Apples SOC is one of the most open of the ARM chips out there, you might think it is a closed platform but that is in fact quite wrong. Things like once the boot manager hands over to your kernel only your kernel code is running there is no device manager engine like on intel/amd systems with full access to the system or a ring level below your kernel running the UEFI etc.

Due to the sheer marketshare and people following the effort. Mainline Linux for Snapdragon devices is there for some devices like the 835 and very smooth sailing as I've found out. And Alwinner devices support is really good under Linux Sunxi, except for the propietary things like video acceleration, however they are mostly on cheap chinese devices and enthusiast single board computers so they are a way less sexy target

@@hishnash never heard of tee apparently.

@@hishnashtor browser

Uplink

Great story and Archimedes holds a special place in my heart!

​@@DavidAspden I did a bit of excavation and although the ARM port first booted in 1994, it didn't start getting merged into mainline Linux in earnest until Jan 1998 with Linux 2.1.80 (when arch/arm appeared in the tree), although some preparatory patches had already been merged before then.

@@Tim_Small Can you give a link to "The History of ARM Linux"? Looking it up isn't quite clear what source your referencing.

I truly love it, thanks for sharing ❤

@Daniel Rumbacher should be a lot simpler for people to get HW accretion support on apple silicon for these things due to the ability to lift macOS up into M1N1 (vm like layer) and track exactly how it talks to any other bit of the system without macOS bothering/detective it is being snooped on.

@Daniel Rumbacher The solution Asahi Linux is taking is to not attempt to replace the proprietary firmware. inserted they pull this firmware during install time: how boot works means that when the pre-boot hands over to your kernel literally nothing is started up apart from a single e-core for your kernel. It is your responsibility to start up any part of the SOC you like and provide it with the firmware blob (this is great as you can pin the firmware version to your kernel even if someone has macOS installed using more uptodate firmwares this does not effect Linus at all). Then when up update your linux install to a version that wants newer firmware it can pull those blobs it wants. This is a key features of these chips! Apple provide fully legal way to pull down the firmware during first install. Apple really does not block free software, they are the main developers behind massive open source projects like LLVM. There have also made key changes to pre-boot and boot loader signing that have no use case for apple but made things much simpler and easier for the linux on M1 team. It is clear apple have no issue with this project and most likly like it quite a lot to put engineering time into making things better, they also keep in putting useful negates of info into the Darwin kernel (the open source part of macOS) that have helps the devs on this platform. The image that apple Is apposed to open source is just utterly wrong, vast amounts of the work apple does depends on it and they contribute back to the community a lot.

@Daniel Rumbacher , @matthaus woolard: I agree with you that hardware acceleration for graphics and videos is essential for these applications. There are no doubts in my mind about that. However, manufacturers secrecy about their drivers and other factors are definitely hindering the needed advances that we are all craving as a community. A potential solution to this problem is offering a huge number of specialized cores in the processor itself. This is a kind of a trend nowadays, especially in AMD's APUs. GPUs embedded in APUs would not be exactly the same as specialized cores, though. They could be directly accessed by a program without using shaders or drivers. High level programming languages are not quite prepared for this, except one: Go. Go has an intrinsic generic assembly that can be further compiled for efficiency in a given architecture. So instead of a low level driver, the programmer would have a library that would use the resource if it exists. It would be desirable to have a hardware scheduler to distribute the tasks back and forth the given specialized cores. This would be potentially highly efficient and would eliminate the need of graphics cards for most applications requiring them. The big problem with graphics cards is their high energy consumption and the communication bottleneck by needing to transfer the whole contents through an external bus. With specialized cores one would eliminate the need of graphics cards internal memory, communication (bus) bottlenecks, drivers and shaders, transferring these problems to a highly parallel and highly efficient managing hardware to transfer the tasks back and forth the specialized cores. Another advantage of specialized cores is that they could be conceived in a standardized way. Thus, future versions could be merely extensions (more cores, for examples). As I said, APUs are already slowly going in that way. It would be a great thing if ARM (and other processors, such as RISC-V) started to do the same.

