Diagnosing faulty MacBook Pro (late 2011) - super slow (permanently 90% System CPU usage), won't wake from sleep

[PPCnostalgic]

I like challenges too so I completely agree with your choice.

BTW Great CPU that Xeon X5670 on your Gigabyte build! I think there's really nothing better than that six cores Xeon for the actual price, I had one some time ago on my HP Z400 and it was very powerful and very stable. Did you know you can boot your GA-X58 from a nvme SSD? Mine boots fine and very fast from Samsung SM951 (see signatures)

[TheBloke]

1 minute ago, PPCnostalgic said:

I like challenges too so I completely agree with your choice.

BTW Great CPU that Xeon X5670 on your Gigabyte build! I think there's really nothing better than that six cores Xeon for the actual price, I had one some time ago on my HP Z400 and it was very powerful and very stable. Did you know you can boot your GA-X58 from a nvme SSD? Mine boots fine and very fast from Samsung SM951 (see signatures)

Yeah the X5670s are awesome - perfect balance of price and performance.  They overclock really nicely too. I had an i7 930 originally and it took me far too long to realise I could replace it with a Xeon in this X58, but I'm really glad I finally did.  I have two in my server as well.

The thing that really surprised me about X58 is that the 24GB RAM 'limit' is untrue.  It can take up to 48GB.  I still don't understand why the mobo manufacturers never publicised that.  Maybe Intel wouldn't let them support that config.

I didn't know that about NVMe, that's interesting to know, thanks.  Unfortunately I have no free PCIe slots so I wouldn't be able to add the necessary port.   I put all my data on my server which has lots of disks and which I access over 2 x 10GB NICS, so that gives me pretty good performance.

[PPCnostalgic]

When Gigabyte built our X58 boards there weren't DDR3 modules greater than 4Gb, I think that's why they stated 24Gb as maximum ;-)

[TheBloke]

I've stumbled into another period of remission.. this time lasting a couple of hours and still going.  The MBP is currently performing exactly as it should, with no slowdown and sleep/wake/startup all fine.

What changed? Well, I picked it up!  Maybe it's just really lonely and needs to be held?  Perhaps.  Anyway, for the first time in a couple of days I picked it up off the desk and carried it with me to the other side of the room so I could copy down the serial number into a Clover config.plist I was creating on my Hack.

With it still on my lap I started it up, and noticed it started immediately.  Since that point it's been running fine. 

I'm pretty sure that the first time it worked (when I only got 30 minutes) I had also recently moved it around a bit. This could be coincidence, but it might suggest that the problem could trigger according to physical motion - maybe there's a damaged or loose solder connection that sometimes gets pushed into the right position when the laptop is moved about.  That's potentially good news, as it might prove that there's a single cause to all the sensor failures.  I don't know if I can repair a bad solder joint on tiny surface mount components, but it's not impossible.  As long as it's not one of those ball-grid-array things, I'm pretty sure that's impossible to fix without the right equipment (reflow ovens and the like.)

With it working I tried again the SMC reset, and that showed me that I had not managed to do that before. I had wondered if that might be the case - because it doesn't startup properly, it wasn't processing my holding Shift+Control+Option+Power.  I know it worked this time because the light on the MagSafe connector changed colour after I did it.

I don't hold out any hope that this SMC reset will actually fix things of course.  It seems certain it must be a HW issue, and quite likely something like a loose connection or bad solder connection.  Still, it hasn't failed again yet, despite my moving it about quite a bit, and even tapping all around the case to see if I could trigger it shutting down (and thus maybe roughly locate the position of the bad component.)

LATER:

OK after about four hours it reverted to the fail state.  It didn't actually reboot, just the kernel CPU usage suddenly went sky high.   It actually happened when I was away from the machine, so there wasn't any physical movement associated with it as far as I can tell.  That somewhat contradicts my earlier thought of it being related to movement.

It was running properly long enough for me to install some software - XCode, finally, so I can later DTrace to work out which kernel module is responsible for the high CPU 'safety measure'.   And also iStat Menus which gives me some monitoring.

And that now allows me to see the nature of the problem - all those sensors that report as failed in ASD are showing as their max values in iStat Menus.  So something is pulling those sensor lines to the maximum.  Going back to the Apple repair 'beginners guide' I mentioned, Louis Rossman says:

Sometimes the data line on which the sensor is communicating is being brought down by something – dirt in a webcam connector, a bad webcam, a bad chip. Check if the line is topping out at the max spec'd voltage of that data line. 

I do see the sensors going to maximum.  So there's still some hope it could be a simple fix, like dirt in a connector - that would also fit with it intermittently working.  Or of course it could be a bad chip, which is likely to be much harder to repair.  I miss the days of when chips were in sockets..

SLIGHTLY LATER AGAIN (last one tonight I promise):

I wiggled the laptop around a lot and the CPU usage disappeared and those sensors changed values.  Interestingly, they didn't all seem to revert to normal levels, but I guess they came down enough to be considered 'OK'. 

So movement definitely is important.  This more strongly suggests bad connection or dirt than a bad chip - as does the fact that the movement lowered the value of the bad sensors, but not down to what (I think) is normal levels.  Or it could still be a bad solder joint.  Or even still a bad chip I guess, but hopefully less likely.

This coming week I really do need to take it apart completely

Edited March 26, 2018 by TheBloke

[BTrain]

Hi, I know this is an old topic, but I was wondering if you ever managed to solve the issue. 

I'm having the exact same problem on a mid 2012 MacBook Pro and I would like to repair it myself without having to replace the whole logic board. 

Any help will be appreciated. 

Thanks a lot in advance.

[Camicia]

Wow.. when I read your first message, I thought .. that is exactly what is happening to me.  I just posted here: 

After a bunch of researches I realized that "High CPU on the kernel_task" is a very generic problem that has a lot of multiple causes including hardware problems. 

I have been reading and a lot including Louis Rossman videos and his forum https://www.rossmanngroup.com/boards . If I remember correctly I saw a thread where somebody had similar errors on the sensors. check on the forum. In that case it turned that the sensor values were legit and the problem was a broken heat pipe. If bended it may break inside. It is not visible but it does not properly dissipate the heat. The CPU was really overheating and a new heat sink fixed the problem.