Apple MacBook Pro 14 and 16 Review: Return to Form
theverge.com
theverge.com
It's a real shame.
1) works, fits
2) doesn't work but fits
3) could work but doesn't fit
4) doesn't work and doesn't fit
Of course theirs a whole other world of USBC where things don't quite work but here I don't think it's the worst, like an escalator breaking down but, hey, you still have stairs.
Finally replacing those dozen terrible USB port variants with a single good design is great, and solves problem (3). You’d finally only need one type of cable.
Routing all kinds of other protocols through that same port creates problem (2), and introduces a completely new kind of cable fragmentation. I think that’s orthogonal.
If only. Certain USB-C devices can be damaged simply by plugging it into an incompatible charger/cable. The original Nintendo Switch being the most popular example.
Old MBP16 from 2019: 1m13s
New MBP, M1Max: 46.71s
295 dependencies, including some common ones like hyper, serde and tokio.
These have good GPUs but poor gaming performance. They perform as good as the best for work related tasks but closer to midrange AMD/Nvidia products.
I wonder if this is a "if they build it they will come" and more games will be optimized for these chips? Obviously Apple also made some design decisions in focusing on "pro" work rather than gaming too.
The real issue is the combination of lacking native 32-bit support, restricted metal API, translation layer overhead, and Rosetta all taking a piece of the performance pie.
I agree about the other issues you bring up though. The lack of 32-bit support is going to be a real stability challenge for anyone who wants to take Mac gaming seriously, and without a decent graphics API like Vulkan it will be exceptionally hard to court developers. Interestingly, it seems like Rosetta's overhead could be the bottleneck for a lot of these games too: if you watch the performance of GPU-light games like CS:GO, people struggle to surpass 90fps even at 720p on the new machines. Since ARM will always be punching up the abstraction layer to emulate x86, I suspect this may always be the case for Apple Silicon.
Future still unclear but there's some GPU horsepower in every mac now.
To name a few I've played and enjoyed:
- The Pathless
- The Last Campfire
- Cozy Grove
- Alba: A Wildlife Adventure
- FANTASIAN
- Mutazione
The games don't _have_ to ship via Apple Arcade to take advantage of the cross-platform tools, but most of the ones that have done the work have come out on AA.
But ultimately the Mac will never be a first-class platform for most game devs.
Maybe someone who understands better than me can explain these observations:
1. DDR5 RAM is supposedly 36-38GBps bandwidth, DDR4 3200 is about 25.6GBps bandwidth. (Supposedly Alder Lake is the first chipset to support DDR5.) 2. Maybe the 200-400 GBps rate advertised is the entire bus bandwidth for the SoC? If this is actual memory bandwidth, then Apple increased that bus speed 5-15x! 3. Maybe Apple divided the RAM into additional banks with more lanes to increase the bandwidth?
I've been very irritated by the pricing, but if Apple figured out how to optimize the memory bandwidth this much then maybe it's worth the extra cost per GB.
Did I catch on to something majorly different here?
https://www.anandtech.com/show/17024/apple-m1-max-performanc...
This is constant with the advantage of the M1 Max being the added GPU cores.
>If this is actual memory bandwidth, then Apple increased that bus speed 5-15x!
Yes because GPU requires lots of bandwidth. Top of the line Nvidia GPU has ~ 1TB/s memory bandwidth.
>I've been very irritated by the pricing, but if Apple figured out how to optimize the memory bandwidth this much then maybe it's worth the extra cost per GB.
You are not only paying for memory aka LPDDR5 chip, you are also paying for the additional memory controller. Which has an implication on yield and die size cost. Whether you think it is worth or still irritated by that pricing is of course subjective.
Thanks for clarifying the extra memory controller, that's helpful too.
No. The single core CPU performance has zero relation to memory bandwidth.[0] i.e You do not gain or loss any performance going from M1 Pro 10 Core with ~200GB/s Memory bandwidth to M1 Max 10 Core with ~400GB/s bandwidth. The additional memory bandwidth are there only to support GPU performance.
The reason why CPU performance is fast is slightly complicated. But to over-simplify it to two point, Apple has access to all the common code path on their platform and can optimise of it. e.g NSObject release is 5 times faster than on x86. Second being Apple managed to do something the whole industry thought ( at the time ) was impossible, having a wide decode and kept them well fed. You can read more about it here [1] on HN.
> what it'll look like in a year or so with the M2/M2Pro/M2MAX then.
You can take a look at [2], which is a chart from Nuvia, started by ex-Apple CPU architects. Given the amount of lead time in CPU design ( 3-4 years ), you can pretty much bet Apple will follow the path as shown by Nuvia to reach a Geekbench Score of 2000 within next 2-3 years.
Hope this answer your questions.
[0] CPU Performance is always latency sensitive. Hence the amount of cache matter much more than memory bandwidth. And Apple's SoC has always had large amount of cache compared to rest of the industry, even when it is a Smartphone SoC compared to Desktop part.
