Apple's Silicon Magic Is Over!
Summary
TLDRThe video script discusses the evolution of Apple's Mac lineup, particularly focusing on the significant impact of the M1 chip on the industry. It highlights how Apple's shift to its own silicon led to a dramatic improvement in performance and efficiency, allowing for sleek designs without compromising on power. The script also addresses the incremental improvements from the M2 and the challenges faced by the M3 due to the limitations in transistor density and the escalating costs of newer chip processes. It emphasizes the need for Apple to innovate beyond silicon and explore unique form factors that leverage the performance per watt benefits of their chips. The video further acknowledges the growing competition from other chip manufacturers like Qualcomm, which is making strides with its Snapdragon X Elite SoC. The summary calls for Apple to continue taking bold risks and to innovate in design to stay ahead of the competition.
Takeaways
- đĽ The M1 chip was a game-changer for Apple, offering significant performance improvements and efficiency, without requiring a fan for cooling.
- đ Apple's M2 chip, while iterative, increased transistor count and moved to a more refined process, but also faced thermal challenges.
- đĄď¸ The M2 MacBook Air was found to be more difficult to cool and experienced performance throttling due to increased heat and power draw.
- đ Despite higher peak energy consumption, the M2's performance per watt improved, allowing for quicker task completion and lower energy usage per task.
- đ The M3 chip's launch on TSMC's 3nm process did not yield the expected massive performance gains, indicating a slowdown in process improvements.
- đ¸ The cost of manufacturing on newer nodes is increasing, which may affect the pricing and efficiency gains of future chips.
- đ¤ Qualcomm's Snapdragon X Elite SoC is a new competitor in the market, aiming to provide performance between Apple's M2 and M3 chips with a focus on efficiency.
- đ The era of significant performance improvements with each process shrink is over, with new nodes becoming more expensive and offering less density gains.
- đť Apple's MacBook lineup has become more uniform in design and features, but there's a call for more innovative form factors that take full advantage of Apple Silicon's capabilities.
- đŽ The suggestion that Apple could explore creating a gaming-oriented laptop with an M3 ULTRA chip, offering high performance with excellent idle efficiency.
- đ¨ A call for Apple to continue taking risks and innovate beyond their silicon, to maintain their competitive edge as the industry catches up.
Q & A
What was the main issue with the Macs from just over 5 years ago?
-The main issue was that they had high TDP Intel chips paired with inadequate cooling, which led to overheating and a significant drop in benchmark scores, making them some of the worst Macs in decades.
What was the significance of the M1 chip for Apple's Mac lineup?
-The M1 chip was a revelation, addressing the overheating issues and providing a significant performance boost. It allowed Apple to keep Macs thin and light without the need for a fan and offered a 3.5x speed increase with a 50% longer battery life, all without a price increase.
What are the three main drivers behind the M1's incredible performance?
-The three main drivers are: (1) The use of the modern Arm64 instruction set architectures, (2) Apple's dedicated on-chip hardware blocks that handle specific tasks more capably, and (3) deep vertical control from hardware to application level which streamlined efficiency.
How did the M2 chip improve upon the M1?
-The M2 chip improved by moving to TSMCâs refined N5P process, which provided a 7% performance improvement and about 15% lower power consumption. Additionally, the total number of transistors was increased from 16 billion to 20 billion.
What was the challenge with the M3 chip's launch on TSMCâs N3E 3nm process?
-The challenge was that while expectations were high for a significant performance increase due to the smaller process size, the actual improvement was only slightly better than the jump from M1 to M2, which was disappointing given the new process technology.
Why might the era of massive improvements from one process shrink to the next be over?
-The era might be over because the improvements in transistor density are becoming smaller with each new node, and the time between each shrink is now years apart. Additionally, each new node is significantly more expensive, which may not yield the same efficiency gains as before.
What is the Snapdragon X Elite SoC and how does it compare to Apple's chips?
-The Snapdragon X Elite is a custom laptop chip built on TSMCâs 4nm process with a high-performance CPU, GPU, and NPU. It is designed to compete with Apple's chips, with benchmarks placing it between the M2 and M3 in performance. However, it targets a different segment and is not aimed at competing with Apple's Pro, Max, and Ultra chips.
What is the current challenge for Apple's silicon lineup?
