Linux on MacBooks and CPU voltage - can it destroy my CPU?

augustl · 2010-02-27 14:27:29

According to someone who apparently is an Apple repairman, there are issues with Linux on MacBooks. The gist of it all is that only proprietary drivers in OS X and bootcamp are able to set the correct voltage on the CPU. The claims go on to state that the wrong voltage caused by Linux' inability to control it, can and will ruin the CPU after 6-9 months of usage.

Here are some pastes from ##mac on freenode (IRC):

10:13:23 augustl | Branes: I've been asking some linux-on-mac questions in here, and every time I did your name was mentioned. CPUs getting fried etc. Got any links or tips?
10:13:48 augustl | or is it a one way ticket to doom?
10:14:23 @Branes | Keep your linux inside a virtual machine.
10:14:39 augustl | so there's no sane way to mitigate the issues?
10:16:05 @Branes | None, short of writing your own SMC driver that can do what no-one others can -- manage Vcore correctly. Apple's hardware is slightly non-standard, and the AHCI module in their BIOS emulator tells inquiring OSes that it has SMBus, when it fact it doesn't.
10:17:08 @Branes | Every other maker uses SMBus properly, which means the motherboard's electronics has enough of its own smarts to look after the hardware until an OS tells it otherwise. SMC *forces* the OS to micromanage the hardware, as it has no autonomous smarts of its own.
10:18:24 @Branes | Without proper control of the SMC, the CPU is left in an over-volted state which causes cumulative, irreversible damage to the CPU matrix via electromigration (sometimes better known as "overclockers' cancer")
10:19:33 @Branes | What narks me off is that Apple know, but they don't care. It's not OSX that's at fault, so to them it's not a problem. It wasn't a big problem with the earlier Intel MacBooks, the CPUs in those were a bit tougher because the fgabbing tech was coarser.
@Branes | The all metal laptop enclosures are very good. The Mac Pro body is stunning (and its motherboard is pretty good too). Electronically, everything else Apple sell is garbage.
10:33:38 Colloguy | listen up everyone, you get a new mobo if you fry your current one by running linux!!
10:34:01 @Branes | No, you get a replacement. There is no guarantee it will be new.
10:34:12 @Branes | 70% of the logicboards I handle are refurbished.

10:48:52 augustl | so how does boot camp not fry CPUs?
10:49:08 @Branes | Simple: Apple provide drivers.
10:50:32 @Branes | Although that said, the "SMBIOS" drivers kit that Apple provide with Boot Camp actually cheap -- Windows can't perform the same degree of micromanagement OSX can, so the SMC 'driver' lowers the CPU's Vcore and then exits.
10:51:01 @Branes | It's partly why MacBooks run hotter and even have less battery-life when in Windows.

So far, this is my only source. Everyone in ##mac vouched for Branes, and he seems to know what he's talking about, so I'm worried that Arch will fry my CPU. In #archlinux, IRC, I mentioned this, and pretty much everyone though I was trolling. No wonder, these are pretty extreme claims. There has to be some truth in it, though? Perhaps this applies to some versions of the kernel, or specific distros?

I'd certainly like to see a few more sources on this matter.

Skripka · 2010-02-27 14:50:55

Whiskey-Tango-Foxtrot?

Why would the CPU be in an "over-volted" state to start with? Over-volting and heat can cause damage true-but I don't understand how or why a stock CPU sold by Apple would be over-volted in the first place. What moron makes the hardware run over-volted at a stock configuration without specific drivers? Especially on laptop systems-which have severe heat issues to start with.

He's either BSing, or Apple hardware/software is terribely designed.

I don't run Intels, but can the CPU voltage even be altered on non-Extreme edition CPUs?

I've known of and myself have run AMD CPUs in desktops at +0.1V and +0.2V from stock for years. The only reason I stopped using them was because better CPUs came out.

Last edited by Skripka (2010-02-27 14:59:16)

lifeafter2am · 2010-02-27 14:53:11

Well I've been running Arch on my Macbook Pro for about 6 months now .... so I guess my system is going to die soon. lol.

augustl · 2010-02-27 15:04:15

Skripka wrote:

He's either BSing, or Apple hardware/software is terribely designed.

According to Branes, the mac repairman, that's the problem Macs are designed to run OS X, and bootcamp is a hack, via a proprietary driver that starts, sets the correct voltage, and exits.

But I'm not sure if I'll run Arch in a VM just becasue 1 person said I should.

Skripka · 2010-02-27 15:17:16

augustl wrote:

According to Branes, the mac repairman, that's the problem Macs are designed to run OS X, and bootcamp is a hack, via a proprietary driver that starts, sets the correct voltage, and exits.
But I'm not sure if I'll run Arch in a VM just becasue 1 person said I should.

I'd sooner think he's BSing and fear-mongering. I can't say for certain though-as I don't run mac hardware, and the folks who *might* actually know have a VESTED interest in you running ONLY their hardware and software.

Electromigration happens in ANY electrical system. Just because you're running OSX doesn't mean electromigration suddenly stops or goes away because you're using "the right drivers". Higher power levels on a CPU (such as a higher VCore) do increase electromigration somewhat. How this effects mainboard/CPU life is anyone's guess.

