See this thread:
http://setiathome.berkeley.edu/forum_thread.php?id=46914#749496

I am having the same problem now. Before I upgraded to Leopard, S@H ran fine on my laptop at 40% CPU utilization. With BOINC 5.x it even obeyed the battery/AC power status and all was fine.

Then I upgraded to Leopard and the 5.x setup could no longer run properly. I decided to upgrade BOINC to 6.6.20 and got S@H running again, but now every couple of days S@H goes renegade and starts consuming all the CPU it can get.

This is a NASTY bug for laptop users.

I suspect it may be related to the completion and starting of new tasks, since I've noticed a couple of times that the work unit was at <1% completion (a new one had just started) when my laptop got hot and the fan started huffing constantly.

Yes, it does that on my 17" MacBook Pro all of the time on Leopard. Heats up the enclosure to 104 degrees Fahrenheit, has the fans spinning at top speed and generally causes concern. And that's with everything else shut down. It does not seem to have this effect on the iMac that also runs Leopard. Go figure.

The above, of course, makes me wonder how hot the machines will run if S@H implements the Nvidia GPU processing for OSX even before Snow Leopard is released (which ostensibly will also give us access to the GPUs for general processing, like adding another CPU to your machine, except that the GPU runs two to ten times faster, depending on the application). Faster processing is great, but it makes me wonder if ill-mannered Boinc will burn up my laptop in the process, especially once we have the GPU processing up and running for OSX.

---

After a few months off, I just rejoined on SETI (been a member since 2000) and had to upgrade Bonic to be able to download work. Now my old iMac G5 1.8 GHz runs out of control (CPUI goes to 100%) under Leopard even though I set all preferences to 20%. It even does it on snooze and/or suspend. Only way to stop is to quit Bonic.
Please advise a fix or I must bow out of a project I would like to support.
Eric Pearson

---

I have forwarded this to the BOINC developer for the Macintosh.

---

Hi,

This is a weird one for me. I have a MacMini running OS X 10.5.8 running S@H on BOINC 6.6.36, and had a 1.67GHz PPC PowerBook running the same S@H and BOINC versions, but on OSX 10.4.11.

Both systems ran perfectly - obeying any throttling commands and suspending themselves when either machine became active through my actions as a user.

With the recent release of the OS X 10.5.8 combo update from Apple, I decided to finally take the plunge and update the PowerBook. This was done first to 10.5.4, then combo updated to 10.5.8.

I re-installed S@H/BOINC from scratch (after doing a comprehensive uninstall and reboot). Since the fresh install, I too have noticed that the PowerBook (laptop) and NOT the MacMini is suffering from this rogue 100% (or as much as it can eat) CPU usage and fan screaming behaviour.

It is not consistent, but does happen maybe 80% of the time. Quitting BOINC manager and client and restarting them fixes it, but often only temporarily.

Rachel

I have forwarded this to the BOINC developer for the Macintosh.

Hi Jord, have you had any feedback as yet?

I'm just curious as I did get a shared memory error a few days ago and went looking for a fix, and found that by following the instructions below (on my PowerBook) the 100% rogue CPU behaviour seems to have gone away...

So for those others who are having this issue, go and look at the article here: http://www.spy-hill.com/~myers/help/apple/SharedMemory.html

If it helps (or doesn't help) resolve your problem - please post an update.

cheers
Rachel

Well, I spoke too soon.

It has finally come back (with a vengeance) and I'm wondering if it is tied to particular work units? Tonight I couldn't even get the machine to unsleep its screen and present me with the screen unlock dialogue (username/password).

Not sure if this helps, but the messages tab shows that when the CPU is red-lined only the network work is suspended. Normal operation sees a line saying the computation work is suspended (user is active) then immediately followed by the network work is suspended (user is active) message.

Going to each task in the task tab shows that each task is supposedly suspended as the user is active, but the CPU is still maxed out. Even manually forcing a "suspend" on each task does nothing.

The CPU only drops back to "normal" levels several seconds after BOINC manager is quit (and I have quit set to kill the client too).

So - I've set the PowerPC to "no new tasks" and when the current 2 are finished, will detach it and fully de-install BOINC.

Both my Intel CPU machines seem well behaved (added a Macbook as another computer in addition to the MacMini).

A fix would be nice - I can see my RAC and world ranking slipping away without the PPC contributing... :-(

R

Can you tell which process is hogging the CPU - BOINC manager, the client or the science application?

Can you tell which process is hogging the CPU - BOINC manager, the client or the science application?

Given the CPU consumption after BOINC Manager has been quit - it takes several seconds for the CPU use to drop from 100% back to whatever the load is without BOINC/S@H running.

So, I would assume it is the client or the science app. I must admit I didn't run Activity monitor up to check which (sorry)...

Bizarre how it is only affecting (for me) the combination of OS X 10.5.8, BOINC 6.6.36 on a 1.67GHz PPC CPU with specific work units.

I also got the shared memory error again, even though OS X now starts with 4x the defaults.

R

hi - any updates?

Try BOINC 6.10.4 from here, see if it does it better.

---

Hi,

I am having the same problem. Boinc version 6.10.4 still causes the 100% CPU load. I am running a new macbook pro....

I also noticed that the screensaver will sometimes go black in the middle with two random green lines across the screen. When I send the input in, it throws me to my password screen with no problems though.

---

Has anyone found the cause of this as yet??

---