The generation of Apple's M2 chip MacBook was born very bumpy.
In the development process of M2 chips, N3, which was originally the "morning train" to catch up with TSMC's 3nm process, is the first generation of 3nm.
However, N3's yield, process level and cost have not reached the ideal state, and after TSMC's trial production in the second half of 2022, including Apple, Qualcomm, and AMD have canceled orders and have not really put into mass production.
Even Apple, a wealthy party A, has not been able to recognize N3, which shows that TSMC's first-generation 3nm does not have any cost performance.
With M2 once again choosing TSMC's second-generation 5nm process, even though Apple changed the new architecture and increased the number of GPU cores, there was only a slight improvement in theoretical performance in the end.
Moreover, what is more deadly is that the MacBook series (Air, Pro) version of the SSD has a slowdown, and it still costs thousands to add memory and storage, and the reputation of M2 comprehensive products is not as good as that of M1.
In addition, there is an objective reason, as working from home gradually becomes history, the demand for PC overdrafts has rebounded, and the PC market has shrunk to a certain extent at the beginning of the year.
Among them, Apple's Mac even lost nearly 40% of the market share.
For subjective and objective reasons, Apple also sent a request to TSMC to stop the production of M2 and cut some chip orders.
At the same time, perhaps because of the less than ideal iteration of the M2, Apple has not adjusted the Mac product line.
Instead, a regular chip iteration was carried out, and even the conference was not held for M2 Pro and M2 Max, and product lines such as iMac and Mac Pro were also absent.
It seems that for Apple, everything explodes, waiting for M3.
M3 at 3nm, will it be a savior?
Around the end of last year, the M3 chip had news of running scores.
From the score point of view, the M3 basic chip is stronger than the M2 Max in Geekbench running score, and the performance is only the level of the basic model.
On this basis, M3 Pro and M3 Max may have a more "terrible" energy efficiency ratio.
At the same time, it also shows from the side that TSMC's second-generation 3nm process (N3E) has become mature, sweeping away the haze of N3.
▲ The eighteenth fab responsible for the production of 3nm chips Image from: TSMC
However, in view of the fact that most manufacturers have cut orders for N3 before, TSMC's previous huge investment has been lost, which has also affected its own 5nm and 7nm production capacity and layout.
In order to make the N3E transition and mass production smoothly, TSMC made a little adjustment to the N3E.
Compared to the initial plan, N3E will use fewer EUV lithography layers, from 25 to 21 layers, reducing the difficulty of going to production.
The cost will fall, but the associated transistor density will be 8% less than N3 and still 60% higher than N5.
In terms of specific energy efficiency, TSMC said that N3E will have an 18% performance improvement and a 32% performance improvement compared to N5.
Simply put, TSMC provides a shrunken version of N3E, which reduces process complexity, which in turn makes the cost lower and more pragmatic.
According to Apple's previous M chip upgrade strategy, the number of transistors directly represents its performance.
More transistors, and does not rule out that Apple in the M3, change the configuration of 8 cores (4+4) CPU, add more cores to it, such as similar to M1 Pro, is likely to make its performance a significant improvement.
In addition, for GPUs, thanks to N3E, it is not ruled out that Apple will add ray tracing characteristics to be consistent with mainstream chips.
▲ Arm 2022 GPU architecture ray tracing demo Demo (GIF with compression), pay attention to the change of light and shadow Image from: Arm
By the way, the A17 with N3E does not rule out joining optical tracing, after all, optical tracing is already the mainstream configuration of Qualcomm and MediaTek's flagship chips.
Previously, Apple executives made it clear in an interview that they would make Mac games and rely on powerful hardware and development platforms to attract developers, so light chasing is essential.
For TSMC, a new process node often reaches a "dessert position" in the second generation.
5nm N4 and N4P have obvious energy efficiency improvements, and with the investment of production lines, manufacturing cost control is also better.
In TSMC's 3nm process plan, there are also multi-generation versions, and the adjusted N3E cannot be determined whether it is a dessert for the time being, but it may be the most realistic process for TSMC to put into production.
