Apple Silicon has been in the fourth generation in five years.
6 Facebook
We are crawling on the five-year anniversary of the apple silicone Tim Cook and for a fifth of servants. This is how the whole saga began, how she launched and what to expect in the future.
After many years of work with the restrictions on the Intel Apple chip, she introduced the transition to Apple Silicon during the WWDC 2020. The efforts to cancel the well-used architecture will see that Apple will go to the new line of the developed internal chips for their Catalog MacBook, starting from the same year.
This two -year transition from Intel to Apple Silicon began serious changes for the company, as well as the industry as a whole. In addition to claims to high efficiency structures in combination with a high level of performance that you expect from the launch of the chip.
Apple Silicon also presented the most promising function of the iPhone chips to the desktop. The addition of a neural engine to the M “M” Chips M “was forced the rest of the computing industry to consider the possibility of moving in similar ways using its own selection of the processor.
Apple General Director Tim Cook, introducing Apple Silicon during WWDC 2020, the lifting of loan: Apple
the chip basically followed the basic idea of the processor with an integrated graphic graphic A processor that has been and remains a fairly ordinary concept during processing. The neural engine, of course, added more performance in tasks that could benefit from machine learning, but another change from the norm also helped.
Instead of using discrete memory bullets for CPU and a graphic processor, as well as other components, Apple decided to use a unified memory. It went against the usual concept, using one large memory pool to which all the processor elements could get.
The idea helped performance in several ways, including permission to components to access the same data in memory without the need to duplicate it.
there is also an advantage in the GPU that the more unified memory was turned on, the more memory it had access for graphic work. This allowed the GPU in Apple Silicon to have access to a much larger amount of memory than you could find in modern discrete graphic processors or video cards for PCs.
While the introduction of Apple Silicon at that time was focused on chips of the M1 and M series designed to use the MAC, this term has expanded in its use. Now it belongs to all the main processing chips of Apple, including its series A used in the iPhone and iPad lines, which were also predecessors for the M-series M.
Apple Silicon everywhere, this is not only processors, and this is not only in Mac. But in order to make things simple this time, we will discuss Apple Silicon from the original Mac-oriented point of view.
, but even just holding the MP series, the Progress Apple Silicon has reached a stunning and industry from the moment of its introduction.
Silicon from the preliminary phenomenon: troubles with Intel
Apple decision to make your own chips was related to the problems that were using the company.
In the 2010s, Intel was the main computational force. It was the status that he earned for years, but this is not the fact that the manufacturer of chips could maintain.
“There was no extremely difficult to continue to keep up with the Moore law” for Intel, due to the constant reduction of the matrix. Each new improvement in the process added new obstacles, while the subsequent contractions of the matrix continued to improve.
ultimately Intel rejected its known frequency of such a change in process change process (Tick), followed the update of architecture in this process (TOCK). Instead, by 2016, Intel switched to the Tick-tock strategy, linking the optimization phase at the end to get more from the process before it is reduced to the next.
inevitably, delays with shrinks of the matrix, such as the introduction of Intel 10-nanometer chips, were too large for Apple could deal.
This was only one of the problems that Apple faced Intel, and the heat created by chips and high energy consumption became great problems that increased. In 2018, this led to the issues of a thermal throttle affecting the updates of the MacBook Pro this year.
the throttle was serious enough to be a problem, and is aggravated by poor firmware. Consumers quickly gained popularity.
After our own testing, Mike Voerites from Appleinsider discussed the throttle on the new Leo Laporte screen.
it is noteworthy that Mike said that the Apple Silicon will ultimately be called in 202020 at the low end of the Mac.
There were rumors that until this year, when Apple brought more of her work with mobile chips, including some went back in 2011. The transition to the development of Mac Chip was, it would seem, inevitably, since Apple's experience continued to grow.
, of course, he had barriers to pass to get there, but over time he had knowledge and the ability to go from Intel.
It seems that the unintentional Intel relationship with Apple gave the iPhone manufacturer more incentives to make a jump on the Mac.
he has already mastered the control of thermal and powerful needs for his chips to extend the service life of the iPhone battery and keep the smartphones from overheating. It made sense that his knowledge can be used on a desktop class processors for a large effect.
working with a longtime TSMC chips partner, this is exactly what Apple did.
M1: Original releases
The first equipment of Apple Silicon AdPly, which Apple introduced Apple. After the announcement of Tim Cook was not using the M1 chip. Indeed, it was an A-Series chip.
The developer’s transitional set was distributed in the form of an individual Mac Mini, gutted Intel components and instead the completed Apple Silicon Hardware. The center of this was the Apple A12Z Bionic, which is a good choice as the most powerful chip of the A series, and it is very similar to the fact that the developers could expect from the first official output based on ARM.
different sizes of the chips of the M1-image Credit: Apple
The M1 series began with M1, created using a 5-nanometer process. He had 16 billion transistors, four high -performance kernels and four energy -efficient nuclei, as well as seven or eight -core graphics processor.
as for the neural engine, it was a 16-core component capable of making 11 trillion operations per second. While memory values were low by 8 GB or 16 GB, they also worked with 68.25 GB/s memory capacity, glacial compared to modern versions.
it took almost a year for Apple to send M1 Pro and M1 Max, which brought more power and performance.
