Channel exception error with NVIDIA card: It’s back

Posted by Pierre Igot in: Macintosh
August 19th, 2009 • 9:53 am

After many weeks without a single occurrence, I am afraid I have to report that the problem with the “channel exception” error linked to the NVIDIA graphics card in my Mac Pro appears to be “back.”

Yesterday, I was working with a number of applications open and I selected a couple of PDF files in the Finder to open them in Preview. I typed command-O, and Mac OS X switched to Preview, but then everything froze, with the Spinning Beach Ball of Death.

There was no way out of it. The mouse pointer was still moving, but force-quitting was not accessible. I couldn’t switch to anything or launch anything. I was stuck again.

I tried logging in remotely from my other Mac Pro, but my local area network is dynamic and I didn’t have the exact local IP address of the machine, and at that particular point I didn’t really have much time to experiment anyway, so I did a hard reset. Now everything is back to normal, but for how long?

This episode suggests that the problem probably never really went away, but was simply under control for whatever reason. Given that the past few days have been particularly warm in my region, for the first time of the year, it might be a confirmation that the problem has an environmental component.

You might remember that, in my case, the problem appeared to be “cured” after I physically removed the graphics card from the machine, cleaned everything, and reseated the card properly. At the time, I suspected that the accumulation of dust (although there wasn’t much of it) and overheating might explain the seemingly random and unpredictable nature of the occurrences of the problem.

For the record, I am now running a system with the latest official system release, i.e. Mac OS X 10.5.8 (applied as a combo update a couple of weeks ago).

A Betalogue reader who is experiencing the same “channel exception” errors, but with much more frequency, has also e-mailed to confirm that 10.5.8 didn’t fix the problem on his machine. He typically experiences the problem when running a high-end pro application such as Maya, because it’s the “worst offender” (likely because it makes heavy use of the graphics card) but also experiences it with other, simpler applications, such as Safari, Preview, etc.

He too has been in contact with Apple via Bug Reporter, but has had little success. He has been sending them the logs produced using the stackshot command, as per the instructions that Apple gave me, but has received zero response from Apple, which is quite discouraging.

In my case, after responding a few times to my bug reports and giving me instructions for the stackshot command, Apple ended up writing:

Moving to verify as we am afraid not much NVIDIA can do to address this without repro.

and then promptly switched the status of my bug report to “Closed.”

In light of this latest incident, I have written to Apple again to ask that the bug report be reopened, so that I can submit more information. However, I also told them that when I try to run

sudo /usr/libexec/stackshot -it

in Terminal, as per their instructions (but on my Mac Pro running normally), I seem to get error messages instead of the expected results (i.e. logs):

Mac-Pro:~ igot$ sudo /usr/libexec/stackshot -it
/usr/sbin/symstacks.rb:53:in `readline': end of file reached (EOFError)
	from /usr/sbin/symstacks.rb:53:in `ksymbolicate'
	from /usr/sbin/symstacks.rb:233:in `readLoop'
	from /usr/sbin/symstacks.rb:301:in `main'
	from /usr/sbin/symstacks.rb:307

This does not look like a command running properly to me. And I certainly cannot find any logs produced by this under /Library/Logs. So I’ve asked them for more information. (They tend to assume that everyone is as proficient in using Terminal as they are.)

Still, I hope that the problem does not recur. But if it does, I want to be able to capture as much information as possible, so that I can provide Apple/NVIDIA with what they need to address the problem once and for all.

There is also a vague hope that Snow Leopard will somehow eliminate the problem altogether. I am not convinced myself, since, based on what Apple is saying, the NVIDIA drivers for Mac systems are developed by NVIDIA and not by Apple’s own engineers. But who knows? We are certainly in the realm of difficult, hard-to-reproduce problems and this particular saga could go one for a while still.

Comments are closed.