On Tuesday, Apple updated the iMac with new SKUs, display panels, and various color options. The new systems are powered by Apple’s M1 CPU, which uses the ARMv8 ISA rather than an Intel or AMD x86 CPU.
That sentence is a bit more revolutionary than it might seem. The three new iMacs, which are priced at $1,299, $1,499, and $1,699, respectively, do not differ by memory (8GB) or by CPU core count. Apple calls the M1 an eight-core, and it nominally is. But it’s better understood as a 4+4 CPU architecture, with four high-performance CPU cores and four high-efficiency CPU cores. The only difference between the M1 you get at $1,099 and the one you get at $1,699 is a single GPU core.
If you want to buy a MacBook Air or MacBook Pro, Apple will sell you an M1. Want a Mac Mini? You get an M1. Interested in the iMac or the new iPad Pro? You get an M1. It’s possible that the M1 CPUs inside the iMac will have different thermal or clock behavior than those inside the systems Apple has already launched, but the company’s decision to eschew clock speed disclosures suggests that these CPUs differ only modestly. The iMac might have the same 3.2GHz base clock but hold its frequency better under load, for example.
But outside of that, Apple is selling a single CPU across a wider range of products than any competing Intel or AMD CPU is ever sold. This speaks volumes as to what Apple believes it has its hands on, namely: a CPU fast enough at the quad-core level — because, scaling-wise, the M1 is effectively a quad-core chip, with four low-power cores to handle low-power workloads and provide a little extra performance boost — to address a huge range of markets, while drawing so little power, it can also be sold in a laptop.
Part of the reason Apple can get away with doing this is that — and let’s be honest — it’s been selling badly underpowered systems at certain price points. The old 21.5-inch iMacs included a $1,099 option with a dual-core CPU and only a 3.6GHz (no turbo) quad-core at $1,299. Only the six-core iMac, at $1,499, had a CPU powerful enough to even arguably be shipping in a 2021 PC. That matters because, when these systems get reviewed, they’re going to be compared in part with the hardware they replaced. The M1 appears to be faster and more power-efficient than current x86 CPUs, regardless, but it’s going to compare particularly well when the other systems are underpowered relative to what a PC OEM would have been selling at the same price point.
But lopsided configurations are only part of the equation. Apple couldn’t position the M1 this way if it wasn’t an excellent CPU in its own right. The M1’s dramatically higher efficiency and improved performance relative to x86 allowed Apple to standardize on a single CPU core across a wide range of products and price points. This is in complete opposition to the way PCs are traditionally positioned.
Where OEMs Fear to Tread
x86 OEMs almost never standardize a product family on a single CPU. If a vendor offers more than one thin-and-light laptop in the same product family, the upper-end SKU will offer a higher-end chip. Workstation and gaming boutiques that offer premade systems follow the same pattern. Step up the price stack, and the amount of CPU horsepower available increases.
There’s an obvious historical reason for this. For most of the past four decades, higher CPU performance cost more money and was more difficult for companies like Intel and AMD to deliver. In the modern era, performance advances have slowed, and minimum spec requirements have slowed with them. Apple is betting that the M1 is good enough to serve a much larger range of markets than x86 can typically address and that the M1’s innate performance is good enough to satisfy consumers at this range of price points.
The implication of slides like this, from the M1’s launch, is that the M1 can address a much larger range of use cases than a conventional x86 CPU, by virtue of being so ideally positioned.
If that doesn’t seem like a fusillade across x86’s metaphorical bow, consider the issue from a different perspective: According to Apple, the M1 is the right CPU for a $699 computer, and a $999 computer, and a $1,699 computer. It’s the right chip if you want maximum battery life and the right CPU for optimal performance. Want the amazing performance of an M1 iMac, but can’t afford (or have no need) for the expensive display? Buy a $699 Mac mini, with exactly the same CPU. Apple’s M1 positioning, evaluated in its totality, claims the CPU is cheap and unremarkable enough to be sold at $699, powerful and capable enough to sell at $1699, and power-efficient enough to power both a tablet and a pair of laptops priced in-between.