-The challenge is that the jumps in performance from M2 to M3 and potentially from M3 to M4 are not as massive as expected. Apple is pushing against technological limits, and competitors are catching up, which means Apple needs to innovate beyond just silicon performance.
Why is it suggested that Apple should focus on hardware design in addition to silicon performance?
-Focusing on hardware design can allow Apple to take better advantage of the performance per watt provided by their silicon. This could lead to innovative form factors, such as even thinner and lighter laptops or more powerful laptops with better thermal management and longer battery life.
What is the potential market for a laptop with an M3 ULTRA chip?
-The potential market includes gamers and creators who typically opt for heavier-style laptops on Windows/Linux platforms. An M3 ULTRA chip could provide the necessary performance with better thermal headroom and maintain excellent idle efficiency.
What is the author's final suggestion for Apple?
-The author suggests that Apple should stop resting on the success of the M1 series and take more risks to innovate and stay ahead of the competition. This includes focusing on both silicon performance and hardware design to create truly differentiated products.
Outlines
đ Apple Silicon's Impact on Mac Design
This paragraph discusses the significant transformation in Apple's Mac lineup following the introduction of the M1 chip. Previously, high TDP Intel chips coupled with insufficient cooling led to overheating issues and poor performance. The M1 chip, however, was a game-changer, offering faster performance without the need for a fan and with a 50% longer battery life, all at the same price point. The paragraph also touches on the improvements in the MacBook Pro and M2 MacBook Air, which corrected past design flaws and enhanced various features. It concludes by noting that while Apple's current lineup is one of the best ever, it must continue to innovate to maintain this status.
đ¤ The M2 and M3: Incremental Improvements and Challenges
The second paragraph delves into the technical aspects of Apple's M1, M2, and M3 chips. It explains the three main factors behind the M1's success: the modern Arm64 instruction set architecture, dedicated on-chip hardware blocks, and Apple's deep vertical control over the hardware and software stack. The paragraph then explores the challenges faced by the M2 and M3 chips, which include the limitations imposed by transistor density and the trade-offs between performance, power consumption, and heat generation. It also discusses the M2's increased transistor count and the thermal issues it faced, contrasting this with the M3's more modest performance improvements despite being built on a newer, more expensive process.
đ The Emergence of Snapdragon X Elite as Competition
In this paragraph, the focus shifts to the competition emerging in the market, particularly with Qualcomm's Snapdragon X Elite system-on-chip (SoC). The author shares their experience with Qualcomm's reference design laptops and notes the significant strides made in optimizing Windows for Arm architecture. The Snapdragon X Elite is described as a powerful laptop chip with a high-performance CPU, GPU, and NPU, capable of competing with Apple's M2 and M3 chips in benchmarks. The paragraph suggests that while Apple's Pro, Max, and Ultra chips will likely remain superior, the X Elite represents a serious challenge, especially with Microsoft's support and the potential for future integration of high-end GPUs from NVIDIA or AMD.
đĄ Pushing the Boundaries of Apple Silicon
The final paragraph emphasizes the need for Apple to continue innovating beyond its silicon capabilities. It suggests that while the M1 series was a groundbreaking success, Apple must not rest on its laurels as competitors are catching up. The author proposes exploring new form factors that take full advantage of Apple silicon's performance per watt, such as a thinner and lighter laptop or a more powerful gaming-oriented laptop. The paragraph also reflects on Apple's past reliance on Intel chips and how the M1 series allowed Mac users to boast about their computers' speed for the first time. It concludes by encouraging Apple to take more risks and innovate further to stay ahead of the competition.
Mindmap
Keywords
đĄApple Silicon
đĄThermal Throttling
đĄTransistor Density
đĄInstruction Set Architecture (ISA)
đĄUnified Memory
đĄM1 Max
đĄSnapdragon X Elite
đĄProcess Shrink
đĄNeural Engine
đĄEfficiency
đĄM3 Series
Highlights
Over 5 years ago, Apple's Macs faced overheating issues due to high TDP Intel chips and inadequate cooling.
Apple silicon's introduction with the M1 chip was a significant shift from the problematic form factor of previous Macs.
The M1 chip allowed for 3.5x faster performance without the need for a fan and with a 50% longer battery life.
Apple maintained the price while delivering significant performance improvements.
The redesign of the MacBook Pro and M2 MacBook Air corrected past design flaws and enhanced various features.