With the quality of code by amazing people like Asahi Lina, I think it will not take long

My bet is that it will take A LOOONG time for them to reverse-engineer FULL hardware support for M2 into Apple's secretive proprietary world. You think Apple will allow them to have Linux running better on a Mac than MacOX?

@@hstrinzel it is no more secretive than any other platform to be fair, and given some of the hardware features such as being able to lift macOS fully into a very light weight VM (M1N1) they are able to sniff what it is doing very easily. This is a much simpler project than trying to reverse engineers most other hardware out there. The other benefit they have is that each of the added SOC features tends to be connected to its own little arm core that they can load the firmware (apple provide at install time). This exposes a high level interface to the CPU. So they do not need to write a low level display controler firmware that talks DisplayPort and handles display port handshakes as that is all done by the display controler cpu and firmware provided by apple all they need to do is see what remove function calls macSO sends to it and gets back from it and implement that interface in linux. (more or less a header file and some of that is even in the open source Darwin kernel that they can read over to help them figure these things out).

what would give better fps on linux steam with x86 games running with proton. an m1 macbook air with asahi, or with an amd ryzen 7 3700u or 5700u? the example game i like to inquire for is no mans sky on playable settings at 900p. but if you have any idea of how those two hardware compare in general for gaming that would appreciated.

@@jangofet555 Currently Asahi linux doe snot have GPU drivers and even when it does the GPU drivers it will get first are OpenGL once as this is much easier to developer. There is a team working on VK drivers but this will be be the TBDR subset of VK (not the set of VK apis that PC games expect and not what proton expects). Many people misunderstand VK it is not a single api it is more of a collection of seperate apis and depending not the hardware you have you implement the selection of features that match that hardware (you do not attempt to emulate other hardware that is againts the concept of VK). Apples GPUs are close to those by PowerVK and are TBDR gpus this means they have quite a different VK feature set to the TBIR gpus made by AMD/NVidia and Intel so PC games (and proton) that are coded for VK are in fact not coded for generic VK they are coded for TBIR PC VK subset of apis. In summary playing an un-modfied PC game on these machines will not run very well as if someone puts a shim in to support the cpus these games expect to be present it will be a shim with a load of nasty overhead.

Probably not from apple lol, that would be a conflict of interest, but surely from other projects.......surely........hopefully.......probably........maybe

@@fuseteam I checked out the progress on the GPU driver and the creator estimates they will have a working driver capable of rendering a desktop environment by the end of the year. Pretty incredible progress and the whole reverse engineering journey is extremely fascinating - worth checking out if you're curious about systems programming

@@DavidAlsh cool that's quite impressive indeed, all without apple's help ;) I'm definitely following the news of where this goes

@@DavidAlsh And depending on the hardware quirks getting the GPU drivers stable may or may not take a long time. For example, the GPU driver for intel GPUs is still a bit unstable even after over a decade of work and some level of support by intel engineers working on it. The issue seems to be that the intel GPU locks for some input and the driver needs to do all kind of tricks to avoid the hardware from locking up. If Apple GPU has similar problems, creating a working stable driver will be an enormous effort. If Apple hardware is actually good, then creating reverse engineered GPU driver will be much easier.

@@MikkoRantalainen that's really great insight. I'm not knowledgeable in driver development or reverse engineering hardware to create drivers so that really helped add to my understanding of the challenges

Out of curiosity what brand(s) are you implying that make genuinely usable laptops that don't drop support of hardware after a decade?

Long story short we users, customers definitely want these computing devices maintained, patched as long as possible independently of any vendor. At least, I think that was the point that the prior post was to make.

@@StormBurnX most laptop brands drop support for (at the OEM level) very quickly after release, they just depend on MS supporting things but stuff like firmware updates typically last for less than a few years if that. Based on apple's history with macOS we can easily expect 10 years of security updates for these devices (firmware and os).. might not get new os features for all of those 10 years but sec updates is what matters.

@@hishnash can confirm from personal experience getting various updates 7+ years after purchasing a macbook air over a decade ago, which is nice, but OP seems to complain it's not enough which is why I was curious what "better" brand they were trying to imply exists, otherwise they're just complaining for the sake of complaining it seems

@@StormBurnX the real nice factor is getting firmware updates as well not just os updates. So many OEMs just do not bother at all in the laptop (and desktop space).

for the performance in the thing an light long battery portable category is ia very competitive!

