MacNN Forums (http://forums.macnn.com/)
-   Mac OS X (http://forums.macnn.com/mac-os-x/)
-   -   Running Linux natively damages Apple hardware (http://forums.macnn.com/90/mac-os-x/499182/running-linux-natively-damages-apple-hardware/)

 
abbaZaba Mar 25, 2013 03:17 PM
Running Linux natively damages Apple hardware
I had an interesting conversation with a guy who claimed to be a 30+ year accredited Apple technician. He claimed that running Linux natively on Apple's Intel hardware runs the risk of permanently damaging the CPU. His justification went like this:

"No Linux distro can control the advanced power management functions of Apple's custom SMC; there is no standard SMBus device in any Apple product. This means the OS must micromanage power flow.

Without that micromanagement, the CPU is left in a permanently over-volted state. Under load, this means delta-T goes positive and electromigration begins."

He went on to explain that electromigration is an irreversible degradation of the ALU matrix, caused by over-potential electrons to 'jump trace' through the insulator. If this happens too often at the same place, it short-circuits the logic. Also known as 'overclocking-cancer': a gradual, incremental, irreversible degradation of the silicon.

He also explains that Apple provided a fairly crippled SMC driver for Bootcamp, which is why Windows is the only other officially supported OS on their hardware.

He says Apple's reasoning for doing this is for battery battery life. The kernel controls CPU power; it is pre-emptive power control vs the traditional SMbus method which is reactive and wasteful of energy.

I cannot find any evidence to back any of this up with my Googling skills. Does this sound legitimate? On one hand I can see a bit of FUD in all this, but on the other hand it's not like Apple's a stranger to adjusting their hardware to suit their needs. His, what I may describe as tinfoil-hat-theory, says that the number of people running Linux natively as the boot OS on Apple hardware is so low, and that when the hardware does eventually fail, it's chalked up to sloppy Chinese manufacturing AND that Apple keeps this information very close to the chest.

I would love to hear some thoughts on this from the more hardware-minded people on here.

edit: just for clarity sake, these conversations all happened in ##apple on Freenode, which is why I say this guy "claims" to be a 30+ year Apple technician; I don't have any proof that he definitely knows what he is talking about. But in my conversations with him, he definitely pulled it off like he was quite the veteran.
 
P Mar 25, 2013 04:07 PM
The CPUs are permanently overvolted for better battery life? Really? No, that part doesn't make sense - increasing the voltage increases the power consumption radically. In fact, there is zero reason for Apple to overvolt anything by default, and I've never seen anything to indicate that they do. Apple probably does hide a lot of these interfaces, but why on earth would they overvolt anything by default?

Without the overvolting, the chips cannot die. Back in the olden days, CPUs could die from overheating (there are spectacular videos of AMD Athlons doing that if you look back 10 years or so) but that is no longer the case. Intel chips specifically will prevent it through a two-step "throttling" process - first inserting NOPs - "no operation" opcodes - into the instruction stream, and then underclocking. If that doesn't help - because say the cooling system has failed catastrophically - the chip will shut down.

Other than that bit, his explanation of overvolting is correct - overvolting CPUs is bad, and can damage CPUs. The exact process he describes is specific to the Intel Pentium 4 "Northwood" which would die in that manner, but other processors will have similar syndromes.
 
abbaZaba Mar 25, 2013 04:16 PM
he didn't say that Apple over-volted them by default, he was saying that without the kernel handling the power flow (their non-standard SMC), the CPU is then left in an over-volted state. Apple has their custom SMC handling it in OS X, and provides a "somewhat crippled" SMC driver for Bootcamp. Thus, since there isn't a driver (or kext?) for Linux, the CPU is then operating in an over-volted state. At least, that is how I understood it: the pre-emptive power management done by the kernel vs the reactive manage provided by the SMbus method is what provides better battery life and why the decision was made to handle it their way.

but I'm just the messenger trying to figure out if this actually makes sense.
 
