Author: Keith Ramirez

The arrival of Apple’s M1 chip to your computers is a perfect time to compare with other computers in its price range and even higher. As a package containing CPUs, GPUs, Apple Neural Engine and more, this is a piece of hardware worth further dissecting .

Let’s see what makes the M1 a revolutionary chip in benchmarks and in real use, with facts and figures that put Apple’s milestone in context, and that may soon be from Qualcomm, HiSilicon, Samsung or Nvidia itself .

The first surprise: the concept of consumption at rest changes with the M1 thanks to its Icestorm cores

After many years with Intel, in which the consumptions could be consulted with Intel Power Gadget and ‘Activity Monitor’, what is most surprising in the M1 is that once all the common applications in daily use are open, the use of the cores is very low and, above all, that the four high-performance CPU cores, called Firestorm, are hardly used once everything was working.

That totally changes the concept of rest , since it can be considered that almost any task such as browsing the web, office automation, playing videos, etc., uses energy close to that which is consumed without doing anything. This is the real key to the amazing autonomy that we have seen in the reviews of the new MacBook Air and MacBook Pro: the chip consumes much less than a watt doing many of the day-to-day tasks , with tremendous consistency. Against this, the tests in Intel say that the energy oscillation of the processor is much higher, and rises more with tasks in the background.

The M1 has an eight-core CPU: four high-performance Firestorms, and four high-efficiency Icestorms, all of which can work at the same time.

When writing this on a 2018 MacBook Pro with an i5-8259U processor, and with only the text editor open, its consumption between half a watt of use and 7 watts. The M1, with the same use, does not rise from half a watt of consumption . Behind this magic are the four highly efficient CPU cores, which for me are the big surprise of the chip .

Apple had bragged that the M1’s four high-efficiency cores on their own were on par with the 10th-generation dual-core i3 of the early 2020 MacBook Air. That was one of the mentions of the critically-criticized Keynote presentation of the M1 that surprised, because when carrying the word efficiency in its name and not performance, the logical thing is to think that they are much less powerful.

For years, the ARM-Android world has been using the Cortex A55 architecture, used for high-efficiency cores complementary to those of the Cortex A75, A76, A77 or A78 architecture. So when arriving in M1, it was understood that those high-efficiency Icestorm cores from M1 were really small cores .

And they have a smaller size on the chip, but they are not small and useless at all . The concept is different from the A55 used in SoCs on Android, as Apple’s architecture is much more modern and powerful. According to Anandtech, in the Apple A14 of the iPhone 12, which also uses Icestorm cores, these cores are on par with Cortex A76 cores at 2.2 GHz, with 3 times more energy efficiency. We are not talking about something minor, as the A76 is used as a high-performance core in recent chips as powerful as the Snapdragon 855, the Kirin 990 or the Exynos 990.

 

Icestorm – decent power, amazing efficiency

 Those for small, high-efficiency cores are one of the great keys to the M1.

Apple said that the four high-efficiency cores consume a tenth of the power of the four high-performance cores , which are considered the jewel in the crown because they are the ones that allow computers to outperform their predecessors with Intel. So far, beyond Anandtech’s analysis, focused on the A14, which shares architecture with the M1 but differs, we have not been able to know if this was true.

This issue interested me both to know how it is that high-efficiency cores have so much prominence in the day-to-day, performing almost all the normal tasks of the system, and to know what they were really capable of , to know what Apple can do with them in the future and if he only used them exclusively.

To do this, I decided to disable the high-performance Firestorm kernels, something that the system does not offer, but which I managed with help, with good luck and using the command line. It’s not difficult at all, but it does require disabling System Integrity Protection. To measure the consumption of the M1 and the rest of the processors that have been tested we have used the command ” sudo powermetrics ” and Intel Power Gadget.

First of all, let’s see how these four cores compare to the fully functioning M1, according to Geekbench 5. This will give us a first idea of ​​which mobile and desktop processors they are comparable to. This is the result of running the famous benchmark using only the Icestorm cores :

 Geekbench 5 test using only the four high-efficiency Icestorm cores.