Also know that CPUs have a great deal of redundency built in. Guys over on Tom's Hardware have run the same benchmark after a period of 5+ years on the same hardware, and found reduced performance--possibly due to any number of things.

The only way for certain to pin CPU death on electromigration for certain is to pullout a very high-power microscope and examine BILLIONS of copper interconnect switches for ion pollution...something I STRONGLY doubt Apple has ever made an extensive survey of doing....examining one computer alone would take weeks or months

"Overclocker's Cancer" indeed. MOST overclockers altering with VCore setting in the BIOS don't use a CPU long enough before electromigration (or anything else) kill their CPU. The main people who need to worry about CPU life are the casual computer users who still have and run P3 and P4 chips.

Last edited by Skripka (2010-02-27 15:19:50)

augustl · 2010-02-27 15:22:05

Perhaps I can find a tool that yields CPU information (including voltages) on both OS X and on Linux. Then again, if Apple actually reports wrong values to the OS, that would be kind of pointless..

augustl · 2010-02-27 15:29:49

Another source. Same guy, though. http://forums.gentoo.org/viewtopic-p-61 … 1f41538666

Skripka · 2010-02-27 15:32:35

augustl wrote:

Perhaps I can find a tool that yields CPU information (including voltages) on both OS X and on Linux. Then again, if Apple actually reports wrong values to the OS, that would be kind of pointless..

The only way to get anything useful out of it would be to compare your readings to official Intel recommendations.

For Windows there's CPU-Z, a GREAT program with no port that I know of for OSX or *nix.

For Arch, there used to be a CPU-Z like utility in AUR such as perlmon. lm-sensors may also give voltage readings on your hardware, depending on support.

lifeafter2am · 2010-02-27 15:36:57

augustl wrote:

Another source. Same guy, though. http://forums.gentoo.org/viewtopic-p-61 … 1f41538666

And everyone there thinks the same thing ... that it is B.S..

I guess my thing is why would only one person know this? Why would there be no others mentioning this?

Last edited by lifeafter2am (2010-02-27 15:39:20)

augustl · 2010-02-27 16:05:18

There's mactel-linux, but it seems pretty outdated. http://mactel-linux.sourceforge.net/wiki/AppleSMC

augustl · 2010-02-27 17:55:31

Another source. I think this comes from info from Blaines as well, though.

http://forums.macrumors.com/showthread. … ost8984536

Last edited by augustl (2010-02-27 18:09:30)

lifeafter2am · 2010-02-27 17:59:10

augustl wrote:

Another source. I think this comes from info from Blaines as well, though.
http://forums.macrumors.com/showpost.ph … stcount=10

Where is the rest of the topic? You show only a single post?

augustl · 2010-02-27 18:09:51

lifeafter2am wrote:

Where is the rest of the topic? You show only a single post?

Fixed the link.

wuischke · 2010-02-27 18:10:02

I don't run Intels, but can the CPU voltage even be altered on non-Extreme edition CPUs?

Just to clarify this: It's changed automatically when the frequency is lowered for all CPUs supporting speed step. (At least in my experience; Same for AMD's Cool & Quiet)

A possibility for the original claim would be that it doesn't change voltage when lowering the the CPU frequency. But I couldn't imagine this affecting live expectancy that much.

Last edited by wuischke (2010-02-27 18:11:56)

augustl · 2010-02-27 18:19:24

Blaines online on ##mac on freenode right now if you want to challenge him

augustl · 2010-02-28 09:38:12

Nothing but speculation in the Ubuntu forums either.. http://ubuntuforums.org/showthread.php?t=1417752

R00KIE · 2010-02-28 14:38:02

If the cpu is overvolted when booting _without_ any OS then it is a terrible design flaw, manufacturers release datasheets with the __safe__ operational limits for a good reason.

Relying on the OS to monitor and manage critical settings is just stupid, it is ok for the software to be able to override some settings after the OS as booted but no software is perfect, so for very sensitive settings such as the cpu voltage there should be some lower level, very stubborn and dumb safeguard to forbid dangerous settings. Everyone knows notebooks have bad cooling so low level safeguards should be in place to keep things from burning.

It raises a good question though, why does apple boot the system with the cpu overvolted in the first place? Boot settings are supposed to be safe and reliable ones. And why does it design broken hardware, if the hardware doesn't support smbus and if it is meant to use only their OS there is no need to report it supports smbus, if the hardware is not meant to run only their os then they should design hardware that does what it claims to do.

Skripka · 2010-02-28 15:31:49

wuischke wrote:

I don't run Intels, but can the CPU voltage even be altered on non-Extreme edition CPUs?
Just to clarify this: It's changed automatically when the frequency is lowered for all CPUs supporting speed step. (At least in my experience; Same for AMD's Cool & Quiet)

I was more thinking of the possibility of over-volting. Underclocking can be automatically done under Linux/Windows, and presumably Mac. But unless you own an Extreme Edition Intel, I didn't think it was possible to overvolt the CPU with Intel CPUs....as that requires an unlocked BIOS, I thought.

