Q: display anomalies

Been fiddling with fix recipes, since my system has been made unstable by using Pacifist to install.

My machine definitely has something corrupting system software resources. I'm now wondering if failing PRAM battery underlies this sometimes....

Still playing with recipe. Had to do an erase/install, which fell over and have been reassembling my system. (a pain) I find the temp monitor history graphs really useful. I still think corruption causes overheating, since when all corrupt resources are OK, system runs much cooler.

I have experimented putting some plastic cards under the middle of the back of the MacBookPro and encouraging the body to bend up a little in the middle as opposed to drooping on the feet - hoping to put a bit of pressure on the heatsink to make the thermal goop thinner! Interestingly, some CPU sensors are now cooler than the GPU.

(in a laptop, where is is harder to ensure pressure between the heatsink and CPU/GPU, there may be some sense in having thicker thermal paste that an stretch a little. Also found interesting article on thermal goop experiments http://www.dansdata.com/goop.htm)



Observations from experimenting with fixes:
* assume corruption in system.kext or boot cache or similar that is hard to get rid of, because need a clean system to install new files cleanly - having a spare drive to install a clean system from DVD would help
(that's next, since I do have a spare drive knocking about)
* $ sudo touch /System/Library/Extensions is the recommended method for rebuilding kext caches according to developer documentation
* horizontal line artefact on login screen is worse
* horizontal line artefact on desktop appears to go away by deleting corrupt prefs and logging in again - a test of 50 logins revealed a cyclic pattern of glitches then returning to normal after several re-logins
* system often starts cleaner on auto login rather than manual login
* deleting suspect prefs, logging out and logging in helps clear up desktop horizontal line glitch
* single user boot is obvious when kexts are loading from scratch (not cache), because there are several lines of progress dots

So the latest recipe modifications:
* deep clean with applejack may be optional, but doesn't hurt, just takes a long time
* safeboot
* replace corrupt kexts using only Apple installers without deleting any - latest OS combo update mostly does it, else archive/install first
* delete corrupt prefs - last list is close but perhaps not complete - assume prefs with current modification dates could have have been messed up
* use "$ sudo touch /System/Library/Extensions" if you think you haven't got a clean boot, then reboot to permit boot/kext caches to rebuild


More later.

Latest report...

Presently assuming a horizontal line glitch on login window means corrupt kext/boot cache + corrupt prefs. A boot with a clean login screen is a good start.

In fixing the archive/install, I have enabled the root user. http://support.apple.com/kb/ht1528

The root user logs in with less frequent desktop corrupt line than normal user. Safeboot never logs in with desktop corrupt line.

Present theory looks like this:
1. Initial problem triggered by overheating - clogged fans, bad thermal paste, bad ATI driver
2. Corruption spreads though prefs files, esp those that load/write PRAM. First stage may be glitched desktop horizontal line artefact.
3. Worse corruption spreads through boot/kext caches.
4. Corruption in prefs spreads to window-drawing memory eventually causing graphics window glitches, hangs, freezes.

If corruption is present on login screen or worse problems and crashing, suggest deep clean and reapply latest system combo update.

DEEP CLEAN (first download and install AppleJack http://applejack.sourceforge.net/)
1. shutdown
2. reset SMC (on MBP: remove power and battery then hold power button for 5 seconds)
3. startup
4. zap PRAM (hold cmd-opt-P-R when booting until the system chimes a second time)
5. single user boot (cmd-S while booting)
6. # applejack AUTO shutdown
7. safeboot (hold shift while booting)

REPLACE SUSPECT RESOURCES - minimal fix (if problems persist, archive/install from MacOS DVD)
1. Run latest OSComboUpater
2. Run latest SecurityUpdate (optional if an archive/install not done)
3. Enable root user

DELETE CORRUPT PREFERENCE FILES - start here if corruption of desktop on login
1. Login as root - if login screen is clean, continue, else try this step first and reboot

2. To show hidden files to delete, in the Terminal copy/paste (I have put these commands in a shell script file since I seem to have to run it a lot):
$ defaults write com.apple.finder AppleShowAllFiles TRUE
$ killall Finder
To revert back if desired:
$ defaults write com.apple.finder AppleShowAllFiles FALSE
$ killall Finder

3. Open windows in list view and sort by reverse date (newest last):
a. {user-root-home}/Library/Preferences
b. /Library/Preferences
c. {other users} /Library/Preferences

(NOTE I presently think prefs in /SystemConfiguration are OK. loginwindow.plist may be OK, not sure)

Delete these files if present - definitely need to get rid of the hidden files:
.GlobalPreferences.plist
com.apple.loginwindow.plist
com.apple.BezelServices.plist
com.apple.systemuiserver.plist

Also in Preferecnes/ByHost
.GlobalPreferences.plist.....

maybe:
com.apple.spotlight.plist
com.apple.finder.plist

In resolving the desktop horizontal line glitch, I have been deleting prefs, logging out and logging in again. Anything with a recent modification date is suspect. When I can login as root 4 consecutive times with a clean desktop, everything seems OK. Then login as user, also repeat a few times. It appears for the prefs than are common to /Library and {users}/Library, both need to be deleted at the same time and suggest deleting /Library one first, since it may be used to replicate the user versions. Repetitive clean login suggests prefs are good.

I did see one transient glitch in text (tearing artefact) writing this, which suggests I may have left out an earlier step. More later....