As we can see, the performance in single core is low , on the level of what Apple provided in the cores of the A9 in the iPhone 6s five years ago. Of course, that weakness is relativized thinking that it is the single core power of a Snapdragon 845, the star SoC in Android in 2018. As for multi core, it is also located in that band, in that of the Snapdragon 845, which had , yes, with eight cores in 4 + 4 arrangement.

IPhone, using all cores, the four Icestorm of the M1 are similar to the iPhone X A11 of power . To find equivalent Macs, we can go, for example, to 2013 or 2014, when my old MacBook Pro scored 515 points in single core. Of course, in multi core, the four Icestorm cores are comparable to what Apple promised : a MacBook Air i3 from 2020, also similar to the Pro with i5 two cores from 2017 of the seventh generation.

For even more perspective on the high-efficiency quad-core data, let’s review how they stack up against the Surface Pro X 2020, the best the world has ARM and Qualcomm in Windows 10 laptops. With quad-core high-performance running at 3.15 GHz and four high-efficiency, the SQ2 chip in the Surface Pro X 2020 is only 56% faster than the four high-efficiency cores in the M1 running at 2.06 GHz (3,100 points in Geekbench 5 versus 1977 of the Icestorm).

But the key to these cores, as we will see, is not their already decent power, but their efficiency. Let’s see, before moving on to that, the result in Geekbench 5 of the entire M1, as it comes from the factory:

Geekbench 5 test using all the cores from the M1 on the MacBook Air, as shipped from the factory. In single core one of the most powerful cores is always used, in this case a Firestorm.

The high performance Firestorm core is 3.32 times more powerful than the Icestorm . But the most surprising thing is that, as we will see, at one-third the power, these cores require only one-tenth the power of high-performance cores. In Geekbench 5, the peak consumption of the M1 is 18 watts, while using only the small cores, it is 1.4 watts.

Icestorm and Firestorm under the microscope with Cinebench R23

The M1 gets hotter on the MacBook Air than on the Pro due to the lack of a fan. Photo: Pedro Aznar.

Geekbench is a good benchmark, but it does not saturate the processor cores too much, and it does not allow only tasks for one core to be executed. Without having the means to test with industry benchmarks, such as SPEC2006 or 2017, the best benchmark to compare CPU performance is Cinebench R23, native to Intel and Apple Silicon and representative of real use , since it uses the same engine to render from Cinema 4D, which belongs to the same company that develops it, Maxon.

This test is where everything that Apple promised in performance is confirmed , and everything that we have intuited in efficiency. Let’s first look at the Cinebench R23 single core scoring results using the high-performance Firestorm core, the high-performance Icestorm core and one of the 16 “MacBook Pro cores released in 2020, featuring an i7-9750H six-core.

 

As the table expresses, the M1’s high-efficiency cores achieve a third of the high-efficiency cores. However, they do it with a tenth of the consumption, responding to the task for which they were born with an efficiency three times higher than that of their older brothers. With a score of just under half, Icestorm’s high-efficiency cores are 31 times more efficient than Intel’s i7 . An important part of this difference is explained by the fact that the Intel processor is still manufactured in a 14 nanometer process, compared to the 5 nanometer of the M1.

The high-performance Firestorm cores in the M1 are ten times more efficient than the i7-9750H, thanks to seven times less power consumption. And all this, operating with a similar frequency, around 3-3.2 GHz. It seems like a milestone against almost the best that Apple had in 2020 in laptops . I feel like calling it a revolution is over the top, but it’s just what it is.

In all tests, we will test the M1 with its eight cores in full and, in parallel, with limiting the chip to the four high-efficiency Icestorm cores (4C in the tables)

In the multi core test of Cinebench R23, we have used the same equipment. In the MacBook Air M1 we have done the test both as it comes from the factory, with its eight cores, as with the four of high efficiency exclusively . This is how it turned out:

 

As we can see, the situation repeats itself. The Icestorm cores in the M1 achieve three times the efficiency of the M1 overall, with about ten times less power. Compared to the i7, which it surpasses in the case of the MacBook Pro M1, which is better cooled than the Air M1, the M1 manages to exceed Intel’s efficiency by almost six times , with consumption also almost six times lower.