That is a very good detail to point out. Agree. Thank you. It's better priced in fact for better hardware. Performance comparisons would be interesting to see. And Virtual Win performance too.

1k? You can but that thing for around $700 when discounted(Amazon etc). You can buy the newer M2 for around $1k on Amazon discount.

@@bdkamil95 Good info to share, thank you.

u can run Windows x86 in a VM?

@@1MinuteFlipDoc running x86 windows is costly as translating the kernel live has a big perf impact. Better to run ARM64 windows with x86 translation embedded (but MS translation is not perfect). Apple this year did expose their Rosseta2 to linux VM guests (and based on the license could also be used by Asahi linux when running on apple hardware).

Same story here - I got all of my M1 units at launch and they are stupendous (for what they are). I'm writing this on an M1 MBA and it's stupid good.

Reasons for switching?

@@1MinuteFlipDoc I have the pro expensive annual Parallels subscription running on the maxed out M1 pro never in my life I struggled this much with a piece of software I love it yet hate it in the same time I could rant for hours about my experience (I mainly do Software engineering and Penetration testing/Red Teaming) in my device also had to do some Data Science (worked phenomenally) during my Software engineering work I struggled a lot cuz I had to go thru some hardware specific code (no emulation will help you there as of today for variety of reasons) also, had to do some Ruby (with Jykell and that was such a pain in the ass that's uncalled for) but the thing I had to struggle with the most is working with certain tool for my job and while I was pentesting a target I had to install certain lib problem is many of tools in the industry are not compiled with ARM on mind that was a struggle as I come from a Software engineering (worked as a Full stack dev and Data science stuff) it's easier for me than the majority of people in the tech industry to modify code on the fly without breaking their flow while they are analyzing a target so I would not recommend M1 for any pentesters, except the main reason why I bought them: Mobile Penetration Testing And bug bounty hunting .. if that's something you don't do at the moment skip the whole M series heck all Macbooks

@@TFirsty I don't disagree, but I lament the fact that our computers are at best turning into sealed devices akin to gaming consoles, and at worst into engineering duplicates of cellphones. Considering what the NSA has done with cellphone technology in placing the entire population under surveillance, this does not bode well for anyone interested in using technological tools to promote critical thought or dissent. This is especially true for journalists and human rights activists around the world. I find all of this quite alarming, to be honest.

@@TFirsty This isn't really true at all. SoCs are good for power efficiency reasons but they don't really contribute to performance at all. Why do you think AMD use chiplets for both CPUs and GPUs? Because they give higher performance and are cheaper and easier to manufacture. Even Apple use two processor chips to make the M1 Ultra. There is actually a type of processor called a SoP (System on Package). All the Ryzen 7000 are SoPs as they have a chipset, GPU, CPU, and memory controller on the same package; this design type if compatible with chiplets as some components are in a separate chiplet but they all wire into the same substrate or package. M1 Ultra is also a SoP as are all M1 products. As for SSDs the actual storage chips that hold the data can't be part of the SoC anyway, as seen in the Apple devices. Only the controller is part of the SoC and it supports multiple sizes of storage, Apple just choose to restrict what you can do with it because it benefits them to make devices non-upgradable. With some firmware updates they could make the M1 Ultra support upgradable storage modules quite easily, although these upgrade modules would have to be supported by the firmware and would be proprietary. It would be almost trivial for Apple to redesign the Macbook Pros using the same up-gradable modules instead of soldered on chips. SoCs are also great for low power and cost devices like smartphones where performance isn't required. Even then you can make variants with up-gradable storage and sometimes RAM as well.

