Apple's Silicon Magic Is Over!
Summary
TLDR视频脚本回顾了苹果M1芯片的革命性影响,它不仅提升了性能,还改善了Mac的散热问题,延长了电池寿命,且无需风扇。随后,M2芯片在性能和能效上进行了迭代提升,尽管面临散热挑战。而M3芯片在3nm工艺上的性能提升并未达到预期,表明工艺缩减带来的性能增益正在减少。视频提出,尽管苹果的硅片性能不再是问题,但面对高通等竞争对手的追赶,苹果需要在硬件设计上进行创新,如开发更轻薄的笔记本电脑或针对游戏和创作者市场的高性能笔记本电脑,以维持其市场领先地位。
Takeaways
- 📈 苹果的M1芯片因其出色的性能和能效比,改变了市场对Mac电脑的看法,提供了前所未有的性能提升。
- 🔥 M1的成功归功于Arm64架构、专用硬件模块和苹果从硬件到应用的深度垂直控制。
- 🌡️ M2芯片通过采用更精细的制程和增加晶体管数量提升了性能,但也带来了更高的能耗和散热挑战。
- 💻 M3芯片在3nm工艺上的性能提升没有达到预期,表明制程缩小带来的性能增益正在减少。
- 💰 随着制程技术的进步,新节点的成本越来越高,这可能影响苹果未来芯片的性价比。
- 🚀 高通正在开发新的Snapdragon X Elite SoC,旨在与苹果的芯片竞争,并已获得微软的支持。
- 🔋 Snapdragon X Elite SoC的设计注重能效,提供了从轻薄笔记本到高性能设备的广泛适用性。
- 📱 苹果的硅片优势不再是其产品的唯一卖点,随着技术的发展,其他厂商正在迎头赶上。
- 🎮 苹果可以考虑开发新的MacBook设计,利用其硅片的性能和效率,推出更多样化的产品线。
- 🌟 M1系列将作为苹果历史上最伟大的计算机之一被铭记,但苹果需要继续创新以保持领先。
- ⏰ 苹果需要在其他厂商赶上之前,重新评估其产品策略,可能包括风险更大的创新设计。
- 📘 苹果的硬件设计应该利用其硅片的性能,探索新的产品形态,以满足不同用户群体的需求。
Q & A
为什么说五年前Mac电脑存在过热问题?
-五年前,Mac电脑使用的是高热设计功率(TDP)的Intel处理器,配合的冷却系统却不足以应对,导致电脑容易过热,进而影响性能基准测试的得分。
苹果硅片M1的推出对Mac电脑有哪些重大改变?
-M1硅片的推出使得Mac电脑在保持轻薄设计的同时,性能提升了3.5倍,且不需要风扇散热。此外,电池续航时间比之前延长了50%,而价格并未上涨。
M1硅片的三个主要性能优势是什么?
-M1硅片的性能优势包括:(1) 使用更现代的Arm64指令集架构,没有x86的遗留问题;(2) 拥有专门的芯片硬件模块,如视频引擎、神经网络引擎等,能更高效处理特定任务;(3) 从硬件到内核再到操作系统和应用程序的深度垂直控制,提高了整体的运行效率。
M2硅片与M1相比,有哪些改进?
-M2硅片采用了TSMC的N5P工艺,性能提升了7%,同时功耗降低了约15%。此外,M2增加了晶体管的数量,从160亿增加到200亿,但这也使得M2在峰值状态下可能比M1更加耗电和发热。
M3硅片在性能上的提升为何没有达到预期?
-M3硅片采用了TSMC的3nm工艺,但实际的晶体管密度提升并不如预期的2.8倍那么多。逻辑密度的提升约为1.7倍,而SRAM和IO密度几乎没有显著提升。因此,性能提升并没有达到预期的大规模增长。
高通的Snapdragon X Elite SoC有哪些特点?
-Snapdragon X Elite SoC是一款专为笔记本电脑设计的芯片,基于TSMC的4nm工艺制造,拥有12个高性能核心CPU、Adreno GPU和内置的Hexagon NPU。此外,还支持高达64GB的LPDDR5内存、Snapdragon X65 5G调制解调器和PCIe NVMe存储。
为什么说苹果的硅片设计不再只是关于性能?
-随着技术的发展,苹果的硅片设计已经达到了一定的物理极限,而且其他竞争对手正在迎头赶上。因此,苹果需要重新关注其硬件设计的其他方面,如创新的外形设计和功能的多样化,以保持其产品的竞争力。
苹果的MacBook产品线目前存在哪些设计上的相似性?