PRO had six or eight performance cores in the processor, as well as two kernels of efficiency. The maximum version had eight performance cores and four kernels of efficiency.
graphic processors on each also saw an increase in the main nuclei, by 14-16 on Pro and from 24 to 32 at the maximum. The neuron engine remained the same, but the memory capacity increased to 200 GB/s and 400 GB/s, respectively.
Another addition that the M1 chip was completely missed by the media engine, or the selected equipment for decoding and encoding video. This function made M1 Pro, and then fixes fantastic for video editors and content, especially those who work on industry standard prores.
very later arrival in March 2022 was M1 Ultra, another innovative chip. Using the relationship, Apple effectively boiled two M1 -max pins M1, creating something, which was twice as much in all respects.
, which meant up to 16 productivity nuclei on the processor, as well as eight effectiveness nuclei, the number of main graphic processors up to 64 nuclei and a 32-point neural engine. Since these are two chips attached together, it made sense that she had two things that Max offered.
As the number of lines and the nucleus grows, you can also see an increase in performance. The capabilities of each core did not increase, but the total amount was even more useful for end users.
, using the latest results Geekbench, it is easy to see this progress in performance. For comparison, we use the upper score for each chip, regardless of the amount of the nucleus or model.
Sonuclear control indicators-M1
in the same-core test there were all four levels, lower or about 2400 points. This makes sense, since the test uses only one quick core, and they are quite identical in all directions for generation.
multi-core control indicators-M1
Multi-core control indicators begin to show some differentiation, and this is primarily due to the number of nuclei for everyone. The more cores for use, the higher the assessment.
Metal tests-M1
this is much more obvious with metal tests, since they rely on the number of main graphic processors. There are significant jumps between each chips.
m2: the next generation
, launched in June 2022, the M2 series began similarly with M1, with the introduction of the introduced chip-show. Again, a 5-nanometer was used, but Apple increased the number of transistors to 20 billion with 16 billion M1.
Basic statistics of chip M2-Rinse Credit: Apple
the prevention continued to have four-effective, four cores, but using the kernel improvement compared to the previous generation. In total, the graphic processor also won with options up to 10 cores.
The neuron engine was more lived, with 16 cores capable of making 15.8 trillion operations per second. Memory throughput was also increased to a round of 100 GB/s.
After his established book on the game M2 Pro and M2 Max arrived in January 2023. Again, there were more cores, since the M2 Pro had six or eight performance cores and four effectiveness kernels, while MAX had eight performance and four effectiveness kernels.
The parameters of the graphic processor also increased, from 16 to 19 to Pro and from 30 to 38 nuclei on Pro Max. The same level of updates of the capacity of memory for their appropriate levels was present again.
until June Apple required to release M2 Ultra. Once again, uniting two M2 Max Chips together, it had 16 performance cores, eight effectiveness nuclei, a 32-point neural engine, up to 76 graphic processor cores and used 134 billion transistors.
The memory manual was also phenomenal, and it is able to use until 192 GB. Everyone benefits up to 800 GB/s in memory of memory.
Sonuclear tests-M2
based on performance, the same productivity was generally better, but with a slight differentiation again. The multi -core score shows much more, and the M2 Pro and M2 Max are quite close in the middle.
multi-core control indicators-M2
Once again, the GPU side shows the M2 Ultra as massive productive. This made him loved by visual creators for a rather long time.