The M1 in real use: what benchmarks don’t always count

After having analyzed the performance of the M1 in dedicated tests, now it’s time to do it in real use, and the most relevant thing is that there are cases in which the benchmarks are no longer representative due to lack of optimization of the applications that we have tested or because we did not have in tell what elements like the Neural Engine could help.

ZIP compression: harnessing the power of the Firestorm

A very basic test in macOS is to compress a “certain” size file like the Ubuntu 20.04 ISO . It is something that we can do a lot depending on how we send files or what we need to compress. As there is the option to “compress” in the macOS Finder and in iOS (in the Files application) it seemed interesting to me to see how the different teams perform, and to see if the M1 has as much in common with the A14 as we think .

It’s a test where it’s the power of a single core that matters, on both iOS and macOS:

 

As we can see from the results, the M1’s single core power makes it outperform the 2020 16 “MacBook Pro and the 2020 four-port MacBook Pro 13” which continues to sell today with i5 for 2,129 euros. It’s the latest generation of Intel on a Mac, so it was interesting too.

Here we see that the M1 in the MacBook Air achieves the same as the i7 in the 16 “Pro in 77% of the time it takes the Intel chip , and in 68% of the time it takes the 10th generation i5 of 28 watts On an architectural level, it’s very interesting to see how the M1 pretty much matches the A14 in the iPhone 12, whose Firestorm core is somewhat slower in clock frequency.

The iPad Pro 2020 is more powerful than the iPhone 12 using all its cores, but having an architecture of two years ago, compressing with one core only manages to tie with the i5 of the MacBook Pro. It is a figure that makes sense, since both are more or less on par in bechnmarks.

Creating PDF from RAW Images

A very interesting optimized tool to do real performance tests is Automator, and therefore we will use it several times in our tests. In this case, we will generate a PDF from 28 Nikon RAW images, resulting in a final 2.48GB file . For this, the system reuses the resources of only one kernel. And therefore, we see once again the superiority of the M1 over the i7.

 

In the rest of the results there is a correlation with what was observed in the Cinebench R23 test, our reference to know the maximum performance.

Convert an H.264 video to Apple ProRes 422

Automator has also helped us to encode a 4K file recorded by an iPhone in H.264 to Apple ProRes. This encoding uses all the available cores quite intensively in both Intel and Apple Silicon , and the M1 is again at the top in the equipment used in the test.

 

Export 30 Keynote Pages to Apple ProRes 4444 Video

A very cool test that uses all available kernels is Keynote slide export to video . In my case, to take the equipment to the maximum, I opted to export 30 slides to a video of 2,880 x 2,160 pixels resolution with Apple ProRes 4444 codec. The logical thing would have been to export in H.264 or HEVC, but there the use of the cores it is very low thanks to the hardware coding, so we would not be comparing the same with respect to the equipment with Intel.

When I did the test for the first time, I realized that the generated file was so large that it took a long time to finish the export, not depending on CPU but on SSD. So since the behavior was the same in all the computers (stop CPU and saturate the SSD), I opted to stop the data timer just when the CPU usage was reduced to become irrelevant .

 

Here the M1 reigns supreme over the i7 in the 2020 MacBook Pro 16 “, which takes almost twice as long to export . The 2020 MacBook Pro 13″ is more than twice as slow. Let’s remember that both teams are now 2.39 and 1.8 times more expensive than the base MacBook Air M1 with which we are doing the tests.

Converting a Magazine Page from PDF to HEIC: The “Magic” of Hardware Encoding

One test I used to do to compare computer CPUs to Intel was to convert the first page of a magazine in PDF to HEIC, the compression-enhanced format that Apple uses to take pictures on iPhones, and for which its chips have dedicated encoding by hardware. Here it is seen that with some help from one of the cores (in consumption), the M1 with eight and four cores is too far from the rest.