-苹果的MacBook产品线在设计、尺寸、体积、重量和功能集方面比以往任何时候都更为接近。尽管MacBook Air的显示屏和扬声器质量略低于14英寸的MacBook Pro,但整体上,这些机器的设计同质化程度很高。
苹果硅片的性能每瓦特优势为何没有得到充分利用?
-苹果硅片的性能每瓦特优势没有得到充分利用,因为没有一款笔记本电脑真正利用了这种性能优势所提供的外形设计潜力。例如,可以设计出比2015年12英寸MacBook更薄、更小、更轻的笔记本电脑,而不必受到低热设计功率(TDP)的Intel芯片和不良I/O的限制。
对于苹果来说,为什么需要继续冒险和创新?
-苹果在推出自己的硅片时采取了大胆而冒险的举措,这在M1系列上取得了巨大成功。然而,为了保持领先并避免其他竞争对手迎头赶上,苹果需要继续冒险和创新,开发出新的产品和功能,以维持其在市场中的领导地位。
苹果未来的硅片设计应该考虑哪些因素?
-苹果未来的硅片设计应该考虑包括性能、效率、成本效益、以及如何更好地利用其硅片的性能优势来创造独特的硬件设计。同时,还需要关注竞争对手的发展,以及市场和消费者的需求变化。
Outlines
😀 M1芯片的革命性影响及M2和M3的迭代发展
该段落回顾了五年前Mac电脑因高功耗Intel芯片和不足的散热导致的性能问题。随后,M1芯片的推出彻底改变了Mac的设计,不仅保持了轻薄,还实现了无风扇设计和更长的电池寿命。此外,M1 iMac和iPad Pro展示了苹果芯片在不同设备形态上的应用。尽管M2芯片在性能上有所提升,但也带来了更高的能耗和散热挑战。M3芯片在3nm工艺上的性能提升未达预期,表明工艺缩减带来的性能提升正在减少,且成本上升。
🤔 M3芯片的性能预期与现实差距
尽管M3芯片采用了TSMC的3nm工艺,但其性能提升并不如预期,原因在于新工艺的逻辑密度提升有限,且成本较高。此外,M2芯片虽然在性能上有所提升,但同时也面临更高的能耗和散热问题。这表明,尽管M2在性能每瓦上有所提升,但总体能耗在峰值时更高。
📈 高通骁龙X Elite与苹果芯片的竞争
高通骁龙X Elite是基于4nm工艺的定制笔记本芯片,拥有高性能CPU、GPU和NPU,支持高达64GB的LPDDR5内存和5G调制解调器。虽然其性能可能不及苹果的Pro、Max和Ultra芯片,但其低功耗设计和接近M3的性能表现显示出高通在芯片领域的野心和潜力。同时,微软和高通在软件优化方面的努力,以及未来可能集成高性能GPU的可能性,预示着苹果在芯片领域可能面临更激烈的竞争。
🚀 苹果硬件设计的创新与风险
该段落讨论了苹果MacBook产品线的设计同质化问题,并提出了创新的硬件设计思路,如开发更轻薄的笔记本电脑或针对游戏和创作者市场的强大笔记本。作者认为,尽管M1系列取得了巨大成功,但苹果需要继续创新和冒险,以保持在行业中的领先地位。
Mindmap
Keywords
💡Apple 硅片
💡M1 芯片
💡热设计功率(TDP)
💡指令集架构(ISA)
💡统一内存池
💡性能每瓦特
💡Snapdragon X Elite
💡N3E 3nm 工艺
💡性能节流
💡MacBook Air 与 MacBook Pro
💡技术极限
Highlights
5年前,Mac电脑因高热和性能下降而声名狼藉,但M1芯片的推出彻底改变了这一局面。
M1芯片不仅保持了Mac的轻薄设计,还实现了3.5倍的性能提升,并且不需要风扇。
M1芯片的电池续航比前代提升了50%,且价格没有增加。
MacBook Pro和M2 MacBook Air的重新设计修正了蝴蝶键盘的问题,恢复了MagSafe和其他I/O接口。
M1 iMac和iPad Pro证明了没有哪种形态对于Apple Silicon的高效率来说是太小的。
Apple当前的计算机产品线可能是有史以来最好的产品线之一。
M1芯片的三个主要性能驱动因素:Arm64架构、专用硬件模块和深度垂直控制。
M2芯片采用了更精细的N5P工艺,性能提升7%,功耗降低15%。
M2芯片的晶体管数量从160亿增加到200亿,但这也导致了更多的热量和功耗。
尽管M2在峰值时消耗更多能量,但其每瓦性能的提升意味着完成任务的速度更快,总体能耗更低。
M3芯片在TSMC的3nm工艺上的性能提升没有达到预期,晶体管密度的提升并不如预期那么大。
随着工艺节点的缩小,每次新的工艺节点的成本都在大幅增加。
Apple Silicon的优势不再仅仅在于硅片,而是需要关注于Apple的硬件设计。
高通展示了他们的Snapdragon X Elite SoC,这是一款为笔记本电脑设计的定制芯片,性能介于M2和M3之间。
Apple Silicon需要在保持足够性能的同时,重新关注极致的能效比。
Apple应该利用Apple Silicon的高能效比,设计出更轻薄、更小巧的笔记本电脑。
对于需要更强大性能的用户,Apple可以考虑在更大、更重的笔记本电脑中使用M3 ULTRA芯片。
自从M1系列以来,Apple需要再次冒险,推动技术创新,以免被竞争对手迎头赶上。
Transcripts
It's hard to believe that just over 5-years-ago, I was ripping into the internals of brand-new Macs
looking to hodge-podge a fix together, such that they wouldn’t overheat and cause benchmark scores
to plummet. High TDP Intel chips paired with inadequate cooling made for a deadly combo—one
that created the worst Macs in decades—but then this happened: [insert M1 unveiling]
Apple silicon was a revelation to what had been deemed an ill-suited
form factor. Not only did they *keep* things thin and light, but they launched with the
literal identical MacBook Air chassis just to prove—HEY—not only is this 3.5x faster,
but we don’t even need a fan. Oh, and by the way, we didn't touch the battery size