After a few days without any anomalies i reinstalled MacOSX (Tiger) for testing. Running very well so far - a few hours later i installed MacOSX SL again (only updated via ComboUpdate to 10.6.4), copied some files via USB to my MBP und one distorted line appeared on my desktop again. Temps were CPU 50° and GPU 49° degrees. I reapplied new thermalpaste today (just for testing, Arctic MX2 give me up to extra 3-4 lower temperatures).

I used Apple Software Update to install the Graphics Update, restarted my MBP and the problem was solved again - for now.

The desktop horizontal line seems relatively easy to banish - see step three in my last post. While it is there, further trouble is likely to propagate through the system.

Which MBP do you have? Is running 10deg cooler than mine.

Thank you for your response, i'll try your steps again if the issues occur again. After the Apple Graphics Update still no anomalies.

I have the MacBookPro 2,2 with Intel Core2Duo at 2,16ghz, 3GB RAM and the ATI x1600 128MB graphics card. Temps with idle system (as i typed this reply) were CPU 49° and GPU 46° degrees, room temperature at appr. 23° degrees with fans at 2000 RPM. The book is as warm as my hands. (measured with "Temperature Monitor" and SMCfancontrol). While speeding up fans to 4000 RPM i get CPU 43° and GPU 41°.

Message was edited by: vinylger (typo)

Still no anomalies. Temps are as cool as already posted above, temp with World of Warcraft running (fullscreen with higher graphics settings) CPU 53° and GPU 43° degrees. Both maxed at 59° for CPU and 49° degrees for GPU (both +/- 2 degrees).

My best temperatures after reapplying new thermal paste (arctic mx2) and without any problems so far. I thought it would be a temperature related issue but as i stated above i got the problem before installing the Snow Leopard Graphics Update yesterday after reinstalling the OS (while temps weren't higher as they are now).

found a site listing temps, but no machines like mine running 10.5.8:
http://www.intelmactemp.com/list

So far have gotten nearly 3 days glitch free running low temps without thermal mods (but still trying to figure out exactly what i did that banished the problem). Once corruption has started, it will spread at any temp.

This as as near as I can find to a possible answer - see post by mmcwilliam:
http://getsatisfaction.com/apple/topics/2006imac_graphics_card_crashfreeze

"Here were the updates from Apple Engineers:

July 1, 2010
We have tested the GPUs in both our system level test and in our socket cards and the parts are good. Both parts pass our tests at 25 C and 105 C.

We are highly confident that the GPUs are good and that this is a memory issue.

Sept 8, 2010
We are waiting on update from the video memory vendor about failure analysis.

Sept 22, 2010
The failure analysis indicated that the problems was in graphics memory, not RAM. There were pins with shorts, and these were found to be damaged by heat due to high current from unknown sources."

I have logged a bug with Apple re: software problem with ATIRadeonX1000.kext. Apple engineering asked me to test 10.6.4 as well and report back.

Interestingly, fresh install of 10.6.4 safebooted runs 100% CPU at no higher than 65degC on my machine - 15 degrees cooler! Normal boot and temperatures climb into the 80s again under the same load (Safari and a few flash windows + $yes > /dev/null ).

As you all know, the problem is present in 10.6.x as well as 10.5.x. Latest upgrade 10.6.5 made no difference to desktop artefact.

So for now I have deleted ATIRadeonX1000.kext. (after usual recipe to clean system+prefs + $ sudo touch /System/Library/Extensions) Xbench happily runs graphics benchmarks but OpenGL does not perform properly.

For the curious. Temps before and after safeboot 10.6.4 (the gap in log for normal bootup) , same CPU loading method - few flash web pages + $ yes > /dev/null. So it looks some software changes have improved efficiency and therefore temperature in 10.6.x, but not fixed our graphics glitching problem.

http://i197.photobucket.com/albums/aa145/cheongi/Screenshot2010-11-11at114642AMp re-postsafetbootsameload100CPU.png

So this is my solution... Take out the battery... and put the open cavity over the fan of a cooling pad... Turn your fans all the up using SMC fan control.. and WA LA.... you can run garageband and imovie... and whatever else without a crash!

Anyone wanting to stress test GPU could try:
Furmark via Windows/Bootcamp
http://www.ozone3d.net/benchmarks/fur/

CineBench
http://www.maxon.net/downloads/cinebench/cinebench-115.html

Yes, that is a solution, but not a complete solution. What happens is that when you take out the battery and run the computer only from the power cords, all the system components (cpu & gpu) will underclock themselves by about 50%. This is a built in feature by apple because what tends to happen sometimes when running very intense computations is that the system will need more power than the cord can provide so it takes it from the battery.

Back again, i could reproduce these anomalies by installing Windows 7 (without BootCamp, just booted from Windows 7 Install DVD). The fans begin to start very late, the case of my MBP tend to be very warm. As i read from Ian Cheongs "stress test" links i started World of Warcraft with Windows 7 again and my machine gets hot. After dragging Windows around the lines appeared again. So i found myself searching i tool for windows like SMCfancontrol. I found something like "remapper" which is able to control the fanspeed as well. So i turned up fan speed up to 6000 RPM and everything went fine again, still with the Game running at its highest (possible) settings.

But: when i start windows 7 some colored flickering pixels appear on the screen. Imagine a triangle in the middle of the black screen, every "corner" is a flickering pixel - I have to go very close to my display to notice that. After the orbs flying into the screen to build up the Windows Logo with "Starting Windows" everything is ok. Any Ideas?

Not sure that Windows 7 is meant to work without boot camp ... since PCs have different BIOS
See http://support.apple.com/kb/HT3986