Apple M1 Chip Ultimate Review (History, Power & Future!)
So what i wanted to do for this video is review the history of the m1, including the original roots of this chip, and then i wanted to see how it currently stands in terms of value, efficiency and performance compared to when it first came out and then Finally, i want to take a little sneak peek into the future apple silicon mac chips like the m1x and the m2 starting off with the history of the m1 ship. I want to go back way back. I did a little digging and i found this very interesting article concerning a former senior apple engineer, who claimed that way back in the day, apple began planning its own alternative to intel. Processors apple started work on what would become the m1 10 years ago in 2010, when they contacted arm to develop a custom 64 bit instruction set architecture, and if you didnt already know, basically, every single smartphone and computer chip today uses a 64 bit isa, which is Why the m1 absolutely had to be a 64 bit chip in order to run the most common, modern apps, which are built to support 64 bit because 32 bit only supports a maximum of 4 gigabytes of ram, which is obviously too little for many apps. So apparently, apple approached arm even before they finished designing their own basic 32 bit chip and fun fact. In 2011, the 32 bit arm architecture was the most widely used architecture in mobile devices and then in 2013 this happened.
The 5s is the most forward. Thinking phone weve ever created, in fact, perhaps the most forward thinking phone anyone has ever made a7 is 64 bit. This is the first ever in a phone of any kind. I dont think the other guys are even talking about it. Yet now phil schillers words were true because a full year later, samsung revealed the galaxy note 4, which came packing the snapdragon 805 chip, which was still using a 32 bit architecture and less than a week after that apple revealed the iphone 6 with their a8 chip, Which was their second generation 64 bit chip, basically giving apple a huge head start in terms of 64 bit chip design and to this day apples a series chips are way ahead of anything qualcomm or samsung can design in single core multi, core and graphics performance, and, If you think about it, its all thanks to apples idea of making their own chips for the mac way back in 2010 and according to chakran apple, had specific plans for how their chip would work. So they asked arm to create the 64 bit isa, which became rm64. So what hes, basically trying to say, is that the arm 64 instruction set architecture that we know today is only here because of apple approaching arm over 10 years ago to create this specific 64 bit architecture and, if youre, having a hard time, believing that apple was Planning for their future m1 ship way back in 2010.
Let me take you through a list of changes that apple absolutely had to make in order to make the m1 max possible today, in 2014 apple introduced their very own metal, graphics, api for ios to replace the open standard, opengl api, which apple believed had way too Much overhead and then a year later, apple brought metal to the mac which improved rendering performance by up to 50 percent and it improved efficiency by up to 40, which are two extremely important aspects of the current m1 max, especially efficiency and then in 2017. Apple introduced metal 2, which further improved performance as well as machine learning performance, and they even announced metal, 4 vr, basically showing that apple wants to use metal for every single apple device going forward, including their rumored vr and ar headsets and then moving forward in 2018. Apple made the bombshell announcement that mac os mojave is deprecating opengl across all platforms and pushing developers towards using metal for better performance and efficiency. Now this really upset game developers because they didnt want to have to rewrite their games to support metal and with that came the announcement that mojave requires a graphics card that supports metal, and this ultimately meant that nvidia gpu support was gone and then in 2019 apple. Officially, dropped support for 32 bit apps with mac os catalina, which basically broke all 32 bit games and made a lot of people upset. But now we finally know that apple was making all of those frustrating moves because they were literally planning for their apple silicon m1.
Chip the entire time. So let me explain why the m1 chip has no support for 32 bit apps in the hardware at all, which allows it to focus on 64 bit apps, so they can run faster without all of the extra overhead needed to support 32 bit. Apps and because the m1 only supports 64 bit apple was able to create rosetta, which automatically translates x86 64 bit code to arm 64 making. Basically, every old mac program run perfectly fine without needing to be updated, with full apple silicon support, which is the number one reason why the m1 max became so successful and because apple was planning to use the same apple silicon architecture across all of their devices, including The mac, the iphone ipad, everything apple needed their own graphics, api that was as efficient as possible to improve performance and battery life, and that was, of course, metal, and only because of metal ipad and iphone apps automatically run on m1 max, which is a huge feature For apple metal is also what allows the m1 chip to have such impressive graphics performance, while barely sipping any power, just 5.6 watts at full load compared to 45 watts on the 16 inch macbook pros 5300m gpu, which is actually only slightly faster than the m1 gpu. So the point that im trying to make is that apple has been planning the m1 chip for an entire decade, with many steps made over the past few years to get developers to slowly support apple silicon on the mac without them even knowing about it.