at all but it lasts 50% longer than before. Price increase? Nah. Send me a G and we coo.
The redesign of the MacBook Pro and M2 MacBook Air righted the wrongs caused by the butterfly
keyboard-sporting laptops of yore bringing back MagSafe and other I/O, improving displays,
speakers, keyboards, and more. Meanwhile, the M1 iMac and iPad Pro proved that no form factor was
too small for this master-class in efficiency. It’s not a stretch to say that Apple’s current
computer lineup is not just Apple’s best ever, but perhaps one of the best lineups ever from
*any* company. But be warned, because changes need to happen—and fast—if Apple wants that to
continue being true into the future. Let's tackle why the M3 series doesn't measure up
to the almost physics-defying standards the M1 set at launch; and, how for the first time ever,
real competition is just months away from almost every PC maker imaginable.
I’ve talked a lot about why Apple silicon has absolutely dominated since launch,
but there were really three main drivers behind M1’s incredible performance: (1) Arm64 itself uses
significantly more modern (ISAs) instruction set architectures that don’t carry the legacy baggage
of x86—nerdy crap like weakly-ordered memory models, a larger number of general-purpose
registers for parallelizable code, etc., (2) Apple’s dedicated on-chip hardware blocks like the
video engines, Neural Engine, matrix coprocessor, and unified memory pool, handle specific tasks
vastly more capably than a general-compute CPU or GPU, and, (3) deep vertical control
from hardware to kernel to OS to application helped eliminate cruft and streamline efficiency.
All of the work had to be paved for M1. Sure, M2 and M3 benefit from that work,
but they’ve been iterative—and Apple is now somewhat limited in the same way everybody
else is: transistor density. M1 launched on TSMC’s 5nm process and unlike the 90s
and 2000s when transistor density and node naming actually correlated with one another,
process names today like “5nm” don’t really mean anything. There’s more to chips than logic gates
and hardly any of those features precisely measure at the marketed process size anyways. Regardless,
the M1—not to even mention the M1 Max—was a an enormous die that was not just the biggest TSMC
5nm chip to date, but one of the largest Arm chips ever produced period. So what do you do
to get a faster chip like M2? There’s really three options: (1) shrink the size and power
consumption of the transistors so you can add more of them in the same envelope, (2) keep
the transistor size the same but increase the number of them which makes for a larger die with
greater heat and power drain, or, (3) keep the transistor size and count the same but increase
the voltage to push up the chip’s clock-speed which creates even more heat and power drain.