Apple's M1 iMac and iPad Pro demonstrated efficiency across various form factors.
Apple's current computer lineup is considered one of the best ever from any company.
The M1's performance was driven by three main factors: modern instruction set architectures, dedicated hardware blocks, and deep vertical control.
The M2 chip improved on the M1 by using a refined process and increasing the number of transistors.
The M2, despite being more power-hungry, showed increased performance per watt.
The M3 chip's launch on TSMC's 3nm process did not yield the expected massive performance gains.
The era of significant improvements from process shrinks is over, with new nodes being more expensive.
Qualcomm's Snapdragon X Elite SoC is a new competitive force in the market, with promising benchmarks.
The X Elite chip is a custom laptop chip with high-performance CPU, GPU, and NPU, manufactured on TSMC's 4nm process.
Apple's focus on efficiency and performance with Apple Silicon faces new competition from Qualcomm and others.
Apple's MacBook lineup has become more uniform in design and features, lacking a machine that truly leverages Apple silicon's performance per watt.
There is a suggestion for Apple to innovate further by creating a thinner, lighter laptop or a more powerful 'gaming' laptop.
The M1 series is seen as a game-changer, but Apple needs to continue taking risks to stay ahead of the competition.
Transcripts
It's hard to believe that just over 5-years-ago, I was ripping into the internals of brand-new Macs Â
looking to hodge-podge a fix together, such that they wouldnât overheat and cause benchmark scores Â
to plummet. High TDP Intel chips paired with inadequate cooling made for a deadly comboâone Â
that created the worst Macs in decadesâbut then this happened: [insert M1 unveiling]
Apple silicon was a revelation to what had been deemed an ill-suited Â
form factor. Not only did they *keep* things thin and light, but they launched with the Â
literal identical MacBook Air chassis just to proveâHEYânot only is this 3.5x faster, Â
but we donât even need a fan. Oh, and by the way, we didn't touch the battery size Â
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 Â
keyboard-sporting laptops of yore bringing back MagSafe and other I/O, improving displays, Â
speakers, keyboards, and more. Meanwhile, the M1 iMac and iPad Pro proved that no form factor was Â
too small for this master-class in efficiency. Itâs not a stretch to say that Appleâs current Â
computer lineup is not just Appleâs best ever, but perhaps one of the best lineups ever from Â
*any* company. But be warned, because changes need to happenâand fastâif Apple wants that to Â
continue being true into the future. Let's tackle why the M3 series doesn't measure up Â
to the almost physics-defying standards the M1 set at launch; and, how for the first time ever, Â
real competition is just months away from almost every PC maker imaginable.
Iâve talked a lot about why Apple silicon has absolutely dominated since launch, Â
but there were really three main drivers behind M1âs incredible performance: (1) Arm64 itself uses Â
significantly more modern (ISAs) instruction set architectures that donât carry the legacy baggage Â
of x86ânerdy crap like weakly-ordered memory models, a larger number of general-purpose Â
registers for parallelizable code, etc., (2) Appleâs dedicated on-chip hardware blocks like the Â
video engines, Neural Engine, matrix coprocessor, and unified memory pool, handle specific tasks Â
vastly more capably than a general-compute CPU or GPU, and, (3) deep vertical control Â
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, Â
but theyâve been iterativeâand Apple is now somewhat limited in the same way everybody Â
else is: transistor density. M1 launched on TSMCâs 5nm process and unlike the 90s Â
and 2000s when transistor density and node naming actually correlated with one another, Â
process names today like â5nmâ donât really mean anything. Thereâs more to chips than logic gates Â
and hardly any of those features precisely measure at the marketed process size anyways. Regardless, Â
the M1ânot to even mention the M1 Maxâwas a an enormous die that was not just the biggest TSMC Â
5nm chip to date, but one of the largest Arm chips ever produced period. So what do you do Â
to get a faster chip like M2? Thereâs really three options: (1) shrink the size and power Â
consumption of the transistors so you can add more of them in the same envelope, (2) keep Â
the transistor size the same but increase the number of them which makes for a larger die with Â
greater heat and power drain, or, (3) keep the transistor size and count the same but increase Â
the voltage to push up the chipâs 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 TSMCâs refined N5P process Â
which netted both a 7% performance improvement over N5 while drawing about 15% lower power; Â
and then to speed things up even more, they increased the total number of transistors Â
from 16 billion to 20 billion. But this increase didnât come free. Do some shotty, Â
ânot really the full storyâ napkin math, and the data would suggest the M2 is more power-hungry Â
than M1ârunning hotter and drawing more energy as a total package over its predecessor. And Â
thatâs not theoretical: we proved back when the M2 MacBook Air launched that the chip was more Â
difficult to keep cool and experienced more rapid and more severe performance throttling Â
due to those thermals. So why wasnât this more widely reported? Well, because M2âs performance Â
per watt increased as wellânot just total package consumption. Imagine a high-performance sports Â
car. The car runs hotter and consumes more fuel when it reaches its top speeds, much like the M2; Â
however, because it's so fast and efficient, it can complete a 'race' much quicker than a regular Â
car, reducing the total time it is running at its hottest most fuel-consumptive state. So, Â
while yes, the M2 consumed more total energy at its peak, that extra compute was able to Â
get tasks done more quicklyâreducing time spent at peak and therefore maintaining lower energy Â
consumption per task relative to M1. Sounds like a win-winâso whatâs the problem? Sand, man.