Metal Bendcharks-M2
M3: “Third Album
” Third Album “
The introduction of M3 in October 2023 was very very Unusual, mainly because Apple did not just introduce the M3 chip. Transistorov was offered for users a very familiar make-up, there was an eight-core processor with four productive and four kernel cores, as the last time, and a 10-core graphic processor again. The processor and graphic processor were 35% and 65% better in accordance with the chip M1.The neuron engine also received some performance, and the 16-core component was updated to work up to 18 trillion operations per second. This time, this time was a technology called dynamic caching, which allowed the graphic processor to highlight the use of local memory in real-time equipment. This meant that he could single out only the amount of memory he needed, increasing use. GPU also received hardware tracing of rays and shading of the grid. This helped create visually complex scenes in games and other graphic applications. M3 Pro and M3 Max really differed from the norm, as Apple differentiated processors more than usual. Pro had a 11-core or 12-core processor, with five or six performance cores paired with six effectiveness cores. Singerly-core tests-m2 M3 Max had 10 or 12 performance cores, as well as four kernel kernels. In addition to the presence of more nuclei than the Pro, the M3 Max had fewer efficiency nuclei. The memory capacity was a little step back for this generation. The rates are 150 GB/s and 300 GB/s this time lower than their analogues M2. When it came to the graphic processor, the M3 Pro had from 14 to 18 nuclei, while the M3 Max had from 30 to 40 cores of the graphic processor. multi-core control indicators-M3 M3 Ultra was even more mystery, since it did not come in a timely manner. The speculation of rumors about whether the chip was killed in favor of a faster transition to the M4 generation. ultimately appeared, however, in March 2025. These were months after the release of M4 Pro and Max, the next generation. Metal tests-M3 as usual, M3 Ultra is carried out by holding out two M3 Max Chips together, like one. This meant the number of cores of the processor of 28 or 32 nuclei, with 20 or 24 performance nuclei paired with eight effectiveness nuclei. also turned on the 32-core neuron engine, as well as 60 or 80-core GPUs. The memory capacity jumped to 512 GB of single memory and offered 800 GB/s through memory. When it comes to performance schedules, the main thing you can see is multi-core results. Unlike generations, M2 and M3, you can see a significant separation between the chips M3 Pro and M3 Max. M4 is the current generation of Apple Silicon for Mac and the highest collection. , introduced in May 2024, M4 used the process of three nanometer and compressed in 28 billion transistors. It was a nine or ten -core processor, with three or four productivity kernels mixed with six effectiveness cores. M4 included six effectiveness cores for this time Credit: Apple there was a ten-core graphic processor with the same updates as the M3 and 16-gray neural mechanism, capable of 38 trillion operations for a second. Memory throughput improved to 120 GB/s. M4 Pro pushes specifications for a 12-core or 14-core processor with eight or ten nuclei and four kernel nuclei, as if Apple has learned its lesson from M3 Pro. The M4 GP-graphic processor was available in 16-explosive and 20-core varieties, with memory capacity up to 273 GB/s. Sonuclear control indicators-M4 M4 Pro also improved on the support of Thunderbolt, from Thunderbolt 4 to Thunderblt. Strips for data transfer between Deffers, AT 80GBP versions. M4 Max went even further than the M4 Max, with the choice of 14-core or 16-core processors with 10 or 12 performance cores and four efficiency kernels. The GPU rose to a 32-core or 40-core version, and the memory capacity increased to 546 Gb/s. multi-core control indicators-M4 during the reporting there is no M4 Ultra, but one is not expected. Apple himself says that not every generation of silicon apple will have ultraplet, so there are no guarantees that can be on the way. Based on performance, the single -core results are again quite even, and a maximum slightly higher than the rest. Metal tests-M4 a multi-core test, again, shows progression in productivity, which is due to the number of nucleus. Although this time the difference between Pro and Max is slightly less pronounced than the M3. it is not surprising that the results of the metal show the same result based on the main results. The more cores, the higher the result. Apple showed that it improved the Apple Silicon as a result of the generation. New versions saw advantages compared to the older version in almost all cases, with several between the generation. The main changes over the years were the main speeds and quantity of the core. This is easily demonstrated by comparing all the core results in one diagram. All M-series ballots for a single-core testing for Geekbench you can see an improvement by 60% between the basic chips M1 and M4, increasing to 62% for M1 Max and M4 MAX. Indeed, you can see rather neat jumps between generations between versions. When you turn to multi-core assessments, it is obvious that the clock speed is only one factor. An increase in the number of nucleus can also increase the differences. All M-series cylinders for multi-core tests for Geekbench The base M4 is 74% faster than the M1 equivalent, but the upper levels see greater growth. By the time you compare the M1 Max and M4 Max, the improvement increases to 203%. Turning to the graphic characteristics under the metal, the biggest benefit is with basic versions. Between M1 and M4 chips, there is a difference of 74% in the results of Geekbench metals. All M-series M-series M-series M-series Metals for Geekbench This does not mean that others and Max see 65% and 62% of equivalents in the difference of three generations. Even a jump in two generations from M1 to M3 Ultra is an improvement of 61%. One thing that is very obvious, looking at the charts together, is that updates are quite consistent in generation. Each time there is approximately the same level of improvement that it portends well for the future. , if Apple is both a point and a significant creature. That is, what will arrive in the next few years for apple silicon should be more than the same. quasi improvements in all directions and the same type of generation structure as now. Friday additions of ultra -microxes to the side. Apple General Director Tim Cook during the introduction of Apple Silicon-Image Credit: Apple is expected that the generation of the M5 will appear soon, and it is supposedly already in mass production. As for the fact that there may be a new thing this time, it was assumed that Apple will shift to separate the processor and graphic processor, which can further increase productivity. there were even complaints that the M6, which takes modemic technology of Apple, to the solution for mobile devices for mobile devices. Think about mobile communications at the MacBook Pro, not in need of binding to the iPhone. Although we do not know which future will accurately hold for apple silicon, we can confidently say that it will still be better when time comes.
Follow Appleinsider in Google News M4: Continuation of growth
from M1 to M4
in the next five years