And what is relevant again is not that it manages to be much faster than the i7, but that to achieve the same goal it consumes 20 times less energy , generates much less heat and does not set the fans to maximum (in the case of the MacBook Pro and the mini, which they do have).

Panorama creation in Adobe Lightroom: a warning about poor optimization

Adobe has recently released the optimized version of Adobe Lightroom (the non-Classic version). Performance is good overall, but after doing the test I had in mind, I was very disappointed. I have chosen the function ” Combine photos” -> “Combine panorama” , feeding it 29 RAW files. It is a demanding test and very dependent on the use of the cores.

The problem is that, despite being optimized for Apple silicon, the application is slower with native code than with the Rosetta translation . In fact, with Rosetta, the four high-efficiency cores manage to be faster than the full eight cores with native code. They are the first steps of Adobe in Apple silicon, but this performance is surprising not being a beta. In this case, as we will see now in Handbrake, the Intel teams do much better than the M1, although what I wanted was to show the importance of optimizing the applications well.

In our tests in both Lighroom and Handbrake, the performance was better with Rosetta than running native Apple silicon code

Converting an H.264 video to H.265 with Handbrake

Handbrake is a fantastic video converter with a multitude of options and, in its beta version, it has been updated to support the M1. It has hardware encoding options, but in our case we want to compare pears with pears, so we have done a software encoding. Specifically, we have downloaded the file “jellyfish-45-mbps-hd-h264.mkv” from this website.

Thus, we convert this 1080p video file with H.264 codec to H.265 codec and with MKV container. It’s a preset in Handbrake, so anyone can try it out. This test is very CPU intensive, and uses all cores .

As the test duration is somewhat longer, the M1 in the MacBook Air suffers without a fan compared to the MacBook Pro M1 with a fan . What is really interesting about this test is that it has helped us to confirm that some applications are not optimized despite running on native code, as Handbrake 1.4 beta already does. A yes, test Rosetta (Intel code) has been faster than the test with the Silicon Apple universal binary , as shown in the chart.

7-Zip compression with Keka: taking advantage of all the cores

Given the fiasco with Lightroom and to a lesser extent with Handbrake, I was looking for an application already updated and compiled for Apple silicon that would take full advantage of the M1, and I remembered that in performance comparisons it is common to use 7-Zip compression as a test of real use . In macOS there is no official client, but Keka performs 7z compression and it is very well optimized, as we will see by the difference between the normal result and Rosetta’s.

Even with Rosetta, which now does have a considerable difference with the native application, the M1 manages to outperform the i7-9750H using all its cores , with a considerable margin, and that in this test the MacBook Air comes to reduce to reduce the speed clock to cool down. In addition, it must be remembered once again that the i7 exceeds 60 watts of consumption while the M1 when cooling is about 11.

In the lower part of the graph, the four Icestorm cores demonstrate once again that the values ​​obtained in Cinebench R23 using all the cores are related to reality, tying with the 13 “MacBook Pro from 2017. Of course, the latter uses a value close to 20 times more energy to perform this same task.

A very powerful integrated GPU that surprises as much as the Icestorm and Firestorm

The case with the M1 GPU repeats itself . We’ve known for a long time that Apple also leads here on iPhones and iPads, and while a dedicated graphics chip is missed, on a day-to-day basis, and outside of specific needs, the performance is exceptional.

For integrated graphics and not having memory allocated exclusively, the M1’s GPU is the best integrated on the market, a far cry from the rest. But as with the CPU cores, the thing does not stop there, because it manages to be much better than the integrated proposals of Intel and AMD , with also much lower consumption.

Compared to the dedicated 16 “MacBook Pros we’ve been able to test, the Radeon 5300M and 5500M, the M1’s GPU is surprisingly close in frames per second in the GFXBench gaming benchmark (Aztec Ruins High Offscreen 1440p). It’s a benchmark already adapted to Apple Silicon and that uses the Metal library on all models.

