This is hysterical! (I see you people all enjoy beta-testing Apple's hardware for them.)
I just read every post and not even a rudimentary attempt at debugging?
No one's installed a Kernel Debug Kit from the Developer site?
No checking for crash dumps?
Not even a mention of remote debugging techniques?
Someone has to be UNIX®-savvy around this place.
Let your afflicted MacBook Pro put its display to sleep.
(Yes - I mean as in, wait for the screen saver/lock to kick in and then the display go to black/blank from the display sleep setting.)
SSH in from another Mac. Make the Terminal window on the other Mac be really wide, like 220 columns wide.
Open 2 other Terminal windows on the other Mac and SSH in to the afflicted MacBook Pro in those 2 windows as well, so you have three total separate logins on the afflicted MacBook Pro.
Become root in the really wide window on the afflicted machine via
sudo su -
giving your login password to "
sudo".
Beat the H3ll out of the machine by running
nice -n -20 top -i 1 -o cpu -s 0
in that big wide window.
Then in one of the other (regular size) SSH windows, do another "
sudo su - " to become root.
In that window, run
vm_stat 1
Then, last but not least, in the last (regular size) SSH window, run the multiple-instances of the "
yes > /dev/null" test.
We've taken the GPU somewhat out of the equation because we're not doing anything on the screen.
Meanwhile we're hammering it with CPU-intensive tasks.
If it
does freeze up eventually, in the "
top" window you should have an exact snapshot of the system at the moment it disappeared beneath the event horizon of the Black Hole. You should also be able to see the VM statistics (in case they're useful, like free memory, page faults and pageins/pageouts) when it locks up solid as well.
--
Conversely, if there are kernel panics (eventually) or any sort of a crash that leaves a crash dump, look in Console to see if there was a crash dump and see if there's a stack traceback. If there is, and it's a repeatable crash, consider getting a Kernel Debug Kit from the (free) Developer site and drop in the kernel from that Kit to your afflicted MacBook Pro and reboot it.
Then, if you can provoke it again and get another crash dump to happen, if it's in the identical place this time the stack traceback will contain symbolic addresses instead of useless Hex gobbledegook addresses. Give that information to Apple.
Happy debugging 😀