This was originally written by the authorDiscuss why Intel can’t make a better chip than Apple?I’m afraid it’s the wrong person.〉The original title, based on past experience, as expected to see a message of “good reviews” (?) (the author still occasionally peeks at the message, especially sensitive to being labeled “fruit fan”), very Thank you for your appreciation, so that the author, who is so busy with work on weekdays, has the motivation to continue writing.
Looking back at the works over the years, “Double A” (Apple, AMD) has eaten up a large “share” of the author’s work, but the author must say that the topic of writing Apple and AMD seems to be very trendy at present, but the reasons behind these things are all across the board. 20~30 years, or even longer historical timeline.Like the author, many RISC-V Supporters Lao Ai described the long-existing RISC instruction set as a great invention that turned out to be a great invention, and felt extremely disgusted. John Hennessy (MIPS) co-authored the computer structure “measurement method” and computer organization “white abacus”, they do not know IBM RISC Is it nearly half a century old? Is it to pretend to forget or, do not want to remember?
I don’t understand that 30 years ago, Apple once hoped to adopt Alpha processorbut hit a wallChoose PowerPCIntel in “Charizard Legend“Afterstrategic turning pointLet Apple abandon IBM to Intel, Apple in 2005 Acquired PA Semi in 2008 In the past, it is impossible to fully understand how Apple Silicon’s amazing power consumption ratio and lateral scalability came from.Regarding AMD’s Zen micro-architecture, there is a whole series behind the success of “Fen VI Yu Lie”The generally untold long story。
Without these historical awareness, the cognition of what you see in front of you is like a castle built on sand, which is in danger of overturning at any time. For example, the processor industry pursues “Performance Per Watt”, as early as 25 years ago DEC When developing a low-power product line, it was not feasible to evaluate the continuation of Alpha, and it began to cooperate with ARM to develop StrongARM (in fact, there should be earlier cases, but it was difficult to verify), and later became Intel’s XScalethe volume is more activated at Intelpendulum（Tick-Tock）When the giant ship reached its peak, it wasn’t that “Double A” seemed to be very popular in recent years, and this shocking “event” happened.
This is the Intel XScale report inherited from DEC StrongARM, 20 years ago, the data is still citing the industry’s well-known Microprocessor Report (Microprocessor Report). “Energy consumption ratio” will attract attention, not because of the rise of “Double A”, but because it has been valued a long time ago. Per Watt” on the lips. By the way, XScale is still the “strongest ARM processor of the generation” that Intel has ever owned, but it was sold to Marvell because of the stupid “x86 Boxer Rebellion”, which not only missed the great opportunity to build the first-generation iPhone processor for Apple, but also Ruined the future of Intel’s smartphone market.
The author summarizes the misconceptions about Apple Silicon related issues in recent years into the following five questions and answers. I hope that Apple will not be talked about for a long time. Call me after the rumored M3 processor with 3nm process is released. .
Question 1: In the foreseeable future, what will be the “current situation” of the x86 duo vs. Apple?
Whether Intel or AMD, based on “Process technology behindApple for at least one generation”, “lack of ultra-power-efficient micro-architecture based on mobile computing” (x86 duo’s product designServer-centric for 20 yearsIt is really difficult to change), “without the shade of its own operating system”, it is impossible to create a “mobile computing processor” (probably extended to desktop computers) that is comparable to Apple Silicon in “energy consumption ratio”. The above are extremely objective facts, and there is no room for doubt.
If you don’t agree, please convince the author (or the reader yourself) first that Intel’s 14th-generation Core processor “Meteor Lake” (Intel “class” 4nm, multi-chip package) and AMD Ryzen 7000 APU “Phoenix Point” are scheduled to debut in 2023 “(TSMC 4nm) There is a slight possibility that the energy consumption ratio will be “four times higher”, catching up withThe current Apple M2。
No one should think that Intel Meteor Lake can “four times the energy efficiency”, even Intel can’t believe it.
In short, after the presidential election of a certain year, “This is an unfair election. I want to propose an election—an election—invalid—an ineffective—an appeal—” (please match the background with the out-of-voice “” The election is invalid~the election is invalid~valid~”), which should be the most appropriate portrayal.
The second question: It is clear that the x86 processor duo and Apple “well water does not make river water”, why do they have to compare?
This made the author have to think of the cold joke of “Why do you want to go to the Internet wirelessly? Can’t you surf the Internet wirelessly?”
To put it bluntly, it’s not that Intel and AMD put their faces together to beat people, IntelCEO Pat Gelsinger deliberately chokes Apple as a ‘Lifestyle Company’ that will provide better products than Apple’s homemade chips, and delusional Apple’s change of heart to use Intel chips; David McAfee, vice president of product management and marketing at AMD, publicly praised: “All x86 processors cannot be compared to M1 in energy efficiency, and M1 is a good incentive for the processor industry.” This kind of behavior is called Self-inflicted sins cannot survive, and even gods cannot save them. Saying such words, what can you do without letting you face it?