But with that said before i explained how the m1 chip is so powerful, i want to take a look at how the m1 chip stands against the competing chips that are just coming out today and some that havent even been released. Yet when the m1 chip first came out, we compared it to the best consumer laptop ship out there, the amd 4900 hs and, to our surprise, the m1 smoked it in terms of single core performance being 42 percent faster, and it was just slightly behind in terms Of multi core performance in geekbench 5., however, in cinebench r23, the amd chip was 30 faster than the m1, but whats. Also important to note is that it was using 53 watts of power to get that score compared to only 15 watts on the m1 and surprisingly, the m1 was somehow faster in some tests, like lightroom, classic raw photo exporting and almost as fast for things. Like 4k video exporting and final cut pro because of optimization and that same optimization allowed the m1 macbook air for 999 outperform a 3899 16 inch macbook pro and tests like xcode programming and certain video editing tests like hevc stabilization and r5 10 bit 422 exporting because Of its modern hardware encoders and for other common tests, like 4k editing, it performed impressively close, considering the massive price difference and then recently we compared the m1 macbook air to laptops that are actually in its weight class like the lg gram 14, which is actually 200 More expensive and the m1 absolutely destroyed it in every single way, from the build quality and performance all the way to battery life.
We then compared it to both the intel and amd models on the new surface laptop 4 from a couple of months ago, and the m1 air was faster in basically every test, especially when you unplug the surface laptops from power, making them slow down even more and For real world editing work, the m1 absolutely killed them, especially for 4k video editing. We tried it again with the new galaxy book pro and once again we saw the same results with the m1, outperforming it in every single test, but now moving on to future chips. We just got some leaks of intels alder lake chips that are going to be competing with apples, m1 chip in premium thin and light laptops when they release in the first quarter of next year and in the leaked benchmark for the 14 core chip model. It scored only around 1250 points in single core and 6800 points in multi core, so its still going to be slower than apples, m1 chip, especially in terms of single core performance and then comparing amds latest zen, 3 5800u mobile laptop chip, which is meant to compete With the m1, the amd chip is consistently slower than the m1 in both single and multi core performance. So we can basically see that even nine months later, the m1 chip is still holding its own against competing chips in its market segment and, as we all know, the next version of the m1, which is the m1x, is coming out within the next couple of months And were expecting it to have insanely powerful performance being the fastest laptop ship in the world, with an estimated multi core score of around 14 350 points, but getting back to the m1 chip and why its so fast.
Let me lay out the basics before getting into some very interesting tidbits that give it a huge advantage. First off the chip itself is built on the most advanced chip node in the world right now: tsmcs 5 nanometer process allowing it to pack in 16 billion transistors. On top of that, everything from the cpu, the gpu, the 16 core neural engine, the thunderbolt controller, the machine, learning accelerators, the dedicated video encoding chips and the memory is packed into a single chip package, which greatly reduces the distance that data needs to travel between each Component, which reduces latency and improves performance. The m1 also uses a unified memory architecture, which means that all of those components, mainly the cpu and the gpu, can access data from the ram at the same time, without having to copy it back and forth to work on it. Like traditional x86 chips do going further each core within the 8 core gpu has its own dedicated tile memory and extra cash so that it doesnt have to access the ram as often which optimizes performance, but now getting into some of the more advanced details that most People dont know about this brings us to an article from eric enchine. According to him, the m1 chip has a total of eight decoders, which is unheard of since most of the best chips from amd and intel are limited to four of them. Decoders basically take machine code, instructions and chop them up into micro ops, which gets sent to the instruction buffer, where they wait for the chip to process them so having eight of them instead of four greatly improves performance, and on top of that, the instruction buffer on The m1 is three times larger than what is normal in the industry.
