kernel_task consuming 150% CPU; system unusable

The miniDVI adapter is in my second-gen. MBA and log_level is tame. If I fire up a PPC application, then everything shoots up.

Hello, it definitely is the update which was supposed to resolve the problem.

You easily can reproduce it:

a) Install a fresh osx from the CD onto an extern drive and you will see the kernel task problem is not there although the CPU gets hotter (which is a non problem for me)

b) Do the latest updates again the kernel task problem is there.

I first thought that renaming flash.xpt fixes it, but it does not it somehow relieves the problem only somewhat!
But if you stress the machine with other things, suddenly you have the kernel again going haywire!

The funny thing is this only seems to happen on the MBA, the Mac Mini I have at home never has this problem.
The problem pretty much is as bad as the core shutdowns because it pushes the machine into an unusable state!

So I revert my former postings here that removing flash.xpt resolves the issue once and for all, it just was one trigger of a problem way deeper and no I do not think the problem really is hardware related, it looks more like a software issue to me introduced by a fix for another problem!

Btw. another thing which can trigger the issue is broken file permissions repairing them also helps somewhat but does not resolve the issue entirely!

No, that is completely unrelated to this problem.

Any news here?

I also have the "log_level" problem consuming all my CPU...

Well, my news is that my computer has become even worse, to the point of being useless. I had my fan replaced, and it seems to have become worse, although I cannot for sure say that's the reason (could be 10.5.7 update etc etc). Just now the computer was essentially idle (the most active application was safari with 5% cpu). I typed a search in spotlight, and immediately kernel_task shot up to 150%. Previously, I have been able to at least do simple things on the computer. Now it essentially only works when idle -- which is of course not very useful for a computer. Also, the fan runs constantly when I have the computer in my air-conditioned office. Previously, it would only run during cpu intensive tasks (such as TM backup), and quickly slow down afterwards. Now it just runs constantly. The MBA (gen 1) definitely has a cooling problem, and Apple has not found a solution.

So far I have stayed away from coolbook, finding the solution worse than the problem, but maybe I need to try it out.

Today, Google brings me here.

Spent a few hours following the threads and testing various suggestions, I have a feeling that the cause of the 100%+ CPU issue could be:

summer time!?

Today I work in a 30C+ room without air-conditioning. A reasonable guess?

I was checking the posts on this issue after working on my Macbook Air in the sun, and all of a sudden, kernel task consumed 150% plus CPU. In fact, about 10 min back inside, it is back to normal. It seems right that one core seems to shut off if there is too much heat.

Exactly.

Yes it is a cooling issue, the kernel task seems to do something to keep the processor load occupied or down, so that the processor does not really overheat.
I could notice that today again, when things became warm! Tasks which normally did not cause the kernel task go haywire went triggered the ominous kernel task today constantly, and brought the machine on the edge of being unusable (waiting for minutes for things to complete instead of seconds)
The reason also, 30 degrees+ in an unairconditioned office!

Yep. Exactly the same behaviour here, when ambient temperature reaches ~30C, and mostly when i'm using the ac adapter. I found a workaround though, that does the trick: +keep the airflow free beneath the MBA.+ Just place a thin object (~10cm height & wide enough to balance the device) on the surface under the MBA. Never had a problem since, even though core temperatures reach as high as 80C. If I place the MBA plain on the table surface, the erratic core overload reappears within 5min. Somehow, the above trick does the job. Try it out.

On upgrading to 10.5.7 my MacBook Air began experiencing this problem. Even after a clean boot and running no applications of note kernel_task eventually triggers and spins, making the machine unusual. I have reported the problem to Apple and will call AppleCare later today.

Ok I have found a temporary solution, have in mind this is not for the weak of heart, and if you are unsure what you are doing then dont do it.
The solution is simply to get the additional cooling to the machine, one way or the other, cooling keeps the kernel task down.

In my case for now I am using gel pads which I put in the freezer, and then wrapped into dry (it must be dry) plastic to keep the condensed water away from the machine. So far it looks good the kernel task seems to be down at 2% no matter what I am doing, this is the definitve proof that the cooling of the macbook air seems to work only under airconditioned conditions properly but not at normal operating temperatures in summer!

I just wanted to give a hint, anyway, have in mind if you do this, try to keep away water or any liquid from the macbook air, it otherwise might damage your machine or even would do worse to your health!

I have a small addition since the gel pad solution works fine, it might help if you add a notebook cooling solution to your desk setup where you can place the air!
You just have to add additional cooling so that the not properly working cooling has additional help.
I just wonder if the cooling issue also can be resolved if you add a backplate to the notebook which is slightly bigger in its volume, it looks to me that it might help!
Has anyone tried to run the thin without the backplate on?

Ok another addition, here is a summary of all things.
First of all the problem seems to be happening most at the first gen MacBook Airs, the thing seems to be the heat. Proof, try to add some additional cooling and see the kernel task go down no matter what.

It usually happens at summer temperatures in rooms, if the machine is stressed a little bit. The stress can happen from anything, the easiest to discover it is to play a flash video on an untouched machine.
The problem itself has manifested itself after a system upgraded around 10.5.4, before the machines became really hot or got core shutdowns.
Also it is more likely to occur during loading of the batteries, because additional heat is produced by the loading cycle!

The reason for this and additional added cooling is the definitive proof is simply cooling for some way or the other the machine cannot deliver the needed cooling in many not all cases to stop the mechanism to trigger!

I assume second generation AIRs are less prone to the failure due to either some fixups apple did in this area or due to the fact that they work on the NVidia 9400m which is most likely less heat dissipating than the Intel graphics chipset which is known to become quite hot.

Watching the processor sensor does not really help because the processor temperature does not seem to be the only cause for the internal heat calculation so that the mechanism triggers there seem to be other sensors in place which also can be a trigger!

Solution for now, add additonal cooling one way or the other, then you have proof. Solution in the long run, I hope Apple will fix it, I think the external cooling is enough that a number of machines are broken so that they cannot talk themselves really out!
Sorry to sound a little bit harsh, but I am more than p ** about the silence from apples side, less expensive machines got a replacement of faulty parts while many of the first gen macbook air owners have to sit on semi working machines.

Anyway I will call Applecare today, to present them my proof lets see what the reaction is, the last time they said they were not aware of an issue, and gave me the standard advices of pram and prc resets etc...

Ok, first of all, I now called Apple Care a second time, I got the usual drone, but before even going into the standard procedure of things to do I told them about the additional cooling things I found out and they told me to give the machine to an authorized service provider for further investigation. I assume that the machine will have its logic board replaced. I will post futher results as soon as I will get the machine back (I will go the the service provider around next week!)