When M3 launched on TSMCâs N3E 3nm process, it was the first chip to do soâand performance Â
expectations from pundits were high. I mean, doing napkin math anew would suggest a 2.8x increase in Â
transistor densityâHUGE performance gains! But then M3 came out and we got⌠a slightly better Â
jump than we did from M1 to M2âand those on the same process! Huh? Well, I guess we learned our Â
lesson â using napkins for calculations can be as messy as using a lipstick for math. First of all, Â
TSMCâs 3nm node uses transistors that are much physically larger than 3nmâtheyâre Â
closer to 3.5. Okayyy, but even that would suggest a 2x density increase. Ah, but only Â
the logic density comes close at 1.7x. SRAM and IO density barely increases at Â
all. And chipsâeven magical onesâcontain all of these components. Realistically, Â
thereâs only about a 1.3x shrink. The era of massive improvements from one process shrink Â
to the next are over. The shrinks themselves are now YEARS apart. And even worse, each new node Â
is orders of magnitude more expensive than the prior, per unit area. Itâs estimated that Appleâs Â
cost on these N3E chips is greaterânot lesserâthan just using a bigger area on an older node. Alas, Â
that would not yield the same efficiency gains weâve come to Â
expect from Apple. More transistors on a bigger chip means more heat. So what do?
Weâve spent several minutes getting really nerdy and into the weeds on a lot of stuff Â
that normal people donât care aboutâand at the end of the day, normal people buy Â
the vast majority of Appleâs products. Sure, the jump from M2 to M3 wasnât as massive as Â
expected and the jump from M3 to M4 will likely be even smaller, but might I suggest something Â
heretical for a minute? Thatâs OK! The silicon isnât the problem in Appleâs lineup any longer.
Look, Qualcomm invited me out to San Diego a few weeks ago and I got to check out their Â
reference design laptops (which basically means theyâre not realâtheyâre prototypes) Â
for the Snapdragon X Elite SoC. You may recall, a year-and-a-half-ago, Â
we bought Microsoftâs Arm-based Windows Dev Kit. It utilized the same Microsoft SQ3 Â
chip (which was just a rebranded Snapdragon 8cx gen3) found in a few quirky low-power, Â
low-performance Windows laptops. No offense to the folks at Qualcomm or Microsoft, but this thing Â
sucked. Its performance under ideal conditions was mediocre and ideal conditions were hard to Â
come by because so much of the Windows experience was wildly unoptimized for Armâeven after a decade Â
following the original release of Windows RT. But that was thenâwe live in the now. Not only Â
has every single native app for Windows made the transition to Arm, but massive quantities Â
of 3rd party apps have tooâincluding big onesâlike Google Chrome. Graphics APIs like DirectX, Vulkan, Â
and OpenGL are said to work through mapping layers and both Microsoft and Qualcomm have made Â
huge efforts to ensure a smooth transitionâthey were quick to volunteer that Apple is better at Â
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 Â
card. It is a bespoke laptop chipânot based on a mobile chipâbuilt on TSMCâs 4nm process coming Â
with a 12 high-performance core CPU, Adreno GPU, and in-house Hexagon NPU. Additionally, Â
the on-board sensing hub houses an additional ISP, on-board WiFi 7 by default, and the capability to Â
be paired with up to 64GB of LPDDR5 memory, a Snapdragon X65 5G modem, and NVMe storage Â
over PCIe. The specs suggest Qualcomm is not messing aroundâand from the benchmarks I saw, Â
it consistently placed itself in between the M2 and M3. Not shabby at all. Now, Â
Apple is still certainly going to have the upper hand with their Pro, Max, and Ultra chips, Â
but this doesnât aim to compete with those. While OEMs can push the X Elite to run up to 90W for an Â
extra performance boost, its reference-design consumes just 24W peak. Very, very close to M3.
