutely cannot wait
to see what 2021 brings. Does anyone else think that the naming is because they're going to be
putting the screws to Intel? You know, M!1, M2, M4, just
bigger and bigger screws? Anyone? Bueller, Bueller? Sponsor? Oh yeah, sponsor! Blinkist takes the best insights and most important information
from thousands of non-fiction books and condenses them
into just 15 minutes. You can either read or
listen from the app, and you'd be amazed how much you can learn in just 15 minutes. Offline downloads are available, and they're perfect for
traveling during commutes. Some of the most popular
titles include The AI Economy, which is a book that discusses the rise of artificial intelligence
and how it's gonna affect our jobs, investments, and lifestyle, and What To Do When
Machines Do Everything, which explains how automation
can be incorporated into current business
models and what to expect as whole industries adopt
this new wave of automation. It's available on the Google Play Store and Apple's App Store,
and the first 100 of you to go to blinkist.com/linustechtips
are gonna get unlimited access for one week
to try it out and 25% off. Also, they now have
full-length audio books. Premium subscribers get special pricing, up to 65% off the regular retail price. So don't wait. Thanks for watching, guys. If you enjoyed this video, maybe go check out our reaction
to the Apple M1 announcement to get a little more context
for why we felt this launch might not be all sunshine and
rainbows like Apple claimed.
I hate these thumbnails...
Whoa they showed my app in this. So cool, never thought I’d be in an LTT video.
I am somewhat confused though, given that the iPhone 8 has faster single core performance than my Intel MacBook Pro, why are tech reviewers surprised that Apple’s CPU offerings are really good? Did they expect when you put it in a better
cookedcooled chassis that it would perform worse?Honest question, because I feel like anyone who’s kept up with Apple (which I would expect tech reviewers to) is like “yeah no shit they’re fast”
Came here for a discussion about the Mac Mini, scrolling half way down it’s still a debate about click bait and thumbnails. Feels bad man.
I watched this video early on Floatplane, he was pretty impressed by the benchmarks and performance he was seeing, and how well Rosetta 2 works, especially compared to previous attempts at x86 to ARM translation (Windows on ARM, for example)
I don’t know why people were angry at Linus that he was a bit skeptical about apple marketing. To be honest we should always sceptical about anything that any marketing says
lol i was close... off by a week... Comment on LTT video: "Apple Silicon Mac Announcement - Slow Motion Dumpster Fire"
Disappointed we didn’t get more comparisons of more CPUs. The M1 vs one intel iMac, one intel Mini, a desktop Ryzen 5... why not talk more about power/performance? Maybe compare it to some newly released low power chips from AMD/intel? Maybe that’s coming in the M1 MbP/Air review. Why not look at other similar form computer options, like the Mini, powered by intel or AMD CPUs, and go pound for pound, price per performance?
This was a disappointing video. I had higher expectations considering how long it took to release the review.
Edit: I mean they did calculations approximating performance if the M1 had 8 performance cores. Why didn’t they also look at power draw and compare it with other chips and look at their power draw for the same performance? Again, the more I think about this review the worse it gets.
Edit 2: and considering how much they split hairs with dollars and cents on performance and builds, does that it stand to reason that they may pick out a selection of Windows PCs that are designed like the mini, price them out accordingly, and bench them? Or at least approximate them? This video is really underwhelming.
It's rare a ltt video is posted here
still he did say that apple chart were misleading and change near the review date. from being the most powerful laptop cpu to most powerful low power laptop cpu.
could have just said "we have the best x86 to arm emulator with little to no lost" still would be great.