[1] https://news.ycombinator.com/item?id=25257932
[2] https://www.anandtech.com/show/15967/nuvia-phoenix-targets-5...
> Even when running long benchmarks, like this 30-minute Cinebench test, the MacBook Pro’s fans rarely spun up.
But left out how noisy the fans were when they did spin up. Can anyone here that has one of these models comment on the fan noise compared with any of the latest intel models?
It sounded promising, like it rarely turns on.
I think you meant never turned on
I think it's likely that Apple will offer Boot Camp for M1 Macs whenever Microsoft gets around to shipping Windows 11 for ARM in a non-beta form.
That was before Windows 11 was released. According to Parallels*, now you can just install normal Windows 11 releases.
For the TINY amount of PC gaming that I do, I have an older Lenovo laptop laying around that suits my occasional needs. If I didn't have that, I'd probably buy an Intel NUC to use as my reverse-Mac-mini.
For business software, I assume that if I had a need for that, then my employer would issue me a PC instead of a MBP. I dunno, I've been at Mac shops for a decade now.
So, yeah. I'm sure that Boot Camp is still a thing for some people. But I think it's like an optical drive or a headphone jack. The vast majority today will never notice that it's gone.
It's not that your life stops if you're suddenly unable to run Windows on your daily driver, and "most" people only use the web browser anyway (why do they need a >$1000 machine for that, by the way?), but these things do cause problems or become a dealbreaker for some people.
Sure. Most won't notice or care. But that's sort of true by definition; Apple wouldn't make a product decision where most would notice and care. Most HN users are nerds, we see these things much more often than the majority.
Even if you just use Web browsers, cheap computers are cheap and not enjoyable to use. A $500 laptop is a steaming pile of crap. Imagine making your well-paid white collar employers use a computer with a bad screen, janky touchpad, and limited performance just because they don't do anything super intense with their machines?
Most work places keep their laptops for several years. Amortizing the cost of that over 2-4 years is not a big deal.
Also, a lot of Web apps and Electron app are ram and performance hogs, so getting a cheap computer doesn't make a ton of sense.
You don't need one of these MacBook Pros for everyone, but a MacBook Air is worth paying for every employee at my work (or an equivalent PC laptop).
But, yes, there is definitely software that is Windows only. Some places ran bootcamp for these needs, but others will get Windows machines just for these special use cases.
For some people who like to use Macs, but need one piece of Windows enterprise software every now and then, they may just go with a Windows machine without bootcamp.
In principle, we could have a $500 version of the 2009 polycarbonate Crackbook, which minus the short-lived screen hinges was the point where Apple hit a usability threshold that made it a dream to use a laptop. Since only Apple realizes this, they could demand a ridiculous UX premium even if they offered a laptop today with 2009-level hardware.
But hey, I suppose modern JavaScript on the web kills the hope of being able to use it with old hardware anyway, eh :P And those of us who care don't think twice about spending >$1000 on a laptop, it's the most important extension of your mind and that's worth the premium.
1: https://blogs.vmware.com/teamfusion/2021/09/fusion-for-m1-pu...
If you’re spending $400 for RAM, then $200 from Pro to Max becomes obvious, but the battery life puts a damper on that since I’m not a creative.
Now, if Tensorflow could be properly supported then it might be definitely worth it. Though I’m surprised in their CPU benchmark the increase in memory bandwidth didn’t yield much better results.
Battery life, fan noise, and performance are my main reasons for upgrading.
I'm on the fence between buying a 32GB 24 core GPU Max version, or going for a cheaper 16GB 14 core GPU and using the leftover money to buy a home server with good GPU and RAM.
I'd like the Mac to sustain 95% of my programming, design, and data science needs, but I'd be willing to run the last 5% (gaming, very deep neural networks, or big datasets) remotely. Especially if I could get more bang for my buck buying a home station instead of maxing out the MacBook.
Many new external devices people use with MBPs still tend to come with USB-A. E.g. quality keyboards and mice. And then there's all of those very random and sometimes very useful devices from the past ~20 years that we all have around.
[0 EDIT] Actually I just remembered the pro controllers for my Switch are USB-C, and of course the unit itself charges over USB-C—but what's the data port on the Switch? USB-A! Because of course it is, because that's what's actually useful.
i wouldn't want to trade usb-a ports for usb-c ports any day on a laptop since the usb-c tb4 ports are so powerful.
Monitors have been shipping with usb-c in for years, and passive usb-c to displayport/hdmi cables exist.
The power brick that ships with the Macbook Pro is a USB-C wall adapter. The magsafe cable is USB-C on one end and magsafe on the other. The extended power range spec allows going up to 240W instead of only 100W per the original USB power delivery spec. The 16" Macbook Pro can draw up to 140W.