For Apple, the M3 based on N3E may also be more like a large version of the chip iteration, and it is more worthy of upgrading.
Of course, it would be better if Apple changed the SSD to slow down.
With the M3 chip, the Mac can sweep away the shame?
Unlike Intel, AMD, and NVIDIA, Apple's M chip is not supplied externally and is a rare self-produced and self-sold.
This means that the M series chips are for Mac, MacBook and other products and services, and the ultimate goal is to sell more computers.
As mentioned earlier, the reason why M2's MacBook does not exist strongly has subjective and objective reasons.
But whether the M3 series of Macs can "sweep away the shame" also has subjective and objective factors.
Subjectively, the new Mac series in the M2 series is only the MacBook Air, the new design style, good looking or not, but does not continue the "wedge" classic design, coupled with complete passive heat dissipation, so that it itself has some controversy.
And, Apple kept the M1 MacBook Air and the old mold M2 MacBook Pro 13 products, quietly increased the starting price of the M2 MacBook Air in order to distinguish.
is beginning to quietly break Air's original bias towards entry, Pro is a price differentiation of the professional level, positioning needs to be re-combed.
In addition, perhaps because the M2 series of chips have not been improved enough, Apple has not changed the iMac, Mac Studio and Mac Pro.
The MacBook Air 15, which has been rumored before, is thought to be released at WWDC, and this node can only be equipped with the M2 chip.
The MacBook Air 15 should be similar in size to the MacBook Pro 16, but with wider bezels
Its biggest selling point is actually 15 inches, positioned somewhat similar to the iPhone 14 Plus, suitable for ordinary needs of performance, with a larger screen, to create a "pragmatic" demand.
Many analysts believe that the MacBook Air 15 actually came about two purposes. One is to test market demand, and the other is to empty the inventory of M2.
M2 chip, there is a product that is temporarily pulled out to support the field, only for transitional use, to be precise, it can actually be called M1.5.
▲ So far, the M chip Mac Pro has not been launched, as if waiting for the M3
The strength of M3 may allow Apple to reorganize the product line, and market segment needs such as iMac and Mac Pro will also be re-accommodated in the M3 lineup, which may be more like an iteration of the entire product line.
Objectively, it is also the consciousness of competitors (in this case, Intel).
Before the release of the M1, Intel was a bit crazy on 14nm, and the lack of innovation in the processor was a little flat.
Even if AMD gives a fairly competitive product in the CPU market, Intel is still unmoved.
Until the emergence of Apple M1, the use of large and small cores, and the ultimate energy efficiency ratio, people re-recognize laptops, it turns out that unplugged can really be used all day.
The M1 has brought a deeply recognized technological "leap" to the masses.
Coupled with Apple's influence, it impacted Intel's weakness and made Intel wake up.
Intel also introduced large and small cores in the 12th and 13th generation products, adding the hard solution of multimedia formats, and Intel 12th and 13th generation CPUs almost regained the attention of high-performance CPUs.
The advantage of Arm-based M chips lies in the energy efficiency ratio, and the high performance is actually inferior to x86, plus the M2 chip was not put into production according to the 3nm plan, and finally failed to launch the M chip Mac Pro, there are also such considerations.
Even if the M3 does reach the ideal state of TSMC N3E, it is difficult for the public to experience a technological leap similar to the M1.
For the public, the adjustment and subdivision of the Mac product line, than the strong chip, may be the winner and loser to promote Mac sales.
The blessing of M3 will make this generation of products have a longer life, and will also attract more Intel Mac users of the old era to update.
Bob Borchers, Apple's vice president of product marketing, said in an interview that Apple's next focus on business growth and goal is to gradually replace the old Intel Mac in the hands of old users with the continuous update of M chips.
Compared to expanding other businesses, this replacement demand will be the greatest potential for the continued growth of the Mac business.
The new process of M3 chips, as well as the reworked Mac series, is Apple's dual strategy of replacing the old devices in the hands of old users.
The 3nm process, for M3 chips, is really a deciding factor. But for the boost Apple wants to sell, it's just one small aspect, and more important is how Apple defines different Mac products.