Hacker News new | ask | show | jobs
by marcan_42 1684 days ago
> Apple Silicon Macs are just one step away from being a completely closed system like the iPhone / iPad platform

Apple Silicon Macs are based on the iPhone / iPad platform. Apple chose to spend a significant amount of developer time adding the ability for users to securely load their own kernels, which is part of the new BootPolicy system that iDevices do not have, and a documented feature with multiple official tools to support it. There is a blog writen by Apple's head of XNU development detailing how to use it. If Apple wanted to lock these machines down they would've just not done any of that.

As for soldered RAM, you would need 8 RAM sticks in individual channels to match the M1 Max's memory bandwidth, at a much higher power consumption. Modular RAM is no longer viable for low-power, high-performance laptops. Modular, low power, high performance: pick two. It's just the way the physics works. Carrying a 512-bit bus across a connector isn't free, it has a significant power/performance cost due to increased capacitance and decreased signal integrity.

(Soldered SSDs, sure, that's a valid concern, but it has nothing to do with the OS.)

> Where as the reality is that unless Apple releases hardware documentation for it, all non-macOS operating systems on the M1 will always offer sub-par performance.

Funny enough, we already have better VM performance than macOS thanks to supporting the M1's vGIC (which macOS does not use yet), and we've also figured out how to work around a USB death issue that affects macOS, and I'm already putting making simultaneous DisplayPort 1.4 + USB3 work if at all possible on my TODO list, because I just found out macOS can't do it.

No documentation doesn't mean we can't beat Apple at their own game.

2 comments

> Modular RAM is no longer viable for low-power, high-performance laptops.

The fact that this is true makes me grumpy for all the obvious reasons, but even at my rather less informed level of understanding it's still obviously true.

My "solution" here is basically to enjoy the better batter life while grumbling quietly to myself ;)

Whatever you said doesn't at all change my assertion that the M1 is now just one step away from becoming a closed system - With the M1, Apple can now easily lock the bootloaders of M1 Macs any time and make it a completely closed platform like the iPhones / iPads. And it is evident that Apple has been planning this for years:

1. The first few Intel Mac Minis allowed you some level of customisation of both the hardware (change RAM or HDD / SSD) and software (install other full featured OS).

2. Then came the Mac Minis with soldered RAM and SSD. You could no longer customise the hardware. Software was still customisable and you could still install other OSes. (Recall that Apple even offered free drivers for another OS, i.e. Windows).

3. The current generation of M1 Mini now doesn't allow you to customise both the hardware (everything is soldered) and the software. Technically you can install other OSes, but the reality is that currently only crippled versions of Linux and xBSD is available and practically the only full-featured OS available for it is macOS.

These are clear indicators of how Apple has been working slowly to lockdown the Mac platform like their ios platforms. (The strategy to keep you in denial - https://en.wikipedia.org/wiki/Boiling_frog - has been working great for them). The reason for this is simple - BigTech are increasingly moving towards selling everything as a service. Services are more profitable because it means they create recurring income (even after the device is sold) which means more profits. and Apple's successful business model for this, that earns them billions of dollars, is the closed-platform model that is evident on iPhones / iPads.

> Modular RAM is no longer viable for low-power, high-performance laptops.

That is deliberately misleading and ignorant. Low Power DDR chips (that are soldered) were invented for mobile phones where every watt drawn matters. On laptops they offer only negligible power savings that doesn't really matter - yes, the power saving is negligible on a laptop as the difference in power used in modular RAM vs LPDDR is between 1 to 0.5 volt. That matters on phones because they have small batteries (5 - 15 Watt-hour), but that power saving is negligible on a laptop as it has bigger batteries (40 - 95 Watt-hour).

It's as good as meaningless when you consider what you lose - soldered ram can't be upgraded and is hard to repair. The loss of this big benefit - repairability - hugely trumps the minor power savings. Apple, and others, know this very well - for them, the real attraction to using soldered RAM is that it allows for planned obsolescence.

In fact, reasonable alternative ram chips like DDR3L (low voltage) and DDR3U (ultra low voltage) that offer lower power on modular RAM that were specifically created for servers and laptops, have been deliberately ignored by computer manufacturers, in favour of soldered LPDDR, and that is why the memory industry too is forced to do more R&D on them.

> the M1 is now just one step away from becoming a closed system

Just like almost every other computer. Any manufacturer can do this at will with new systems or a firmware update. You are making a strawman argument. The question is whether they will, and the answer is they seem to have no intention to do so given how much time they spent not doing it. Soldering things down has nothing to do with locking down firmware.