M2 did a combination of options 1 and 2. They were able to move to TSMC’s refined N5P process
which netted both a 7% performance improvement over N5 while drawing about 15% lower power;
and then to speed things up even more, they increased the total number of transistors
from 16 billion to 20 billion. But this increase didn’t come free. Do some shotty,
“not really the full story” napkin math, and the data would suggest the M2 is more power-hungry
than M1—running hotter and drawing more energy as a total package over its predecessor. And
that’s not theoretical: we proved back when the M2 MacBook Air launched that the chip was more
difficult to keep cool and experienced more rapid and more severe performance throttling
due to those thermals. So why wasn’t this more widely reported? Well, because M2’s performance
per watt increased as well—not just total package consumption. Imagine a high-performance sports
car. The car runs hotter and consumes more fuel when it reaches its top speeds, much like the M2;
however, because it's so fast and efficient, it can complete a 'race' much quicker than a regular
car, reducing the total time it is running at its hottest most fuel-consumptive state. So,
while yes, the M2 consumed more total energy at its peak, that extra compute was able to
get tasks done more quickly—reducing time spent at peak and therefore maintaining lower energy
consumption per task relative to M1. Sounds like a win-win—so what’s the problem? Sand, man.
When M3 launched on TSMC’s N3E 3nm process, it was the first chip to do so—and performance
expectations from pundits were high. I mean, doing napkin math anew would suggest a 2.8x increase in
transistor density—HUGE performance gains! But then M3 came out and we got… a slightly better
jump than we did from M1 to M2—and those on the same process! Huh? Well, I guess we learned our
lesson – using napkins for calculations can be as messy as using a lipstick for math. First of all,
TSMC’s 3nm node uses transistors that are much physically larger than 3nm—they’re
closer to 3.5. Okayyy, but even that would suggest a 2x density increase. Ah, but only
the logic density comes close at 1.7x. SRAM and IO density barely increases at
all. And chips—even magical ones—contain all of these components. Realistically,
there’s only about a 1.3x shrink. The era of massive improvements from one process shrink
to the next are over. The shrinks themselves are now YEARS apart. And even worse, each new node
is orders of magnitude more expensive than the prior, per unit area. It’s estimated that Apple’s
cost on these N3E chips is greater—not lesser—than just using a bigger area on an older node. Alas,
that would not yield the same efficiency gains we’ve come to
expect from Apple. More transistors on a bigger chip means more heat. So what do?
We’ve spent several minutes getting really nerdy and into the weeds on a lot of stuff
that normal people don’t care about—and at the end of the day, normal people buy
the vast majority of Apple’s products. Sure, the jump from M2 to M3 wasn’t as massive as
expected and the jump from M3 to M4 will likely be even smaller, but might I suggest something
heretical for a minute? That’s OK! The silicon isn’t the problem in Apple’s lineup any longer.
Look, Qualcomm invited me out to San Diego a few weeks ago and I got to check out their
reference design laptops (which basically means they’re not real—they’re prototypes)
for the Snapdragon X Elite SoC. You may recall, a year-and-a-half-ago,
we bought Microsoft’s Arm-based Windows Dev Kit. It utilized the same Microsoft SQ3
chip (which was just a rebranded Snapdragon 8cx gen3) found in a few quirky low-power,
low-performance Windows laptops. No offense to the folks at Qualcomm or Microsoft, but this thing
sucked. Its performance under ideal conditions was mediocre and ideal conditions were hard to
come by because so much of the Windows experience was wildly unoptimized for Arm—even after a decade
following the original release of Windows RT. But that was then—we live in the now. Not only
has every single native app for Windows made the transition to Arm, but massive quantities
of 3rd party apps have too—including big ones—like Google Chrome. Graphics APIs like DirectX, Vulkan,
and OpenGL are said to work through mapping layers and both Microsoft and Qualcomm have made
huge efforts to ensure a smooth transition—they were quick to volunteer that Apple is better at
this than anyone and they hope to be compared to them this summer when the X Elite laptops ship.
So, what is the Snapdragon X Elite? Well, they gave me one in a cute little acrylic trading
card. It is a bespoke laptop chip—not based on a mobile chip—built on TSMC’s 4nm process coming
with a 12 high-performance core CPU, Adreno GPU, and in-house Hexagon NPU. Additionally,
the on-board sensing hub houses an additional ISP, on-board WiFi 7 by default, and the capability to
be paired with up to 64GB of LPDDR5 memory, a Snapdragon X65 5G modem, and NVMe storage
over PCIe. The specs suggest Qualcomm is not messing around—and from the benchmarks I saw,
it consistently placed itself in between the M2 and M3. Not shabby at all. Now,
Apple is still certainly going to have the upper hand with their Pro, Max, and Ultra chips,
but this doesn’t aim to compete with those. While OEMs can push the X Elite to run up to 90W for an
extra performance boost, its reference-design consumes just 24W peak. Very, very close to M3.