But look at MacBook sales have grown significantly after the release of the M1, For the sake of gradually eroding the Wintel notebook market, even if the damage is limited from the perspective of the x86 duo, the notebook manufacturers whose market share is damaged will never remain indifferent. It can improve the product. On the bright side, consumers who pay for it will not suffer after all.
Question 3: If you want the x86 duo to “immediately” make a product that can catch up with Apple Silicon in terms of energy consumption, what should you do?
It’s very simple, see who can convince Apple to make a “server” processor, after all, a server that “has something to do” and a mobile computing platform that “often sleeps to death”, the logic of “power saving” is very different. The same, not to mention the introduction of simultaneous multi-thread architecture (SMT), more complex highly scalable system architecture and many RAS (Reliability, Availability, Serviceability) functions to maintain high reliability in order to improve the output rate. huge trouble.even thoughThe PA Semi that forms the backbone of the Apple Silicon R&D team It’s not that there is no technical foundation in this area, but the complexity of the server platform is far from what it was a decade ago. It is too difficult (but not impossible) to rebuild this aspect.
In the past 20 years, starting from Intel Pentium Pro (1995) and AMD K7 (1999), until the birth of Intel Xeon (1998) and AMD Opteron (2002) brands, the x86 dual-hero main processor core micro-architecture, They all prioritize high-profit server applications, which have not changed so far, and almost wiped out the “RISC gods” that dominated the server and workstation market in the 1990s. Only IBM Power barely survived and maintained stable development.
At the level of the processor microarchitecture, something like an x86 processor (with the addition of IBM’s Power and Z) dynamic branch prediction function (Branch Prediction) that maintains the “violent” level for a long time, unlike theUphold “minimalism”ofWhere Apple is willing to invest hard, because Apple Silicon does not need products like x86 duo, facing all kinds of strange application behaviors and the heavy burden of compatibility with old software, and to do everything possible to run faster, this is There are as many subtle differences as you want to enumerate. This is where the technological development direction of x86 processors has remained unmoved since the 1990s – no way, due to commercial factors, the most difficult part of developing x86 processors is “compatible with all software”.
also,Innate complexity of the x86 instruction setAnd the lack of industry standards, it also greatly extends the product development time, and needs to concentrate resources on the development of a single general-purpose core, forcing the x86 duo to make a lot of compromises between servers, desktops and laptops. Intel Strategic choices for the Nehalem generationA prime example of this, and then sacrificed is the laptop dragged to the 32nm Westmere.
Since the early 2000s, whether Intel or AMD, the main x86 core microarchitecture has to take into account servers, desktops and laptops (due to profit considerations, the priority is often “server > laptop > desktop”, but Intel’s Nehalem In order to defeat AMD as soon as possible, the laptop was sacrificed first), so it is bound to make a choice, which is a burden that Apple does not have.
But for the same reason, Apple, which has briefly launched Xserve, is even less likely to return to the server market, and there is no need to build server processors for its own data centers.Enlarged version available for laptops and desktops”, so the power consumption ratio of the x86 dual-core laptop processors will continue to be rubbed by Apple on the ground for a long time. The author predicts that there will be at least 3 years, just when Intel expects that “the process technology will overtake TSMC”. Absolutely, should, maybe, hopefully. Even if it doesn’t, Pat Gelsinger, 65, will probably be ready to retire by then.
Question 4: Does ARM have a way to shake up the x86 PC processor world?
There is only one sentence: it all depends on whether Microsoft is willing to seriously manage the ARM version of the Windows operating system, common applications and “re-establish a new ecosystem”, whether the word processor consumes power or not, it is just a blind man, and it does not make any sense. . Apps are king, and Windows users who are not compatible with macOS and MacBook’s rotten keyboards will not jump to a closed Apple Park because Apple Silicon saves power. To put it nasty, if the mere “low power consumption” and “high energy efficiency” are enough to “shake” the x86 personal computer processor, it should have been shaken for 20 years, and we will not see the x86 two heroes in the personal The computer market is still alive and well, and there is really no need to make a fuss about something rare. Ugh.
The author speaks more plainly, unless there are obvious commercial interests, such as Intel pushes WiMAX The main reason is “to allow more users in backward areas to go online to stimulate the sales of personal computers”, whether the establishment of the Windows On ARM ecosystem will allow Microsoft’s cloud business to have more room for development, and so on. The incentive for Microsoft to go all-in on Windows On ARM is not limited to desktops, but they also have to consider the server market possibilities.
Moreover, this also involves the reality of business competition. When a processor manufacturer is desperately trying to make an ARM desktop processor with performance that can compete with the x86 duo and put it on the market, it will immediately find that it is full of low prices everywhere. The stock of low-priced x86 processors leads to only price-cutting competition, which is not profitable at all. It’s the same sentence: In business, Microsoft doesn’t seriously “wash its head down”, and no one in this processor manufacturer, which mainly focuses on the mobile phone market, dares to take huge risks rashly.
Having said that, how much resources and time should be invested behind the ARM processor to allow users to install various operating systems freely like the current x86 processor, instead of preloading the BSP (Board Support Package) in the past? Won’t Microsoft turn the abacus to assess whether it’s cost-effective? Stepping back 10,000 steps, most of the existing ARM processors mainly based on SoC type “install Windows and use it normally”. Do you think this is an easy job?
