Kernel_task issue?

[leftroom]

Hello MacRumors.

I'm getting a bit tired of this now.. Feels like I'm pushing the power-button on my MacBook Pro every week because of the kernel_task.

Here it is:
Kernel_task is up at 100%. The computer feels very slow and I want to do a restart. Some of the apps have stop responding as well.

I'm doing a normal restart which means the Apple menu - restart button. The grey screen appears with a loading picture. I wait and wait.. Nothing happens. So I hit the power button. And boom, the computer restarts and all is fine for a week or two. Then back to 100% cpu kernel_task and I need to do this all over again.

I have done several Pram resets but nothing seems to work. I'm on a retina macbook pro with 10.8.2. I'm happy with the computer but not the OS. In the console.app I can't find anything useful.

Maybe a fresh reinstall helps? I'm out of ideas.

[skirmisser]

Hi,

I also have *exactly* the same issue. Kernel_task at 100% every few days. Retina MBP with 10.8.2 (came already preinstalled with ML, not an upgrade over Lion).

Have tried disabling PowerNap, wake on WiFi. Just reseted NVRAM/PRAM and SMC.

If this happens again I am going to the genious bar. Has anybody solved this issue?

[leftroom]

Now it has happen again while I was watching a episode of South Park online.

So frustrating!

http://i47.tinypic.com/505gt3.png

When I restart often the computer hangs at the grey screen trying to reboot. I see a wheel spinning and nothing happens for like 2-3 minutes. So I shut down the whole computer with the power-button.

[skirmisser]

Yes, it does happen completely at random. Sometimes it takes 8, 7, 5, or 3 days to manifest.

My system came with ML 10.8 and I updated it first to 10.8.1 and then to 10.8.2. I assume it is the same case for you?

I have reset NVRAM/PRAM and now I am problem free for 2 days...no guarantee though it will not happen again. Rember and Apple Hardware Test were run and have found no hardware/memory issue.

I have read elsewhere this may be related to WiFi. I have a linksys WRT54g router. What kind of router are you using?

[leftroom]

Originally Posted by skirmisser

Yes, it does happen completely at random. Sometimes it takes 8, 7, 5, or 3 days to manifest.

My system came with ML 10.8 and I updated it first to 10.8.1 and then to 10.8.2. I assume it is the same case for you?

I have reset NVRAM/PRAM and now I am problem free for 2 days...no guarantee though it will not happen again. Rember and Apple Hardware Test were run and have found no hardware/memory issue.

I have read elsewhere this may be related to WiFi. I have a linksys WRT54g router. What kind of router are you using?

First of all, excuse me of my language because I'm from Sweden

Yes, totally random it happens. Boom, kernel_task at 100%. Nothing you can do about it.

No, I got a retina MacBook Pro which came with Lion 10.7.4 I think so I updated to ML trough Mac App Store. But from there I updated to 10.8.1 and now 10.8.2.

I also have done a NVRAM/PRAM reset several times but no luck after that.

Interesting! I have a Linksys WRT54GL (L is for Linux I think. I have DD-WRT firmware)

Soon I will be upgrading my broadband-connection. Maybe it will be OK then

[thoughtcrimes]

I am having the exact same issues with my MacBook Pro Retina with 10.8.2. The kernel_task is over 100% CPU, the computer runs "warm," and every few weeks I have to force restart by holding down the power button.

Now that we know others have the same symptoms, has anyone found any causes or fixes for these issues?

MBPr 2.6GHz Intel Core i7 - 16GB 1600 MHz DDR3 - OS X 10.8.2

[skirmisser]

Originally Posted by thoughtcrimes

I am having the exact same issues with my MacBook Pro Retina with 10.8.2. The kernel_task is over 100% CPU, the computer runs "warm," and every few weeks I have to force restart by holding down the power button.

Now that we know others have the same symptoms, has anyone found any causes or fixes for these issues?

MBPr 2.6GHz Intel Core i7 - 16GB 1600 MHz DDR3 - OS X 10.8.2

I just got it again, after roughly 6.5 days of uptime. So pretty much identical case as you. Just updated to 10.8.2 12C60 (the Supplemental Update 1.0) from 10.8.2 12C54 (this is what I was running prior to today's 100% event), though I somehow doubt this will fix it. What type of Wifi router are you using? Today it happened while I was streaming video.

[gredman]

Same issue here.