@@TFirsty Your literally talking about Intel as an example of performance coming with efficiency when they are the least efficient compared to AMD, Apple, and Qualcomm. Sure efficiency and performance can go together but it's really not that simple. Just look at RTX 3000 and RTX 4000 for an example. Nobody actually puts their main DDR memory on the same chip as their processing in high performance systems. Even Apple M1 dosen't do this. It's on the same package not in the same chip, so you have a misunderstanding of what an SoC is here compared to an SoP. I actually agree that putting memory on the same package can improve performance but I doubt it will be the only viable solution for quite a while. However the only things which substantially benefit from this approach are devices which use HBM such as GPUs and FPGAs as they actually make use of the better connection between processor and memory that MCM allows. This is partly because the pace of modern processors outstripped RAM decades ago. That's why we have cache. Even putting RAM on the same package isn't fast enough to remove the need for cache and you can see this with M1 itself. Anyway all that aside even with SoPs that have integrated RAM you could always put the SoP on a separate processor card or mainboard that can be upgraded independently of storage and other hardware like IO ports, keyboard and screens. You can actually see this with the MNT Reform - although that design is far from ideal. The framework also has a swappable motherboard allowing upgrade to 12th gen processors from the original 11th gen without needing to replace the other parts. The only sustainable computer industry is one that promotes reuse of existing laptops and other products with new internals.

@@TFirsty What Intel chip are you talking about? Alder lake runs stupid hot so it's obviously not that.

Silicon-based computing is dangerously anticompetitive - I want my computers to be built wholly from individual battery-powered transistors, so I can tweak each one when it fails.

They use same architecture like vulkan or opengl

Jeff Geerling has been covering PCIe GPU support on ARM for a number of years now. Most SBCs do not have a great PCIe implementation and need to be redone in order for GPUs to work out of the box - in the cases where it works (with great support), it is enterprise-only and priced well out of reach for the ordinary consumer.

Look into Nvidia's ARM SBCs. It's not quite what you're looking for (yet) but they have Nvidia GPU cores on die with ARM CPU cores (essentially an ARM APU), which seems to be the most practical way to get traditionally usable graphics on an ARM system at the moment. If you must use a PCIe add-in card I believe I've seen people try to do it with the Pine64 Rockpro64 if I'm not mistaken, but I believe it was a massive effort with limited success.

FYI we are now up to the M2, here is my video about it ukposts.info/have/v-deo/r3R1gHprbn1eqok.html

The M series architecture is not a derivation of the A series chipsets. It’s from a completely different ARM architecture that Steve Jobs purchased the rights to before he passed away. This is Job’s masterpiece. Not something he paid TI for. Hence he hired a bunch of ARM systems engineers and TI guys and had Apple build a plant in Texas near TI so they could scalp TI engineers. This is Steve still showing the tech industry who the boss is, even from the grave.

The thing with arm and part of the revolutiomn it brings, is that there _is_ no standalone cpu. The M1 tauted by linus is an SoC, all pf them are. It's a 'system on a chip' it contains, in the case of the m1; the cpu, the gpu and the ram on a single chip which improves its performances drastically but an SoC can also contain other processing units like a video processing unit, an image processing unit, a neural processing unit. and it still requires a motherboard. Interesting enough apple's m1 even contains an ssd controller so that raw memory modules are used as a storage device SoC _could_ potentially be expandable with extra memory, 'external' gpus and what not but that remains to be seen. But basically wat we're asking is to renew the computer build to be more modern with ARM SoC's rather than the legacy x86 CPUs, which has power(both computational and battery) benefits, see the M1 ;)

A love how people comment (in generally, not only in the tech space) how "we need" and "we want" something just because they are just want Marketing and production works in a specific way. The things that are created are the things that "we need" and "we want". The truth is that most people don't give a single shit about if their devices will user ARM, X86, RISCV, MIPS or whatever! Hell, most people don't even know what these stuff is. Most people just understand and care about performance, battery life, features and cost. Until ARM (or anyone else) can offer something that will set it apart, the average consumer won't care. And unfortunately, unless it can excel in 2-3 sectors, if it's just 1 sector, that 1 sector needs to be performance because the average normie user just gets "craaaazyyy" about performance. I don't say that to say something bad about you or to offend you. I just wanted to point out that realization that I see more and more often as time goes. And again, it's a phenomenon that happens in every space. We humans (and I'm putting myself inside, so you don't call me a hypocrite) do so much self-projecting without even realize it, lol!