No single x86 CPU is sold this way or positioned as a solution to such a broad range of use cases. There are three reasons why. First, PC customers generally expect higher-end systems in the same product family to offer faster CPUs. In the past, both Apple and x86 systems were sold in such fashion. Second, Intel and AMD both benefit from a decades-old narrative that places the CPU at the center of the consumer’s device experience and enjoyment and have designed and priced their products accordingly, even if that argument is somewhat less true today than it was in earlier eras. Third, no single x86 CPU appears to be capable of matching both the M1’s power consumption and its performance.
Apple’s willingness to position the M1 across so many markets challenges the narrative that such a vast array of x86 products is helpful or necessary. It puts Intel and AMD in the position of justifying why, exactly, x86 customers are required to make so many tradeoffs between high performance and low power consumption. Selling the M1 in both $699 and $1,699 machines challenges the idea that a computer’s price ought to principally reflect the CPU inside of it.
Apple’s gamble, with the M1, is that its custom CPU performance is now so high, at such low power consumption, that the choice of chip inside the system has become irrelevant within a given product generation. It challenges OEMs to consider how they might spec higher-end systems if some of the higher price tag didn’t have to pay for a faster CPU. It threatens both AMD and Intel with the idea that a CPU with just four high-performance cores is fast enough for the company to 1). Call it an eight-core chip and 2). Position it against the 6, 8, and 10-core Intel and AMD systems that sell for $1,500+ in the desktop and laptop markets.
There Are Still Higher-End CPUs Coming
The rumors of higher-end M1 configurations aren’t wrong just because Apple is using the same CPUs for this set of iMac refreshes that it launched in 2020. Unless Apple intends to drop the Mac Pro altogether, it will have to build a higher-end chip to support professional users.
The M1 is, at heart, a 4+4 CPU configuration with four FireStorm and four IceStorm cores and an attached pool of on-package DRAM. We know Apple can scale up to 16GB of DRAM for the M1 already, but the current Mac Pro offers up to 1.5TB of RAM, and that’s a bit much for an on-package array. Even if we assumed every FireStorm core was 3x faster than an x86 core, Apple has just four FireStorm cores in the M1 and it currently sells a 28-core Xeon W. There will be a need for faster M1 CPUs with higher core counts in order to replace x86 in these markets.
Apple can stretch a single M1 SoC design across desktop, tablet, and mobile, but it can’t stretch a 4+4 CPU with 16GB of on-package RAM to address the workstation market. The most likely outcome here is a future M-class CPU with eight to 32 “big” cores and a conventional DRAM interface, based on either DDR4 or DDR5. This version of the SoC would presumably omit the on-package DRAM in favor of more CPU and possibly GPU cores. A different cache structure with a larger L3 or L4 would also be a possibility.
There’s no chance Apple can address the entirety of its market with a single SoC, but it might be able to do so with just 2-3 different core designs, depending on whether it wants hypothetical future 27-inch iMac refreshes to offer their own CPU differentiation. But whether Apple ultimately introduces a few SKU options or not is less important than the message the company is sending here: CPU performance, according to Apple, is impressively fast, surprisingly low-power, and quite inexpensive. The M1 is such a good CPU (according to Apple), that non-professional Apple customers no longer need to care about their CPU at all. Professional Apple customers at the top of the market will end up needing to care minimally, but the Cupertino-based company will probably still offer fewer CPU configuration options than it did in the past.
PC users who don’t like Apple’s ecosystem or who need more CPU performance than Apple sells obviously won’t be convinced. But it’s a unique, distinct argument from anything we’ve ever seen. We’ll see if the company sustains it as it introduces new ARM-powered hardware later on this year.
Now Read:
The Link LonkApril 22, 2021 at 06:03PM
https://ift.tt/3sH0IU8
Apple's M1 Positioning Mocks the Entire x86 Business Model - ExtremeTech
https://ift.tt/2YXg8Ic
Intel
No comments:
Post a Comment