The author would like to remind you that no one in the world stipulates that ARM processors can only be made into “System On Chip” (SoC) that are difficult to expand peripheral devices, like the only independent independent device currently surviving. ARM server processorManufacturer Ampereno matter how their products look, they are genuine “highly scalable server platforms”, and the x86 processor world has been around since then. Cyrix MediaGX Later, there are also many SoC-type products, like the ancientNational Semiconductor NS486 andTaiwanese Rise mP6-derived SoC with strong colors，The x86 duo also persistently develops SoCs for different applicationsfrom game consoles (Sony PlayStation, Microsoft Xbox) to set-top boxes (Chunghwa Telecom MOD), the latest Intel Atom x6000 processor “Elkhart Lake” is a fake “system single chip”, and the family is not as complicated Backup.
Although it only becomes one through multi-chip packaging, at the application level, Intel’s new-generation Atom processor is also called System-On-Chip (ST Micro’s fourth-generation set-top box flagship SoC STiH416 is also similar. ), not to mention the many application-specific products that have appeared in the past, such as Chunghwa Telecom MOD used Atom CE4100 “Sodaville” and numerous DOCSIS 3.0 cable modems used Atom CE4200 “Groveland”.
“Application” is the key point. Saying that ARM processors are difficult to replace x86 desktops is “the former is SoC”, but it’s just a matter of consequence. If Windows On ARM really has a head start, who would dare to say that there will be no compatible pins for existing x86 processors. The product? say nothing else,AMD’s K12and SkyBridge It’s really close to achieving this seemingly impossible “feat”.
If AMD hadn’t canceled the K12 processor and SkyBridge chipset, maybe now we would see a system platform architecture compatible with both x86 and ARM, and we wouldn’t see Jim Keller scolding “AMD is stupid to cut off K12”. Who knows!
Last question: How do I view Apple as a company? Is the seemingly ignorant author really a “fruit fan”?
There should be no normal people in the world who spend more than 95% of their “IT life time” facing Windows and Android who are “fruit fans”.
As a long-time user of the “King of Commercial Laptops” ThinkPad “Xiao Hei” (several expeditions to Japan, just to clamor for a traditional seven-column keyboard from the R&D team of Yamato Lab), I recently chatted with a friend about Apple After ten years of notebook computer design, the common feeling is that “this company is always doing its own thing behind closed doors, and it doesn’t care about what other notebook computer manufacturers do. They only need to take care of their own market, and they will Will deliberately force unlucky customers.”
Therefore, when the ThinkPad flagship product line X1 Carbon is getting lighter and lighter (I have the seventh generation, the latest one is the tenth generation), the MacBook (currently M1 MacBook Air) still brings a thick and heavy “aluminum block”, and It is also easy to deform the case due to strong collisions, and the keyboard is still hard enough to hit. The only invincible touchpad is the touchpad. Most of the time when I go out to work, I still have to carry the X1 Carbon, and even the larger and heavier ThinkPad T14s AMD Gen1. Just for the keyboard feel, because it directly affects work efficiency, it is more important than whether the processor can save power for a long time.
By the way, in order to maintain the “consistency of home and work operation”, the author deployed Filco Majestouch Convertible 2 NINJA keyboard (Cherry brown switch) and Logitech MX Master 3 mouse on both sides, and also configured the original Filco keyboard palm rest. It should be enough to explain the author’s “picky degree” of human-machine interface.
The author is definitely not a fruit fan. Really, please believe me, the rotten keyboard of the MacBook alone is enough to eliminate all the possibilities for the author to become a fruit fan. I heard that Cook was born in IBM? Should Apple launch a “commercial MacBook” to compete with the ThinkPad also from IBM in the field of laptop keyboards?
In the end, Apple’s so-called “impeccable and good user experience” also relies on the inherent “asymmetric advantage” of the closed ecosystem, and each link is perfectly interlocked. Starting with the iPhone 4, the author who has been running dual systems for a long time has both an Android phone and an iPhone (now Samsung S20 and iPhone SE2). Under the premise of installing similar software, the speed of idle power down every night is enough to highlight the “software and hardware” on both sides. Integration”, not to mention a bunch of “inconspicuous little places”.
But does this make me think that Apple is a special company? Of course not, because this is the result of “time” combined with “place” plus “people”, or a little bit of “luck”, and the most important “mission” – Apple’s most admirable part, At the very least, not too many people will question Apple’s paranoid determination to make its products perfect, even if the keyboard is so bad that it looks like it’s never going to be saved.
At this point, the author is still in a coffee shop near the Nangang Exhibition Hall on Saturday, struggling to finish this article with the bad keyboard of the M1 MacBook Air, even if the author does not particularly like Apple products, but maybe outsiders see, let’s not talk about it. Like middle-aged Wenqing, the author really looks like an out-and-out fruit powder. This is quite a big misunderstanding.
(Source of the first image: Screenshot of the video)