Starting at $1,301.40 😏

​@@GaryExplains I got my MBA M1 for $899 1,5 years ago. The price is the critical point here. It's what makes the base MBA M1 amazing and the MBA M2 a bit disappointing.

If Linus can endorse framework to work on this side. Framework will be a boom.

Apple deliberately designed it's ARM systems so you could boot whatever OS you'd like on them, they just fon't provide assistance on you doing it, you congratulate the Asahi Team but not enough to pay attention to what they themselves say on the matter, it's not closed hardware.

Apples SoC is one of the most open ARM chips out there. If they were anti open source (Like MS, Intel and AMD) they wouldn’t do this at all

Apple is just locking down iPads and iphones

Yes, but the Cortex A76 is quite old now, it was used in flagship smartphones in 2019.

The difference to Asahi: Apple does not lock down the hardware cryptographically (like Nvidia does). This theoretically allows easy reverse engineering given sufficient time (a few years)

Nice comment but I dont think Apple cares about end of life of its products. I suppose Apple doesn’t want publish a part of their technologies to others

Amdgpu drivers are all open source 🤨

@@TheFredFred33 When your a company the size of apple you end up cross licensing IP etc from many vendors. It would be a legal rats nest for them to just open source everything even if they wanted to. They do open source quite a bit but are very intentional about that (kernel is open source of source, not to mention all the work they do on LLVM).

I have heard of it but never seen one in real life. From the specs the CPU performance is near (worse) than a Raspberry Pi 4, so not really what we need.

so if your very skilled at soldering you *can* upgrade it people have upgraded by SSD and RAM on these devices but these tend to be shops that already are used to doing this to phones.

The processors in Chromebooks tend to be quite low end.

Would that apply to system76 machines?

Why do you want RISC-V laptops? I don't think there is a mechanism to verify what is on a chip. A chip maker can say that it is using certain source files, but then use something else. There is no checksum that the hardware can perform to verify it is the same RTL (or whatever) as the source.

@@GaryExplains Guess we'll have to build our own CPUs, might even start sourcing our own silicon

@@GaryExplains RISC V is more open than Intel or ARM - in fact there are designs that come with the CHISEL source - and you can verify those with a checksum. You can get those fabbed right now. The argument here is that because you can't be 100% certain all the way down at this moment, there's no point in bothering moving towards more open systems. The CPU is a critical layer that needs opening up.

Only the RISC-V specification is open source. Not (all) RISC-V processors. You can't verify what comes from the fab with the checksum of the source.

Eh?

Power PC only "took off" on the desktop because Apple picked it... See any similarities?

You mean the $1300 laptop?

Old processor with only Cortex-A53/A72. Barely faster than a Raspberry Pi.

Why? What advantages will it bring?

Just install Linux on your LG Gram then? Or dual boot?

You mean that laptop that is barely faster than a Raspberry Pi 4?

will be difficult for Qualcomm to functionality use the Nuvia team as they need to be very careful to not do anything that is to close to what apple does/had internally as it would then be way to easy for apple to get it blocked. That team will be required to document their working and start from first principles to ensure they have a paper trail to prove it is not stolen tec.

LOL, it isn't the fact that the CPU is part of an SoC that makes it efficient. It is the CPU design that makes it efficient. Also, Arm's server CPUs in its Neoverse range are also more efficient than the Intel and Amd counterparts that is why Google, Amazon, and Microsoft now offer them in their cloud services.

They don't have to prevent it, they just have to NOT open their system and let them properly support it. Don't get up in the few normies. The majority of people will not buy a laptop to run an OS that is not officially supported in it. If you do that, you are up *AT ANY TIME* for your PC to stop working, either at some part or completely. So yeah, Apple just doesn't care.

Thanks

This is the reason that Intel/AMD will probably remain the main processor for PCs going forward I think.

@@StaceyAyodele hard to stay on the top, when your x86 device use almost the same power just remain in idle and showing the screen saver, like the ARM based device on full load. It will limit your capacity in the future, you need sometimes 1000W power supply to reach that x86 performance, when Apple not far from that performance with a bit more than 100W. So what do you think who can double the performance, and power consumption more, before they will reach the limit?