There aren't currently available any standard USB-C cables that support extended power delivery. The higher current requirements entail more robust cables that will have a compatibility marking on them.
So they are clearly not "ditching USB-C." They are just re-adding convenient ports; plug in HDMI in a pinch, throw your DSLR memory card right in the card reader.
Those things work very well without needing to remember an extra cable or adapter.
Basically, Apple's been pushing USBC-only for over half a decade now, and you can still plug directly in to 99% of video devices in the wild with HDMI, and only 10% (even that much? Probably more like 1-2%, because so many video displays are TVs) with USB-C, without resorting to an adapter or special cable that (again) users of most other brands don't need. It's not "legacy", it's "not that weird crap that only Mac users need".
I’m clumsy, I tend to accidentally stumble of cables that are plugged in. (It doesn’t happen that often, of course, but once every few months is sufficient to kill my laptop.)
MagSafe is wonderful. The fact that it’s a proprietary connector is far less important than the fact that it protects me from my clumsiness.
I blame the designers of USB-C for not coming up with something that can offer the benefits of MagSafe.
(Of course, the battery life of the new laptops makes it less necessary to plug in while you’re sitting on the sofa…)
Now I wonder, can the NITS unlocked by HDR content be controlled in some way? Can I write an HDR terminal emulator? I want to be able to use neovim outside without issues.
I think it would be a fun hack to try.
otoh maybe it's something like only a small portion at once of the screen can be mega-bright, in which case the terminal emulator idea isn't going to work out. (though maybe you could super-bright an area around the current line, which would be better than nothing...)
(all of the above pure speculation, obviously)
Yep, that's going to be a big question mark for the next several months.
> This was underlined sharply by an After Effects update that arrived in the middle of our testing, which enabled better multi-core acceleration. After Effects is still an Intel app, which runs in the Rosetta emulation mode on these machines, but the update allowed the app to light up all 10 CPU cores and all 32 GPU cores, cutting export times on our test from 4:29 to… 41 seconds. (It also substantially sped up performance on our Intel-based Mac Pro, which is nice.)
From this point forward, all consumer processors have a GPGPU - either discrete or embedded on CPU package (AMD Zen has been this way since the beginning). Adapt or get supplanted by a new piece of software that takes advantage.
You specify "either discrete or embedded" but how is that different from before, i.e. "from this point forward"?
Alder Lake CPUs also offer chips with our without integrated graphics, just as past Intel chips did. And, just as before, they are either paired with discrete graphics, or they have integrated graphics.
None of that seems to answer the OP question, though.
> Will Apple’s GPU ideas get fully supported by app developers?
Since you're talking about Alder Lake, I'm guessing the change you're really trying to speak on is the pairing of performance and efficiency cores, which has been common in mobile devices for quite a long time, but was brought to laptops with Apple Silicon, and finally reaches desktops with Alder Lake.
If you want your software to stay competitive, you'd better start using it. Notice the quote - a software update to After Effects caused the execution time of some job they were doing to go from 4.5 minutes to 0.75 minutes. Same hardware. Same workload. The only difference was software that took advantage of this new hardware paradigm that is going to start showing up everywhere...
> I'm guessing the change you're really trying to speak on is the pairing of performance and efficiency cores, which has been common in mobile devices for quite a long time, but was brought to laptops with Apple Silicon, and finally reaches desktops with Alder Lake.
It is that, AND guaranteed access to a programmable GPU
I watched a youtube video before, their M1 Max is twice faster than RTX 3090 machine on their Premiere test. What makes the diff :s
The Keyboard reliability problem has nothing to do with that.
Magsafe has always been great IMO, but I believe you can also charge the laptop via USB C. May just need the right specs on whatever brick you're using.
The slimmer design of the 2016 generation was really nice and had just as much battery life. A shame they threw out the touch bar entirely, with the hardware ESC it was absolutely fine. And they certainly didn’t take back the price increase it came with.
Performance of what metric, or for what software, is lacking with the hardware you run Windows on?
Or are you talking more about efficiency and portability/battery life?
I would expect these to be generally similar as they seem to use the same switch style. The important thing is that they don’t have the butterfly garbage.
https://macsx.com/which-macbooks-have-the-butterfly-keyboard...
For backend developers it's a little different - it is handy to have an x86, considering a lot of your workloads will run on x86 servers (at least for the foreseeable future), and a Linux machine will match the server runtime more closely. You can emulate one, and the M1 is so powerful it can emulate an x86 faster than most x86's can be themselves, but it's more expensive. A better approach I've been using (works with Python) is to treat macOS as a Unix and don't look too closely to the OS running the program. There aren't many surprised down this road.
I've been working as a back-end developer primarily from Macs since 2016 and they are sweet machines - even the Intel ones. I also use a Linux laptop a lot and, quite frankly, if you match specs, your Dell will cost more than your Mac and be a lot uglier.