BLAZE_MkIV Mar 25, 2013 04:40 PM
This would mean that any machine left in a kernel panicked state would die. Its more likely that without the special driver the system remains under volted. I doubt either one is true.
 
P Mar 25, 2013 05:09 PM
It doesn't have to be a kernel panic. Simply disconnect any storage from the Mac and turn it on. Since it can't find an OS to boot from, it would then automatically die eventually. Even better - during a firmware update, when there is absolutely nothing to keep it alive.

Anyone who has updated the firmware knows what happens - no fan control from the OS, so the fans run at max. Same thing if you leave it at a KP for long enough. Obviously this is a safety measure. Why would a company that does that leave the CPU in an overvolted state?
 
abbaZaba Mar 25, 2013 06:06 PM
So this is bunk and something to not even remotely worry when running Linux? Thanks for the replies. Guess I need to get my bullshit meter adjusted

Edit: another clarification. This wasn't something that would kill the CPU right away. More of a gradual damaging over time.
 
cgc Mar 25, 2013 06:34 PM
I gotta say this sounds like BS.

Does the EULA state you MUST run OSX or some other Apple-branded OS on Mac computers? What happens if you install BootCamp then forget to install the BootCamp drivers?
 
mattyb Mar 25, 2013 06:38 PM
How does Linux running on all those Intel CPUs in servers deal with this 'issue'? Surely the architecture isn't THAT different?
 
abbaZaba Mar 25, 2013 07:08 PM
Quote, Originally Posted by mattyb (Post 4223420)
How does Linux running on all those Intel CPUs in servers deal with this 'issue'? Surely the architecture isn't THAT different?
Well, my guess would be that these servers are designed using the SMbus method since battery life isn't an issue.

Did P debunk this? or was it a "I can't imagine this would be how it actually works"? I guess I am still a bit confused...
 
P Mar 26, 2013 04:53 AM
The reasoning here is that any other OS routinely adjusts voltage up and down, but that the way they do so is not supported by EFI, so they can't do that. That's just not true. By default, there is a voltage to the CPU, and the OS doesn't touch it. It can be adjusted in some cases, but it's not routine to do so.

Also: For a CPU to fail in the manner described here, it has to be significantly overvolted. Not just a step or two - you need to be over by some 25% or so at least. That can't happen by accident.
 
shifuimam Mar 26, 2013 09:08 AM
i think this "30+ years Apple tech" just really, really wanted to sound smart on the Internet. :D
 
abbaZaba Mar 26, 2013 03:10 PM
Quote, Originally Posted by shifuimam (Post 4223472)
i think this "30+ years Apple tech" just really, really wanted to sound smart on the Internet. :D
Definitely a high possibility. I trust the smart folks here and P was actually the person I was expecting to lay down some troof.

Too bad NewsPoster decided to make this some sort of sensationalist headline.
 
macgeek06 Mar 26, 2013 05:48 PM
EFI...
I think its BS too, the voltage and fan control, doesn't that reside within the EFI where the OS has very little to do with the CPU? Because even in Windows Boot camp, Windows doesn't even touch the EFI where it has any control over the fan or CPU Voltage.
 
arrannen Mar 29, 2013 05:53 PM
Linux on apple hardware...
I have been running ubuntu on a Macbook pro (1st model of the C2D's) for almost 3 years now and its run like a champ. no problems, always runs cooler than when it had osx on it.


I am always extremely skeptical of those who post such sensational things without any sort of proof to back them up.
 
turtle777 Mar 29, 2013 07:50 PM
Quote, Originally Posted by abbaZaba (Post 4223533)
Too bad NewsPoster decided to make this some sort of sensationalist headline.
That post whore needs to be banninated.

-t
 
All times are GMT -4. The time now is 07:08 AM.

Copyright © 2005-2007 MacNN. All rights reserved.
Powered by vBulletin® Version 3.8.8
Copyright ©2000 - 2016, vBulletin Solutions, Inc.


Content Relevant URLs by vBSEO 3.3.2