@@randomcomment9992 ARM devices that manufacturers make currently are locked down and have no upgradability. I'm not spending $1000+ on a computer I can't upgrade down the line if I want to. For this reason, I don't see ARM computers being a good option for those of us that want upgradeable systems.

@@StaceyAyodele than you can use this forever, and never get a better and faster device in the future.. since they are already limited in a few country.. even the more powerful desktops are over 6-800W power consumption, they are limited to max 600W, over that they can't sell anything.. So bye-bye RTX 3090 Ti (5-600W power consumption possibility) + good CPU (2-300W power consumption possibility). You can't get both in California, and most country will follow this, when the 1000W common power supply will even raised to increase the performance to more..

@@randomcomment9992 3090ti is like 4k$ and youre worried about power consumption? why? you can easily afford electricity if youre spending money on a 3090ti.. come on

I think the ability to reach high performance at low TDP will have gamers interested. Look at how power-hungry the next gen video cards will be. Wouldn't it be nice to have a 90W CPU that performs like an x86 250W TDP CPU? It'll help with heat management and also spare high wattage requirements on your PSU. So... TL;DR I think gamers could be interested, if high-performance low-TDP chips were available.

@@vinnytube1001 you have a point. But, a RISC may not always beat a CISC. e.g. my raspberry pi 4 generate more heat than a Intel NUC i3.

In terms of performance that sounds about right. ARM is generally more efficient by not needing such a large decoder, but high performance CPUs translate to micro-ops anyway, so they're more similar than different. This should mean x86/64 would only be slightly behind ARM on the desktop. (I could be wrong, of course.) I'd still like to see ARM on desktops so developers can more easily support it. But I'd rather see RISC-V. Not for any particular technical reasons. I just find it appealing and I like what it stands for. If I make a homebrew CPU, emulator, or FPGA project using RISC-V (might do someday for fun), I'm protected against attack by lawyers. That's no small thing. They're the thugs that destroy projects, companies, and lives. They steal money, get people imprisoned, suppress innovation, and generally just create fear and misery. That's not a problem for enthusiasts of RISC-V. It's better with ARM these days because RISC-V forced them to be better, but they're still in it for the money and protected by law.

@@JB52520 That's a good point. ARM instruction set has grown to be more CISC-y, and x86 processors translate to an internal instruction set anyway. The cost of those translators is estimated to be 3-5% in modern CPUs. But... not only not paying that cost, but also freeing up die space for other things, is definitely an advantage.

"ARM instruction set has grown to be more CISC-y" could you clarify that? AArch64 is a clean, modern, efficient 64-bit ISA.

True. But the reason why the PC has been successful over the decades is its modular approach. We need that, not more custom built computers where nothing is upgradable.

@@GaryExplains i totally agree. Just pointing it out. I've been bothering my friend for the last decade about why phones don't come with sockets for CPUs, I'm sure he's tired of it (i know why, but still)

No, not relevant.

MARK!!!!

It is an expensive option.

@@GaryExplains You should do a follow-up to this video!

But the opposite is also true, you can't port the software until you have a platform to run it on.

@@GaryExplains Yeah, thats true, classic chicken-egg-problem. I think the best laptop manufacturer to push forward would be Microsoft themself, because they would not only be able to grep more attention than most laptop manufactures, but also if there is a problem with the emulation layer, they do not depend on another company to fix it. (And they already do build some ARM machines so this might help to move in the right direction eventually)

Why would you need RISC-V if you have Arm? What value does RISC-V offer the consumer? As you say it is "coming" it isn't here yet and it is still years behind Arm, Intel and AMD.

@@GaryExplains ironically there's already a roman brand laptop that uses RISC-V processor and the laptop listed for power users. Check it out if you want.

That isn't irony. I didn't say that RISC-V processors don't exist. I have a review of the VisionFive 2 SBC here on this channel. I said they are years behind. The performance of the VisionFive 2 is average at best, and the Roma laptop isn't for power users, that is a joke, what they mean is that it is for nerds as it is underpowered, overpriced, and you will be fighting every day to keep things working. You didn't answer my questions: 1) Why would you need RISC-V if you have Arm? 2) What value does RISC-V offer the consumer?