Now, the reason the regular x86 amd and intel chips dont have larger instruction. Buffers is because they only have up to four decoders, so they wont even be able to fill up a larger buffer. Anyway, compared to the m1s8 decoders that actually can fill up a three times larger buffer and the biggest issue for regular x86 chips is that they cant add more decoders than four, because the chip has to analyze each instruction. In order to see how long it is before using a decoder to chop it up into micro, ops, because x86 instructions can be anywhere from 1 to 15 bytes long and adding more encoders makes the whole process much more complicated. So amd themselves said that four decoders is an absolute limit for them, but for the m1 chip each arm instruction is always four bytes long, no matter what so it doesnt get as complicated, so apple can basically add as many decoders as the chip can handle, which Basically means that they can theoretically add even more in the future, while x86 chips are basically stuck with only four, and this is what allows apples m1 cores to essentially process twice as many instructions as amd and intel cpus, while at the same clock speed now it Doesnt stop there because there are even more huge advantages with the m1 chip which brings us to an article from andre at anantak. The m1 chip actually uses eight memory channels, which is super impressive, because the 10 gen intel chip and the high end 1800 macbook pro.
Only supports up to two channels and even the newer 11th gen intel i7 chip still only supports two channels, so having eight channels gives the m1 chip a super impressive, 68.25 gigabytes per second peak memory bandwidth, but making it even better. A single m1 high performance core achieves memory copies of 60 to 62 gigabytes per second, almost fully saturating. The memory controllers, thanks to the eight memory channels, which is apparently something that andre, has never seen in a chip design before all of this basically helped the m1 achieve the highest single core performance, seen in any chip higher than the flagship 5950x from amd and the High memory, bandwidth and unified memory architecture also helped the m1s eight core gpu be the fastest integrated gpu ever made and heres. Another fun fact from david smith. Retaining and releasing an ns object takes around 30 nanoseconds on a current gen intel chip and only around 6.5 nanoseconds on the m1 chip or 14 using rosetta 2 to emulate x86 intel code. And if you have no idea what this means. Glitch parrot explained that creating and destroying objects in memory, which is done all the time in modern programming, is much faster on m1 than it was on intel, and this is basically why our 900 m1 mac mini with only 16 gigs of ram, was able to outperform Our 15 000 mac pro with 192 gigs of ram in terms of xcode programming, with our max tech, xcode benchmark so yeah.
The m1 chip is incredibly impressive, so impressive that its actually getting windows focused youtubers to recommend the new m1 max. In short, this is the best computer ive ever owned shocking news. I know considering how many notebooks have gone through my hands, how many pcs ive built myself, but welcome to my long term review of the macbook air m1 base model and why? I think you should get one its weird right as a reviewer, its becoming harder and harder to recommend windows devices over what m1 macbooks offer, and this is only for this type of like an ultrabook right stuff for work or school or just like productivity based stuff. I guess what im saying is: even if you were to go out of your way to point out the flaws its hard not to recommend either of the m1 macbooks. They really are just that good and now let me finally finish off with what you should expect for. The future, of course, were expecting the new redesigned macbook pros to come most likely in october with the new m1x chip. So let me expect what to expect from that. The m1x will most likely use the same exact course as the m1 built on tsmcs five nanometer process, except that it will have double the performance, cores, half the efficiency cores and either double or quadruple. The number of gpu cores, basically making it the fastest ship in any laptop now the m2 chip is basically going to be the replacement for the m1 coming out sometime next year, using a similar chip layout, but instead using apples, newer, a15 cores built on tsmcs, updated.
Five nanometer plus process and mark german believes that it will remain with the same 8 core cpu layout, except that it would add two more gpu cores to boost graphics power a bit more, and i personally believe that the m2 will be built on the newer arm. V9 instruction set, which will come with a bunch of improvements like sve2, to improve performance even further and going deeper into the future. I expect there to be m2x, m3x and 4x chips and so on for the high performance consumer max, like the high end, macbook pros. As well as the m2, m3, m4 chips and so on for the lower end, macs like the macbook air and the smaller imac and the way i imagine apple, pulling that off is by slowly increasing core counts of the cpu and gpu, as well as continuing to Use better process notes from tsmc like four nanometer, three nanometer, two nanometer and so on and theyll probably rotate between using those two methods to improve performance between chip generations.