Each time it occurs I've been running hotkernel.pl, and each time there seems to be a lot going on in NVDAResman, which I assume is something to do with the Nvidia graphics:

NVDAResman`gpumgrIsParentGPU                               18   0.0%
AirPortBrcm4331`wlc_bmac_suspend_mac_and_wait              20   0.0%
mach_kernel`__bzero                                        21   0.0%
NVDAResman`threadStateResetTimeout                         23   0.0%
NVDAResman`gpumgrGetDeviceInstance                         28   0.0%
NVDAResman`osReadFb032                                     29   0.0%
NVDAResman`inforomSetParityBufferState                     30   0.0%
NVDAResman`insert_registration_func                        34   0.0%
mach_kernel`lck_mtx_lock                                   52   0.1%
AppleIntelFramebufferCapri`SafeForceWake(bool)             54   0.1%
NVDAResman`rmGenerateSha256Gid                             69   0.1%
mach_kernel`lck_mtx_unlock                                 74   0.1%
NVDAResman`osMemAddFilter                                  75   0.1%
NVDAGK100Hal`NVDAGK100HAL::~NVDAGK100HAL()                170   0.2%
NVDAResman`pdbGetProperty                                 184   0.3%
NVDAResman`gpuExecRegOps                                  307   0.4%
mach_kernel`ml_set_interrupts_enabled                     332   0.5%
NVDAResman`odbGetClass                                    351   0.5%
mach_kernel`machine_delay_until                           443   0.6%
mach_kernel`dtrace_unregister                            1106   1.6%
NVDAResman`osPciReadDword                                1982   2.9%
mach_kernel`_rtc_nanotime_read                           4655   6.8%
mach_kernel`machine_idle                                57622  84.2%

Doesn't look like a lot but NVDAResman`osPciReadDword is normally around 0.1%, so I'm suspicious.

Seems to only ever occur while running integrated graphics, and trying to subsequently switch to discrete graphics (using gfxCardStatus) has no effect ("Discrete only" is checked but it's still using integrated). Also, trying to open the Displays preference pane hangs.

My theory is it's a graphics driver bug, possibly around switching between integrated and discrete. I'm going to try running only discrete graphics for a week or two and see if it reappears.

[skirmisser]

That's pretty interesting.

So to clarify: did it happen while you were using integrated graphics? If I correctly understand it did not happen while you forced a switch using gfxCardStatus? If that is true it is interesting since NVDA is causing it even though the machine was in integrated graphics mode.

[skirmisser]

Originally Posted by gredman

Same issue here.

Each time it occurs I've been running hotkernel.pl, and each time there seems to be a lot going on in NVDAResman, which I assume is something to do with the Nvidia graphics:

NVDAResman`gpumgrIsParentGPU                               18   0.0%
AirPortBrcm4331`wlc_bmac_suspend_mac_and_wait              20   0.0%
mach_kernel`__bzero                                        21   0.0%
NVDAResman`threadStateResetTimeout                         23   0.0%
NVDAResman`gpumgrGetDeviceInstance                         28   0.0%
NVDAResman`osReadFb032                                     29   0.0%
NVDAResman`inforomSetParityBufferState                     30   0.0%
NVDAResman`insert_registration_func                        34   0.0%
mach_kernel`lck_mtx_lock                                   52   0.1%
AppleIntelFramebufferCapri`SafeForceWake(bool)             54   0.1%
NVDAResman`rmGenerateSha256Gid                             69   0.1%
mach_kernel`lck_mtx_unlock                                 74   0.1%
NVDAResman`osMemAddFilter                                  75   0.1%
NVDAGK100Hal`NVDAGK100HAL::~NVDAGK100HAL()                170   0.2%
NVDAResman`pdbGetProperty                                 184   0.3%
NVDAResman`gpuExecRegOps                                  307   0.4%
mach_kernel`ml_set_interrupts_enabled                     332   0.5%
NVDAResman`odbGetClass                                    351   0.5%
mach_kernel`machine_delay_until                           443   0.6%
mach_kernel`dtrace_unregister                            1106   1.6%
NVDAResman`osPciReadDword                                1982   2.9%
mach_kernel`_rtc_nanotime_read                           4655   6.8%
mach_kernel`machine_idle                                57622  84.2%

Doesn't look like a lot but NVDAResman`osPciReadDword is normally around 0.1%, so I'm suspicious.

Seems to only ever occur while running integrated graphics, and trying to subsequently switch to discrete graphics (using gfxCardStatus) has no effect ("Discrete only" is checked but it's still using integrated). Also, trying to open the Displays preference pane hangs.

My theory is it's a graphics driver bug, possibly around switching between integrated and discrete. I'm going to try running only discrete graphics for a week or two and see if it reappears.

I just got the same case as above. NVIDIA driver running at the same time as Intel driver and kernel_task shooting up to 100%. I experimented with switching to NVIDIA and back to Intel (after restart) and in each case only one of the drivers is active. So it seems somehow all hell breaks loose and both drivers are working at the same time leading to this mishap.