Now do I think that Qualcomm’s going to come out blazing with the best laptop chips within the next
3-years? Not really, no. But they’re hungry, they’ve got Microsoft behind them, and they
alluded to the fact that using heavy-duty GPUs from NVIDIA or AMD wouldn’t be off the table in
the future—something Apple has zero aspirations for. And just like Apple, Intel,
and everybody else, they’re really leaning into their NPU for tasks that can use software-defined
hardware for maximum efficiency and speed. It’ll be exciting to see what form factors
the Snapdragon X Elite embodies given its massive power envelope available to OEMs—from netbooks to
power-hungry beasts. Layer in the fact that this is just their first foray into the X Elite line
and that higher-performance chips are on the roadmap, and well, we’ve got competition, baby.
So what’s Apple to do? Rush TSMC to the next process shrink? Pivot to developing hotter
more consumptive chips in the name of speed? No. And Apple knows that’s not their core
competency. Apple Silicon has always been about sufficient performance with extreme efficiency,
but physics are a cruel mistress and many watching this channel don’t realize they’re
already pushing up against boundaries that didn’t exist for the M1 series. I still see comments that
Apple silicon laptops are dead silent and that’s just… dead wrong. We edit videos for
this YouTube channel on a 14” M3 Max MacBook Pro and the fans run at full-tilt nearly all
the time—not just when exporting. And even with fans ablaze, our NLE struggles to get exports
out even close to the time the 16” MacBook Pro can. It’s throttling—hard. Now, does it
throttle to the point that it’s no faster than an M3 Pro 14” MacBook Pro? No, but its sometimes
slower than an M1 Max Mac Studio—something that benchmarks would very much suggest is impossible.
My point here is that for years, Apple had the exact same Intel chip SKUs as other
computer makers. The silicon was never their selling point. That is until M1,
when that formula got flipped on its head and Mac owners were—for the first time ever—able to
be braggadocios about their computer’s speed. But Apple is pushing against technological limits and
others are catching up—so lets sit the silicon aside for a minute and focus again on Apple’s
hardware design. As I see it, the entire MacBook lineup is basically the same. Sure,
the Air has a lower-quality display and worse speakers than the 14” MacBook Pro and the 15”
Air is cheaper than the 16” Pro, but I mean come on… these machines are closer in design,
size, footprint, weight, and feature-set than ever before. There’s no laptop that truly takes
advantage of the form-factor provided by Apple silicon’s insane performance per watt. Imagine
a laptop even thinner, smaller, and lighter than the 2015 12” MacBook—a computer that still feels
impossible today—nearly a decade after its release. Only this time, it doesn’t have to be
hamstrung by a crappy low-TDP Intel chip and lousy I/O. Would it be slower than an M3 Air? Sure. But
how many people own—heck—a base M1 MacBook Air and have never even approached the limits of that
chip? I’d venture to say MOST—and if your silicon can enable those impressive form factors… do it!
On the other end of the spectrum, why not put an M3 ULTRA in a 16” laptop that’s a
bit on the hefty-side—the style that gamers and creators buy all the time on team Windows/Linux?
A chip that just absolutely screams when needed with the thermal headroom to do it,
but while maintaining the excellent idle efficiency offered by Apple’s low-power
cores. It could be the first “gaming” laptop with a battery that doesn’t die in like 4 hours.
What I'm getting at is this - when Apple decided to make their own silicon,
it was a bold, risky move. It paid off massively. The M1 series will be remembered as some of the
greatest computers ever. But it also feels like that was the last time Apple really took a risk,
and I think its time they stop sitting on their laurels and get back to work before
the rest of the industry catches up. What do you think? Let me know in the comments below,
but most importantly—and as always—stay snazzy.
5.0 / 5 (0 votes)
If You’re Tired of MacBooks Winning…
Don’t buy a MacBook Pro - MacBook Air M3
Why I spent $3600 on the iPad Pro M4.
How to Optimize Performance in Unreal Engine 5
ROG 幻14 Air/ 幻16 Air评测:14英寸16mm+1.48kg的4060全能本ROG幻14 Air/幻16 Air評測:14英寸16mm+1.48kg的4060全能本 | 笔吧评测室
The MSI Claw is a Mess: Gaming Handheld Can't Compete | Review & Benchmarks