 

Although not listed in the table, the latest dedicated notebook solutions from Intel and AMD are a far cry from the M1 , with 44 and 30 fps respectively on the latest generation processors released on the market. It is an advantage that is not surprising, since the iPad Pro of 2018 achieved 50 fps and the iPhone 12 Pro 36 fps.

As in other tests, the M1 has remained practically cool when doing the GPU test, while the two 16 “MacBook Pros with Radeon 5300M and 5500M have even put the fans at 5,000 rpm.

As revealed by Mark Gurman, it is expected that there will be computers with Apple Silicon with 16 and 32 GPU cores, the first being the one that would be expected in a MacBook Pro 16 “base with Apple Silicon. Seeing all the thermal space that Apple has to grow From 10.2 watts on the M1 on the MacBook Air to 61 watts on the 16 “MacBook Pro with the Radeon 5500M, you can expect the M1’s graphics performance to double . This has only started.

The Neural Engine in all its glory with Pixelmator Pro: what you don’t see in a classic benchmark

One of the most interesting sections of the M1 is that not everything is CPU or GPU power. As my colleague Javier Pastor said in his anatomy of the M1, in the context of heterogeneous computing, “Apple’s chips have a good number of specialized chips with cores dedicated to very specific tasks.”

Among all of them, the Apple Neural Engine or 16-core Neural Engine stands out , “capable of performing 11 billion operations per second.” With it, the system is able to perform specific tasks using machine learning much faster than using the GPU or CPU, with much lower power consumption.

At the moment, it is not easy to find dedicated applications that already make use of the Apple Neural Engine , but the future is very promising. Adobe has already launched some applications compatible with Apple silicon in beta, and thus shows the advantage that the M1 has with its Neural Engine over the 16 “MacBook Pro. As we can see, it does scene detection in four times less time.

However, thanks to Pixelmator Pro we have been able to do our own tests with the “ML Super Resolution” function, which uses machine learning to triple the resolution of an image “without losing quality”. In a first test with very small images, the M1 is twice as fast as the MBP 16 “2020 entry i7. However, we wanted to do the test with a large image (8000 × 5806 pixels), to see how it works. The different teams performed. These are the results:

 

The M1 wins again with solvency thanks to the fact that it is only using the Neural Engine for a task in which the MacBook Pro 16 “uses GPU and CPU, reaching 33 watts. It is 26 seconds slower, using 12 times more power consumption In smaller images, such as a cropped screenshot, the M1 was up to twice as fast as the i7, and this is just the beginning, waiting to see what the developers are able to squeeze out of it.

Maximum consumptions of the M1 on the MacBook Air

Before Apple announced the landing of Apple Silicon, there was much talk about the extent to which the company would be able to outperform Intel or AMD with chips “like those of an iPhone.” The reality is that the M1 is more than that, and thanks to an improved cooling capacity compared to mobiles and iPads, Apple allows the consumption of the chip with saturated large cores (something we have done with Prime95) to skyrocket up to just over 18 watts with the high-performance cores running at 3.2 GHz .

Image of the high-performance quad-core cluster of the M1 (P = Performance) performing almost 100% of its capacity. 18,384 mW = 18.38 watts of consumption.

Of course, that consumption and clock speed drops quickly (in 30 seconds, approximately) in the MacBook Air due to the lack of fan, and stabilizes at about nine watts in sustained performance. This drop in performance is not noticeable in the day-to-day, where as we have already said, high-performance cores are rarely used. Still, it should be noted that on the Mac mini and MacBook Pro, both equipped with fans, this performance loss is non-existent .

For Icestorm cores, with high-performance Firestorms disabled, the maximum efficiency cluster consumption is 1.7 watts with all cores running at their maximum 2.06 GHz . Again, it is impressive that with such low peak power consumption, these cores run the rig so well without help from their older brothers.

 

Image of the high-efficiency quad-core cluster of the M1 (E = Efficiency) yielding almost 100% of its capacity. 1,777 mW = 1.77 watts of consumption.