> install other full featured OS

All current M1 machines allow you to install any full featured OS. Apple just isn't writing the drivers for us. There is nothing locked down when you run your own OS. I know because I also run macOS kernels under my hypervisor that way and everything works exactly as it does when booting in Apple-signed mode.

Your claim seems to boil down to "Apple changed their hardware in a way that Linux doesn't support". Well, duh. That's what we're fixing. Nothing about that means they're locking anything down.

> That is deliberately misleading and ignorant.

You didn't even bother to read my comment. The M1 Max has a 512-bit RAM bus. That is equal to 8 RAM sticks (DIMMs are 64 bits). Do you want a laptop with 8 RAM sticks?

Each one of those 4 RAM chips on the M1 Max has 8 RAM dies internally, each handling 16 bits as independent channels, for a total of 128 bits per chip. Typical DDR RAM chips as used in DIMMs are 8 bits each (which is why you get 8 per DIMM to make 64 bits). You would need 64 such chips to reach the same bus width (and thus similar bandwidth).

You just can't make that modular short of putting your RAM on a thousand-pin LGA package like CPUs themselves, and that'd still increase power consumption (and significantly increase physical size, which again makes the power consumption problem worse as it makes your interconnect longer). It just can't work with DIMMs and with standard non-LP DDR technology.

> On laptops they offer only negligible power savings that doesn't really matter.

Yeah, until you make a laptop that is actually power efficient like this one, and then suddenly it does. These things draw milliwatts when idle and have aggressive RAM power saving that puts channels into low power modes after mere microseconds (I know because I've been investigating the memory controller power management config and benchmarking memory accesses). That's part of how they get amazing battery life. You can't do that with regular DDR RAM. These things will run for a week with light usage, and they can do that because they are based on extremely efficient mobile architectures. Your phone lasts a day; make the battery larger without increasing idle power usage and that's how you get a week.

> the difference power used in modular RAM vs LPDDR is between 1 to 0.5 volt.

It is evident you have little hardware engineering experience if you think power is measured in volts. That's not how it works.

I would advise you to spend some time reading up on high-speed digital interfaces and learn about concepts such as ohm's law, IR loss, capacitance, impedance, eye diagrams, and insertion loss. LPDDR RAM is much lower power than regular RAM precisely because it can optimize for very short interconnects, which is why you won't find DIMMs of it. It's not just about the voltage.

Your whole tactic during this debate is to just throw lot of technical jargons to try and confuse the reader and to evade the actual fact - all the hardware changes Apple has made has also been done to create a closed system to support the software services.

All the techno-babble spouted by you, to make yourself appear more knowledgeable, and the cheap potshot on me (also an engineer), desperately tries to hide the fact that high performing modular RAM architecture already exist without all the modifications Apple has made to their hardware with the aim to deliberately create an un-repairable and closed system.

I have clearly pointed out how Apple has been converting the macOS into a closed system, on both the hardware and software front, over the years. This is an undisputable fact as the only thing that now distinguishes the M1 Macs and the iPhone / iPad platform is the bootloader.

And looking at Apple's business model, it is only logical that Apple will soon be locking the bootloader of the Macs too, once it reaches a critical level of acceptance (and we are far off from that, for now).

The M1 is undoubtedly a great piece of hardware - but it has been deliberately designed with built-in planned obsolescence. That's good for Apple's profit margin. But not for us consumers.

Ah yes, "technobabble".

I'll go back to porting Linux to this thing then; clearly you're not interested in hearing about how these things are designed.

It doesn't take being a senior hardware engineer to be able to count bus width and calculate bandwidth, though, so I'm still curious how you want a laptop built with the 8 DIMM slots it'd take to match the memory bandwidth of the M1 Max. Remember, these things have high performance integrated GPUs, that perform at the same level as discrete ones. Ever wonder why discrete GPUs haven't had modular RAM for a couple decades now? Yeah. Bus width.

> high performing modular RAM architecture already exist

The M1 Max has a memory bandwidth of 409 GB/sec. A top spec EPYC server (Rome) chip has 410 GB/sec of memory bandwidth, with 8 channels populated with the fastest RAM they'll take.

Indeed, high performing modular RAM architectures do exist. In servers that eat huge amounts of power and require 8 DIMMs to go that fast. Good luck fitting that into a laptop.

But I guess this is all still just technobabble to you :-)