With AMDs CPUFreq under linux cuts the frequency multiplier down drastically for throttling (I didn't think it did anything to VCore). Windows7 cuts the multiplier a great deal too, possibly the VCore although I can't remember if W7 futzes with it or not.

wrote:

A possibility for the original claim would be that it doesn't change voltage when lowering the the CPU frequency. But I couldn't imagine this affecting live expectancy that much.

WinXP didn't have any throttling by default and always ran at stock maximum frequency. Although people did tend to notice slowdowns, which usually got blamed on registry bloat over time. Some folks over at Tom's Hardware have had hardware 5+ years and did the same benchmark, and even with a fresh OS noticed slowdown; their theory being possibly that ion pollution over time damaged ICs on the CPU and the CPU rerouted around the damage interconnects. I haven't read anything certain about this though.

Last edited by Skripka (2010-02-28 15:42:03)

thestinger · 2010-03-01 04:52:16

you can overvolt any cpu as long as your motherboard is capable of it - the only time you would have a "locked" bios is if you buy an OEM computer (HP, Dell, etc)

I think you're confusing the multiplier with voltage - extreme editions have an unlocked multiplier which gives you more options when OCing (you don't need to find a precise base clock and memory ratio, you just pump up the multiplier)

Last edited by thestinger (2010-03-01 04:53:35)

R00KIE · 2010-03-01 14:07:23

thestinger wrote:

you can overvolt any cpu as long as your motherboard is capable of it - the only time you would have a "locked" bios is if you buy an OEM computer (HP, Dell, etc)
I think you're confusing the multiplier with voltage - extreme editions have an unlocked multiplier which gives you more options when OCing (you don't need to find a precise base clock and memory ratio, you just pump up the multiplier)

Sure, if you neglect the problems of electromigration as the guy from apple mentions, the lower marging to the maximum gate voltage that degrades the life expectancy of the semiconductor, not even taking into account poor/cheaper/less complex voltage regulation circuits or less than optimal pcb designs that can lead to voltage overshoot and ringing, and the extra generated heat, yes sure why not, just not on my next laptop please.

augustl · 2010-03-01 14:52:19

I don't see any reason to doubt this Blaines guy, so I'll be running Arch in a VM for now.

lifeafter2am · 2010-03-01 15:08:56

augustl wrote:

I don't see any reason to doubt this Blaines guy, so I'll be running Arch in a VM for now.

Excpet for the tons of people (like me) running linux on thier Macs with no problems.

R00KIE · 2010-03-01 15:18:39

augustl wrote:

I don't see any reason to doubt this Blaines guy, so I'll be running Arch in a VM for now.

He must have a point because he is in a position to know how things work, however it doesn't change the fact that they (apple) sell broken by design machines (by his own words), and it is in their best interest (just like any other maker of proprietary things) to spread some fud.

If they want to keep in business, either they change their ways or find new and improved fud and vendor lock-in techniques to keep people buying their stuff. But maybe it's just me who can't see the attraction of buying something I should be able to use as I want and I can't.

litemotiv · 2010-03-05 12:28:32

the cpu voltage should be readable by programs like cpu-Z, everest or maybe even lm-sensors on linux, so it's a matter of checking if it stays within normal bounds on non-macOs systems. note also that the above story would mean that anyone running Windows on their mac without bootcamp would also face the same 'problem'.

this has fud written all over it, but i'd be interested in seeing some actual measurements.

Peasantoid · 2010-03-05 21:11:14

shenanigans();

I have been using Linux on an mb3,1 more or less continuously for well over six months. In the very unlikely event that this assertion *isn't* made of stupid, I suppose I'll find out soon.

** The guy's probably an Apple meat puppet, come to think of it...

Last edited by Peasantoid (2010-03-05 21:14:18)

Arch Linux

#1 2010-02-27 14:27:29

Linux on MacBooks and CPU voltage - can it destroy my CPU?

#2 2010-02-27 14:50:55

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#3 2010-02-27 14:53:11

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#4 2010-02-27 15:04:15

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#5 2010-02-27 15:17:16

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#6 2010-02-27 15:22:05

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#7 2010-02-27 15:29:49

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#8 2010-02-27 15:32:35

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#9 2010-02-27 15:36:57

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#10 2010-02-27 16:05:18

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#11 2010-02-27 17:55:31

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#12 2010-02-27 17:59:10

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#13 2010-02-27 18:09:51

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#14 2010-02-27 18:10:02

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#15 2010-02-27 18:19:24

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#16 2010-02-28 09:38:12

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#17 2010-02-28 14:38:02

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#18 2010-02-28 15:31:49

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#19 2010-03-01 04:52:16

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#20 2010-03-01 14:07:23

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#21 2010-03-01 14:52:19

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#22 2010-03-01 15:08:56

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#23 2010-03-01 15:18:39

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#24 2010-03-05 12:28:32

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

#25 2010-03-05 21:11:14

Re: Linux on MacBooks and CPU voltage - can it destroy my CPU?

Board footer