According to the latest report from Bloomberg, Apple is developing a new ARM processor with up to 32 high-performance CPU cores, which may appear in Macs in the second half of 2021. The processor may also appear in the new "half-size Mac Pro" in 2022.
According to reports, the company is also developing a CPU design with 16 high-performance cores and 4 energy-efficient cores, which will be used in new versions of MacBook Pro and iMac.
The new processor may be available as early as the spring of 2021. Future Apple Silicon designs can also use GPU designs with up to 128 dedicated cores.
When Apple released its first Macs with its own chips, news of their new processors came out. According to reports, the company's new MacBook Air, MacBook Pro and Mac mini all use the company's M1 chip, which has a CPU with four high-performance cores and four high-efficiency cores.
However, the company's more powerful machines (such as Mac Pro) continue to use Intel chips. Apple has stated that it intends to transition the entire Mac product line to its own chips within two years.
In addition to the increased number of CPU cores, Bloomberg reported that Apple is also developing chips with more GPU cores. Although the current M1 chip has seven or eight GPU cores, Apple is currently testing models with 16 and 32 cores, and is developing chips with up to 128 chips by the end of 2021 or 2022.
Although Apple is working on a processor design with 16 high-performance CPU cores, Bloomberg pointed out that depending on the mass production method, it may choose to only enable 8 or 12 cores to release it.
Before Apple announced the switch to ARM-based processors, Bloomberg reported that it was developing a processor with a CPU with eight high-performance cores and four energy-efficient cores. A processor with this precise core combination has not yet been officially released.
The report also revealed several tidbits. First, products directly related to the M1 will power the new iPad Pro models that will be launched next year.
Secondly, the faster M1 successor for MacBook Pro and desktop computers will also have more GPU cores for graphics processing-especially 16 or 32 cores.
In addition, Apple is working on "advanced graphics upgrades for 64-bit and 128-bit dedicated cores dedicated to its high-end computers" by the end of 2022 or 2021.
When we reviewed the M1 version of Mac mini and MacBook Air, we found that their performance beats similar Macs with Intel chips very well.
Part of the reason for these performance improvements is that, according to Apple, it is mainly due to its new and improved unified memory architecture, which allows CPU, GPU and other components to quickly access data in the same shared fast memory pool , Without reducing the efficiency of moving or copying data around.
When we interviewed Apple executives Craig Federighi, Greg Joswiak and Johny Srouji, they claimed that the M1 is just the beginning of a leap in Mac performance based on this architecture.
Considering Apple's ambitious plan to transfer its entire Mac series to its own chips within the next two years, it makes sense that it has a more powerful chip under development. Its first ARM-based Mac has left a deep impression because of its high energy efficiency and performance, but its self-developed chip equipment must match a more powerful Intel-based machine (such as Mac Pro). May face greater challenges.
Apple's chip strategy will create a massively parallel world
This change of Apple is of great significance to technology decision makers, because it heralds how the war between data centers and high-end professional computers will be carried out-using a large number of micro-processing cores to provide a combination of power and performance, this is the Intel Xeon and Core chips may be difficult to deal with.
Apple’s obsession with chips that consume less energy and run cooler temperatures may redefine the traditional server array. In this server array, rack-mounted computer towers currently consume countless amounts of power, and adding more processors has historically been Upper means increased rack and power consumption. Now, the cloud computing company has installed four Mac minis in a single 1U rack. , The number of Macs in the rack is about to surge.
Last week, AWS announced that it is now offering Elastic Compute Cloud instances on Mac mini. They will start with Apple’s last Intel-based computer, but plan to use Apple Silicon equipment in early 2021. Compared with smaller cloud hosting competitors, it is admiring full integration with AWS services, fast startup, and the ability to scale to multiple machines, which is why it prefers its products.
Amazon specially launched Mac products using six-core Intel Core i7 chips, which means that the transition to the first-generation M1 Mac will not produce the above-mentioned multiplication effect. Even without a substantial increase in the number of cores, M1 provides about 70% single-core improvement and 30% multi-core improvement when compared to i7. For cheaper machines with lower operating temperatures, this is not an easy task.
But if the report is correct, the Mac mini using the second-generation Apple Silicon can not only be faster, but can also handle a large number of tasks in parallel. The micro server can be moved from today’s eight cores to next year’s 20 cores, without any changes in appearance, and even Apple can reduce its size so that more Mac minis can be installed on the shelf.
None of the first three Mac computers with M1 adopted a new design, but this situation will change in 2021. Among them, the new iMac inspired by the iPad Pro is almost certain, reflecting Apple Silicon’s upcoming tablet-thin Mac. Similar rumors indicate that Apple will release a Mac Pro that accounts for only a quarter of the current model volume, which may reduce the current 5U-sized rack-mount chassis to a 2U rack.
Although this is not a specific guess at this stage, it may mean that the data center will be able to put five Mac Pros and Apple Silicon into a rack that can only accommodate two.
It is worth mentioning that two long-term development trends enable Apple Silicon to provide such outstanding performance. One of them is that energy-efficient RISC processors based on ARM technology continue to march towards smaller chip manufacturing nodes. The most famous is that this year's 5nm process was perfected by Apple's manufacturing partner TSMC (TSMC). The other is the powerful multi-threaded OS technology created by Apple, such as Grand Central Dispatch, which can effectively route application tasks to multiple cores without requiring developers or users to be vigilant.
The direct result of these innovations is that Apple's chips and operating system can now seamlessly expand to as many cores as can be accommodated on the die, so that each new generation of products can obtain significant speed and function improvements.
Conversion will have a multiplicative impact beyond the CPU performance. Even Apple’s original M1 chip is equipped with 16 dedicated AI cores, which can provide 11 TOPS performance support for general machine learning tasks and computer vision applications. Just this month when Qualcomm’s new Snapdragon 888 system-on-chip increases the mobile AI index to 26 TOPS, Apple will bring higher-performance AI hardware to its mobile and Mac chips, and each device will be equipped with a decade ago. Unimaginable performance.
Apple’s GPU strategy appears to be similar to that of Nvidia, which triples the number of processing cores in its graphics card to meet the needs of demanding gamers and imaging professionals. However, although Nvidia is now selling graphics cards with more than 10,000 graphics cores and Tensor AI cores, they require the maximum total power of the system to reach 750 watts. However, Apple's biggest ambition in graphics is only 1/100 of that. Assuming that it actually releases a 128-core GPU and uses it on devices released in 2021, Nvidia's professional applications will continue to maintain its leading position, and it will also maintain its lead for the foreseeable future.
More importantly, Nvidia also mastered the ARM architecture on which Apple's own chips depend, although Apple relied on ARM technology to create its own CPU and GPU cores instead of relying on off-the-shelf ARM chip designs.
In summary, the main result of all these developments is-we will see a surge in the number of Apple processor applications. This is not only in consumer applications. Once Apple Silicon enters the data center, each Mac will be supported by multiple Apple CPUs, GPUs and AI cores; each rack can accommodate multiple Macs; each server farm will be able to handle more Multiple workloads provide support for all functions from B2B applications to enterprise data warehouses and cloud games.
It is not clear whether AMD, Nvidia, and Intel see Apple’s growth as an existential threat, or only through traditional annual competitive improvements to deal with, but compared with the future, the future of processing will obviously be larger and will develop in parallel . Because Apple will construct its own world to win over more data center markets, developers and money.