Regarding the GPU, the maximum consumption observed doing the GFXBench benchmark is 8.5 watts in the MacBook Air with seven cores in the integrated GPU. In the case of the Mac mini M1, connected to the current, it amounts to 10 watts with the eight-core GPU.

Regarding the maximum consumption of the last component of the M1 that the system reports, the Apple Neural Engine or neural engine, the maximum consumption that we have managed to obtain with Pixelmator Pro is 2.2 watts , a figure also spectacular considering everything that is squeezed. The Neural Engine is also activated every time we use the camera, in those cases showing a consumption of 0.12 watts.

Lastly, I set out to fully saturate all parts of the chip at once, with some success. In a very limited time, I got a consumption of 17.36 watts of CPU, 8 watts of GPU and 1.03 watts of RAM. Altogether, the “pack” put out a figure of 27.2 watts . Perhaps, if I had managed to use the Neural Engine at the same time, it would have touched 30 watts.

 

Notes on energy and possibilities

 The autonomy of the MacBook Air M1 is really surprising.

With all that we have talked about the energy consumption of the M1 and its high-efficiency cores, you wonder what possibilities this gives for other future equipment . The first thing to say is that I also do not see that launching a device that only had high-efficiency cores changes Apple’s game much.

It could launch a much cheaper chip, and even equip it only with a CPU with eight high-efficiency cores , which, if it grows in performance like all four, could be almost compared to the best that Intel has with the eleventh generation. Of course, single-core performance is very important to Apple to keep it as “low” as the Icestorm offer. Probably, without an architecture change, improving it would be feasible by raising the frequencies of the high-efficiency cores somewhat, which is still low.

At the level of possibilities, something with that consumption and heating could perfectly power a Mac that was the size of an Amazon Fire Stick 4K , and that, although it did not have many ports, would fully fit with the company’s strategy since the adoption of USB- C / Thunderbolt 3.

 Apple has in its hands the possibility of getting, for the first time, a device like the Intel Compute Stick, but with enough power for most tasks.

Moving on to more realistic scenarios, using only high-efficiency cores does not make sense when you see how the system manages to govern the two groups of Icestorm and Firestorm. The use made of the high-efficiency ones is so good that by testing video playback for one hour on Twitch with fixed brightness I have obtained similar results with four cores and with eight . The same goes for video calling in Zoom. This is due to what I said that high-performance cores only intervene when they are needed, such as to open an application at maximum speed. In the mentioned scenarios they are not used too much.

While watching a direct from Twitch a 2020 13 “MacBook Pro consumes about 4 watts, a MacBook Air M1 consumes less than medium, and with much less power fluctuation. Therein lies the real savings.

However, high-efficiency cores are going to play a major role in making future computers with Apple silicon, such as the successor to the 16 “MacBook Pro, have enviable autonomy. With the same use as a MacBook Air, and growing somewhat the consumption due to having a larger screen (which seems to be more efficient soon thanks to the use of miniLEDs), it does not matter if Apple gives them four or twelve more high-performance cores: the Icestorm will continue to be tremendously capable and fulfill the same role .

Thus, for light use, it will not be strange to see in a 16 “MacBook Pro more 30 hours of autonomy in video playback , for example, compared to the 18 that Apple talks about in the Air. It is not unreasonable considering that they maintain chassis and current battery capacity, set at 99.8 watt-hours (down from 49.9 watt-hours for the MacBook Air M1).

Although the M1 is more efficient than contemporary Intel, when it is pushed to the maximum it will drain the battery at a good speed: there is no revolution

Of all the news, having the Icestorm seems to me the most game-changing . Because something must be clear: if a MacBook Pro 13 “with M1 is required for a long time with all its cores at 100%, that is, consuming about 17-18 watts, the battery will drain quickly. In that scenario, The autonomy results will not be so different from those of a low consumption Intel i5, although it will achieve a lot of performance, but it is in light tasks where the real gain is that allows us to leave the house without scares.