When it happened the machine was (according to gfxCardStatus) at internal GPU mode. Note that I never once used gfxCardStatus to switch GPUs. Just use it to monitor.

Thus seems like a driver issue. Will also try to use discrete GPU and report after a couple of weeks.

[gredman]

Over two weeks now locked on discrete graphics and I haven't had any issues.

[BreeeZe]

Any one else try this, did this indeed fix the problem?

I'm having the exact same issue with my rMBP.

Is there any place we can file a bug report at Apple?

[skirmisser]

One solution is to do what the above poster did: switch off automatic graphics switching.

The other solution that I tried and I have 10.5 days uptime already is to just not use Safari.

It seems it relates to graphics switching + Safari.

[leftroom]

I have found the issue. It maybe be like you said skirmisser, Safari. It completely hanged and gave me a beachball. So I clicked Force Quit and terminated Safari. After that kernel_task went up to 100% and stayed there.

I clicked on reboot and the I'm getting a spinning wheel and nothing happens so I press the power button.

Maybe I shall stop using Safari?

[MikieMcBikie]

I have had this issue with my 2012 13" macbook pro. I havent had the issue for several days now, and what I "believe" my issue was I had trim enabler enabled on my vertex 4....although i am still waiting on it to happen again since it was soo random I wont really know for many more weeks lol. Odd thing is a buddy at work has the exact same machine, and upgrades, but his is still flipping out after doing the same fixes I have, so I am truely doubtful i have this fixed.

here are the things i did before it stopped locking up.
onyx, all automation options selected.
disabling trim enabler
changed from linksys router to apple time capsule(had this on hand this already, but i had read somewhere someone thought it may be the linksys router) If i stay solid, i may go back to the linsys for a bit and see if the issue reappears.



my thread
http://forums.macrumors.com/showthread.php?t=1470911

[MikieMcBikie]

mine just did it again.. i am lost lol

[skirmisser]

Originally Posted by leftroom

I have found the issue. It maybe be like you said skirmisser, Safari. It completely hanged and gave me a beachball. So I clicked Force Quit and terminated Safari. After that kernel_task went up to 100% and stayed there.

I clicked on reboot and the I'm getting a spinning wheel and nothing happens so I press the power button.

Maybe I shall stop using Safari?

Well I hope that they will address this issue soon. In the meantime just by using Chrome and avoiding to use Safari and leaving automatic graphics switching on I am close to 13 days up time. Using Safari I would get the kernel_task 100% event every now and then. The shortest period was 3 days and the longest 8 and a half.

Safari and driver switching must really be messed up somehow.

[leftroom]

Originally Posted by skirmisser

Well I hope that they will address this issue soon. In the meantime just by using Chrome and avoiding to use Safari and leaving automatic graphics switching on I am close to 13 days up time. Using Safari I would get the kernel_task 100% event every now and then. The shortest period was 3 days and the longest 8 and a half.

Safari and driver switching must really be messed up somehow.

Ok, thanks for the tip. I will be using Chrome for now on. When I first got the computer when I launched Chrome I got an kernel panic. But maybe that issue is gone now.

I have never been able to have about 13 days of uptime. Maybe 4-5 days. Because I had the freezing issue also but the computer hasn't been freezing for several days now. Only the kernel_task went up to 100% because of Safari.

[skirmisser]

Got almost 15 days of uptime just by using Chrome and automatic switching on.

Just updated to the newest 12C3006 10.8.2 update (just do software update, came out today) so uptime is reset . Started using Safari to test to see if this update resolved the issue.

Worth noting that Chrome is bug-free and does not bring down the system (kernel_task 100%), but in terms of UI scroll performance and rendering quality (both text and images) Safari is in a whole different league...Chrome is both *slow* and *ugly*...

[BreeeZe]

Installed the update a few hours ago and safari seems to work a lot smoother.
I haven't had any weird graphics issues like partial screen tearing yet on websites which had it previously.
Also scrolling seems a lot smoother.

[BreeeZe]

yey.. the update introduced a new bug.. my rmbp didn't want to wake up from a nights sleep just now...
The screen just turned on and off and showed a black screen with just a mouse cursor..

Had to do a hard reboot..

[skirmisser]

After more than 2 weeks of uptime I am happy to report that 10.8.2 12C3006 fixed the stability issues with graphics switching and safari.

Still I encountered the following issue if I play StarCraft 2. If after playing the machine hibernates and then resumes and I load SC2 again, even though the NVidia GPU is again activated it becomes horribly slow and only after an SMC reset (following shutdown) the issue is resolved.