Now do I think that Qualcommâs going to come out blazing with the best laptop chips within the next Â
3-years? Not really, no. But theyâre hungry, theyâve got Microsoft behind them, and they Â
alluded to the fact that using heavy-duty GPUs from NVIDIA or AMD wouldnât be off the table in Â
the futureâsomething Apple has zero aspirations for. And just like Apple, Intel, Â
and everybody else, theyâre really leaning into their NPU for tasks that can use software-defined Â
hardware for maximum efficiency and speed. Itâll be exciting to see what form factors Â
the Snapdragon X Elite embodies given its massive power envelope available to OEMsâfrom netbooks to Â
power-hungry beasts. Layer in the fact that this is just their first foray into the X Elite line Â
and that higher-performance chips are on the roadmap, and well, weâve got competition, baby.
So whatâs Apple to do? Rush TSMC to the next process shrink? Pivot to developing hotter Â
more consumptive chips in the name of speed? No. And Apple knows thatâs not their core Â
competency. Apple Silicon has always been about sufficient performance with extreme efficiency, Â
but physics are a cruel mistress and many watching this channel donât realize theyâre Â
already pushing up against boundaries that didnât exist for the M1 series. I still see comments that Â
Apple silicon laptops are dead silent and thatâs just⌠dead wrong. We edit videos for Â
this YouTube channel on a 14â M3 Max MacBook Pro and the fans run at full-tilt nearly all Â
the timeânot just when exporting. And even with fans ablaze, our NLE struggles to get exports Â
out even close to the time the 16â MacBook Pro can. Itâs throttlingâhard. Now, does it Â
throttle to the point that itâs no faster than an M3 Pro 14â MacBook Pro? No, but its sometimes Â
slower than an M1 Max Mac Studioâsomething 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 Â
computer makers. The silicon was never their selling point. That is until M1, Â
when that formula got flipped on its head and Mac owners wereâfor the first time everâable to Â
be braggadocios about their computerâs speed. But Apple is pushing against technological limits and Â
others are catching upâso lets sit the silicon aside for a minute and focus again on Appleâs Â
hardware design. As I see it, the entire MacBook lineup is basically the same. Sure, Â
the Air has a lower-quality display and worse speakers than the 14â MacBook Pro and the 15â Â
Air is cheaper than the 16â Pro, but I mean come on⌠these machines are closer in design, Â
size, footprint, weight, and feature-set than ever before. Thereâs no laptop that truly takes Â
advantage of the form-factor provided by Apple siliconâs insane performance per watt. Imagine Â
a laptop even thinner, smaller, and lighter than the 2015 12â MacBookâa computer that still feels Â
impossible todayânearly a decade after its release. Only this time, it doesnât have to be Â
hamstrung by a crappy low-TDP Intel chip and lousy I/O. Would it be slower than an M3 Air? Sure. But Â
how many people ownâheckâa base M1 MacBook Air and have never even approached the limits of that Â
chip? Iâd venture to say MOSTâand 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 16â laptop thatâs a Â
bit on the hefty-sideâthe style that gamers and creators buy all the time on team Windows/Linux? Â
A chip that just absolutely screams when needed with the thermal headroom to do it, Â
but while maintaining the excellent idle efficiency offered by Appleâs low-power Â
cores. It could be the first âgamingâ laptop with a battery that doesnât die in like 4 hours.
What I'm getting at is this - when Apple decided to make their own silicon, Â
it was a bold, risky move. It paid off massively. The M1 series will be remembered as some of the Â
greatest computers ever. But it also feels like that was the last time Apple really took a risk, Â
and I think its time they stop sitting on their laurels and get back to work before Â
the rest of the industry catches up. What do you think? Let me know in the comments below, Â
but most importantlyâand as alwaysâstay snazzy.
5.0 / 5 (0 votes)