A bright future ahead after dispelling many doubts

A much larger battery than that of 13 “computers and greater dissipation capacity can give the 16” MacBook Pro a range and performance never before seen in notebooks.

After having carried out many tests on the M1 and some of its predecessors and direct competitors in the range that Apple has for sale, what has started the M1 seems like a revolution to me. One that, because it does not change the industry too much in its early days, because it does not modify the use on a day-to-day basis, perhaps goes a bit unnoticed . But it will not happen for those who see that the autonomy of their laptop is tripled, that the fans no longer sound, that many specific tasks are executed with a speed never before seen on the desktop, etc.

In any case, after verifying that the promises of the M1 are real, we can only wait for their older generation brothers . In that sense, Mark Gurman is usually a reliable source, and on what will come after the M1 has offered information with some interesting details.

In early December, he told Bloomberg that Apple is developing Mac processors with up to 32 CPU cores. For the next generation of MacBook Pro and iMac, he spoke of Apple working on chips with up to 16 high-performance CPU cores and 4 high-efficiency cores (16 + 4), although he also pointed out that they might first opt ​​to launch 8+. 4 or 12 + 4, depending on the production.

Taking into account that in the M1 I have been able to deactivate the cores one by one from the fifth (the first high-performance) to observe how it behaved at a growth level, and that its performance and consumption grows in a practically linear way (about 1400 points with each nucleus added), I have taken the license to make this estimate from the eighth nucleus, always maintaining the base provided by the four high-efficiency Icestorms. Obviously, it is not a prediction , but it can give to understand where the M1X or M2 can move, whatever their name.

The image itself won’t tell us much, but if the estimate were correct, with 8 + 4 cores, the M1X would outperform the Ryzen 5600X by about 2,000 points, which consumes about 61 watts with all cores in use. The M1X would do it with 29.3 watts, less than half. If Apple launches a model with 12 + 4 cores, the estimate says it would reach 18,774 points with 43.3 watts, beating the Ryzen 5800X by about 2,000 points and being on the same level as the Ryzen 3900X. Finally, with 16 + 4 cores, the estimation model returns 24,414 points with 57 watts, catching up with giants like the Ryzen 3950X and 5900X (94 watts with all cores active). In favor of AMD, we must be fair and mention that Apple achieves its current consumption with lithography of 5 nanometers , versus the 7 nanometers of the Ryzen 5000.

Right now, considering how much progress the Ryzen 5000s and the Zen architecture itself have made, it seems like sheer madness. But it also seemed to many that the M1 could outperform Intel’s low-power chips in CPU and GPU (and let’s not talk about outperforming chips like the i7-9750H with which we have compared so much), and it succeeded, by far. margin. As my colleague Javipas said, the M1 was the great no surprise of 2020 for those who had been paying attention to the progression of Apple’s chips on iPad and iPhone.

 This chart was quite a statement of intent from Apple, and considering that the M1 is in the lower part of the blue zone, it is realistic to think that the estimates fit in the middle and upper part.

In that sense, the leaked Cinebench R23 tests of the development kit for the transition (DTK), which is still a Mac mini with the same CPU of the iPad Pro of 2018, have shown that two years ago, Apple was already at the height of what Intel offers in low-power laptops in the 11th generation when using all cores. And all this, with the limited dissipation that an iPad allows.

What the DTK & macOS 11 showed is that Apple probably could have shipped ARM Macs long before now — from beta one, macOS was so stable on ARM it felt like it had been doing this for years — but they would merely have been ‘as good’ as Intel Macs, not the blowaway experience M1 is

— Steve Troughton-Smith (@stroughtonsmith) January 16, 2021

As the well-known and influential developer Steve Troughton-Smith, who now has a Mac mini DTK, said, “what the DTK and macOS 11 have shown is that Apple could have released ARM Macs much earlier.” But, yes, “they would have been as good as the Intel Macs, not the incredible experience that is the M1.” And in squeezing the M1 to the fullest from the least powerful core to the most capable, I totally agree with that perception .

Source : Engadget