Q: NVDA(OpenGL): Channel exception!

Any news? This is not a minor issue.

Does the 10.6.4 update fix this problem? There's nothing specific in the release notes, but I've read that the version number of the driver changed.

Same deal. Just want to add another unsatisfied guy to the growing list.

Also having crashes like this:

NVRM-RC: 0 4e565f504649464f5f494e54525f303a307831303030
NVRM-RC: 0 4e565f504649464f5f4341434845315f444d415f53544154453a307838303030
NVRM-RC: 1 363232300a
NVDA(OpenGL): Channel exception! exception type = 0x6 = Fifo: Parse Error
00000069

I experience them while developing OpenGL applications. The same programs run fine on Debian Linux machines, using NVIDIA hardware and drivers (mostly GTX285 or better cards). On the Mac I am working on a MBP5.1, NVIDIA 9400 or 9600 GPU, OS X 10.6.4, latest XCode.

This started after combo upgrade to 10.6.4 on a 3 GHz iMac with NVIDIA GeForce 8800 GS:... and happens when playing runescape on www.runescape.com after a few minutes


Jun 19 18:08:23 iMac kernel[0]: NVRM-RC: 0 4e565f504649464f5f494e54525f303a307831303030
Jun 19 18:08:23 iMac kernel[0]: NVRM-RC: 0 4e565f504649464f5f4341434845315f444d415f53544154453a307838303030
Jun 19 18:08:23 iMac kernel[0]: NVRM-RC: 1 363232300a
Jun 19 18:08:23 iMac kernel[0]: NVDA(OpenGL): Channel exception! exception type = 0x6 = Fifo: Parse Error
Jun 19 18:08:23 iMac kernel[0]: 00000069

hi folks...sorry if this has been covered but:

is this limited to imacs? my brand new mbp i7 has been crashing non-stop. deciphering the various log messages has been a bit challenging but sometimes it has definitely had the channel exception error. i seem to have other symptoms too...sound repeating and usually a frozen screen (not black).

i'd write more but i don't want to lose the reply before another crash!

Message was edited by: jpodtbc

If your new i7 has been crashing all the time why haven't you just taken in it to be replaced???
Or call Apple if your not close to a store.

i did. i gave them all the detailed information i had at the time and the dufus replaced the hard-drive. obviously it's still crashing.

a new logic board is on order but i'm trying to do a better job troubleshooting the precise problem than most lame techs. i guess i'm trying to find out if these issues will plague me or not.

there's some drivers on the nvidia site that seem to have solved my window server crashes.
http://www.nvidia.com/object/geforce-macosx-19.5.8f03-driver.html. a german site claims they restore 10.6.3 opengl framerates in portal, ( http://www.fscklog.com/2010/06/sammelsurium-nvidia-veröffentlicht-treiber-für-10 64-faq-zu-hdmi-am-mac-mini-weitere-patentklage-gegen-htc-etc.html ) but i'm just happy with the stability. they seem to be labled as the same versions that 10.6.4 instals but are different. they'll only install on a macpro so everyone else will have to use "pacifist". these are straight off the nvidia site, but back up your stuff first. they worked wonders for me, i've been crashing out of photoshop for months.

I just started receiving this error today.

My screen briefly flashes then everything except mouse pointer movement halts. The first two times it resumed after less than a minute. The third time the screen started to flash rapidly and wouldn't stop; I powered off by holding the power button down.

The first two times it was when running a bunch of stuff including Skype, Mail, Firefox 3.6.6, and VMWareFusion 3.1.0 (which was running Windows 7, and several development tools including VS 2010). The third time I was just running Firefox and watching a video on Youtube.

I'm running 10.6.4 on a 15" Macbook Pro 2.2ghz Intel Core 2 Duo with the Santa Rosa chipset (I think this is from the summer of 2008). I also have an external Samsung SyncMaster 2443BW attached to this machine, if that makes any difference.

Coincidentally I had this error happen in the middle of writing this message. If someone from Apple is paying attention to this thread, contact me and I can provide you with a text file containing the relevant entries from my console log file (it's 1100 lines long and was generated over the space of about 2-3 minutes).

edit: I'd like to add that this was only the fourth time the problem has occurred and that the system recovered after going crazy for a few moments. At one point the mouse pointer input was all screwed up (inverted) and this is reflected in the portion of the log file that I captured.

Message was edited by: C.Batt

steepleton wrote:
they'll only install on a macpro so everyone else will have to use "pacifist".

Note that in the "Additional Information" tab at http://www.nvidia.com/object/geforce-macosx-19.5.8f03-driver.html it says that the minimum requirements include:

• System Interface
° Dual slot with one (1) open PCI Express x16 slot
° 2 6-Pin PCI Express Power connector cables (included with graphics card in retail package)

Also, the "Supported Products" tab mentions only the GeForce GTX 285.

I'm only guessing, but it seems unlikely that this will work with anything other than Mac Pro 3,1 or 4,1, or later models.

true, but it actually contains the full suite of kernel extensions, libraries and drivers for all the nvidia cards. I'm using a gt120. as i said, no guarantees, but it solved the problems for me with opengl exceptions and window crashes and resulting freezes and logouts that have been plaguing me and others since 10.6.3
i'd think certainly worth a punt if your mac is going down twice a day.

steepleton wrote:
i'd think certainly worth a punt if your mac is going down twice a day.

Maybe yes, maybe no. From what I can tell the update contains slightly different versions of just about everything compared to what Apple installs. This includes the Hal kext's which I believe are the hardware abstraction layer components as well as the main GeForce.kext & the plugins. (You can see this by drilling down into the package contents & comparing the info.plist files.)

It may be that these versions are meant for the full-size cards' dual slot PCI Express x16 & heavy duty power supply configurations (& the cooling capacity that goes with that), which would presumably be no problem in a Mac Pro but might overtax lesser iMac & MBP systems.

Also keep in mind that using Pacifist to extract & install files is not exactly equivalent to what Apple's Installer does. Pacifist is primarily designed to reinstall damaged components, not to defeat the checks the Apple installer does that may prevent users from installing something not intended for their Macs or to bypass the system configuration steps it does.

If it were me, I would try this on anything other than a Mac Pro only with great caution: first, I would clone the current startup drive so I could easily revert to the original drivers if anything went badly wrong, & then I would install something like iStat Pro & carefully monitor GPU temperatures during heavy duty graphics operations until I was satisfied that the GPU or anything else was not running dangerously hot.

i agree, the precautions you list are wise. but, and correct me on this of course, the drivers have definitions for all the editions of the cards, and'll be able to identify what's installed, so the risk would be theoretical, and be jumped on by a thermal sensor induced shut down anyway?
using pacifist for this is a grubby hack, i agree.

I'm not so sure it would be obvious what is installed in these very large files without access to the source code. I've had some limited success analysing the differences in some other files using the FileMerge.app provided in the Developer Utilities, but it is really intended for text comparison & merges, & it chokes on large UNIX executable files like these. So no help here from that.

But one thing I did notice in comparing the Apple-supplied & NVIDIA-supplied packages is that the NVIDIA-supplied executables were sometimes substantially smaller than the Apple-supplied versions. There is no way for me to know for sure but this could mean that important or necessary pieces of code for keeping the iMac & MBP versions of the GPU's happy are missing from the NVIDIA release.

Basically, hacks are hacks, grubby or not, & any undesirable consequences of using them falls on the user. As long as people understand & accept that risk, I guess it is OK to try it.