"ERROR - Get temperature failed for Adapter=1"

JabulonJabulon Member Posts: 45
Haven't been able to overcome this. Same problem on three different installations: amdconfig --odgt --adapter=all only returns temp for adapter=0, but gives the above error for any and all other GPUs. All cards are mining, lspci |grep VGA correctly identifies all cards. I'm on Ubuntu 14.04, on fresh install of latest AMD crimson drivers with no other apparent issues. Have spent hours googling. Checked everything I could find on this error, but found no coherent diagnosis, let alone a solution. Any ideas? I'm worried about not being able to check temps on all cards.

Answers

  • dont12dont12 Member Posts: 60
    look up atitweak
  • JabulonJabulon Member Posts: 45
    Nope. I've messed around with some tweaking and sensor utilities and always hit the same wall. Tried making some edits (as sudo) to my xorg.conf as well, as I suspect the culprit is there. But saving changes doesn't not make them stick. On reboot, it's back to what it was before editing. Clearly out of my depth here.
  • dont12dont12 Member Posts: 60
    no search for a program called atitweak
  • JabulonJabulon Member Posts: 45
    Ok, finally solved my issue. Long story short, if you encounter the same problem, uninstall your catalyst drivers and reinstall using the automatic install via GUI method. I am used to building the packages myself, but in this case the GUI installer gets a superior result. Still have no idea where the flaw was.
  • peepeedogpeepeedog Member Posts: 32
    Jabulon said:

    Ok, finally solved my issue. Long story short, if you encounter the same problem, uninstall your catalyst drivers and reinstall using the automatic install via GUI method. I am used to building the packages myself, but in this case the GUI installer gets a superior result. Still have no idea where the flaw was.

    Jabulon, I had the same issue for my second GPU. I couldn't pull any GPU info from it. However after rebooting it and re-initializing, it seems to work fine. I didn't touch anything else.

    I run Ubuntu 15.04 with a 7970 and a 280X.
  • JabulonJabulon Member Posts: 45
    Lucky for you if it was that simple. Didn't work out that way for me. Tried all kinds of approaches. In the end, removing and re-installing the drivers in a different manner was the only thing that helped. Now --od can successfully query all GPU's. **addendum: Playing around with AMDOverdriveCtrl, interestingly, it still only shows GPU 0 as active. Also, it does not work at all on Ubuntu 15.10. (14.04, yes). Mismatched libraries or something I guess. Perhaps not too surprising, seems like it hasn't been actively maintained since 2013. Too bad, it was a nice little utility back then.
  • malefactormalefactor Member Posts: 12
    Jabulon - any further detail or insight as to how you made this work? I can't tell you how many times I've reinstalled from scratch, both 14.04 desktop & 15.04 desktop, trying different permutations to find a way to make this work.

    15.04 seems more broken, where even "sudo aticonfig --initial" requires "-f" to save at all, and on the next reboot appears to have been overwritten/ignored.

    I'm wondering if it was some weird sequence, such as: install, install catalyst from AMD from command line, GUI "Additional Drivers" that was needed to get it all working.

    Or perhaps it cares what monitor is attached? I'm amazed this is so difficult.
  • JabulonJabulon Member Posts: 45
    edited January 2016
    Hey Malefactor - yeah, I feel your pain. This drove me nuts for over a week, and nothing I googled or asked on any of the forums helped. Like you, I also tried manually adding devices to xorg.conf and found my changes overwritten on reboot. Clearly there is a script operating outside your control, probably there as a safeguard. Xorg is an antiquated mess, and one of the easiest things people break. If you want to try making your changes stick, you can use sudo chattr +i /path/to/your/config before you reboot, which makes your changes immutable. Not even you as root can alter them without first removing that flag. However, there's a chance you will completely screw yourself, so you might want to make a backup image of your drive before trying this.

    In the end, I solved the issue another way, by doing a full remove/purge of the fglrx drivers, followed by reinstall using the GUI amd installer (automatic install). This worked on both Ubuntu 14.04 and 15.10, with the latest driver release (in my case, for r9 200 series gpus). No more error, I get temps on all devices by way of aticonfig --odgt --adapter=all

    Of course, after installing the drivers and before rebooting, create your new xorg.conf with sudo aticonfig --initial -f --adapter=all

    As far as the sequence for installing the drivers, just download the appropriate zip file from amd, right click>extract here, open the folder, right-click the .run file, choose properties>permissions, make sure the option to allow running executable file as a program is selected. Then go to the panel menu, edit>preferences>behavior tab and make sure 'run executable text files when the are opened' is selected. Go back to the .run and run it. You may not see the graphical environment appear immediately. Wait for it. You'll get a graphical popup asking for your sudo password. Oh, and before you do any of this, I'm assuming you have gksu. If not, install it, then follow the steps above.

    If it all works and you get to 'installation complete' be sure to click to see log. It will tell you if there were missing dependencies.

    **EDIT: as I think I mentioned elsewhere, my original driver install was manual, building the packages myself. This worked, except for the annoying error when querying gpu sensors for temps.
  • malefactormalefactor Member Posts: 12
    edited January 2016
    so, no sooner do I post that ... than I find myself in the same situation as you, Jabulon. Hah!

    I got it working, and I have absolutely no idea why. There is no rhyme or reason. And on this particular system it could have been absolutely anything, as I have uninstalled and reinstalled probably ~12 times.

    edit: your post appeared before I clicked "post". Thanks for the insight, maybe someone will come along and figure out the precise surgical fix. I for one am just going to image this system and never, ever change it.

    I suppose I'll be forced to using windows if it ever comes down to precise tuning with catalyst versions. That wasn't my first choice because it's a pain to manage nodes remotely like that (no ssh)
  • blitzcoolblitzcool Member Posts: 23
    Jabulon said:

    Hey Malefactor - yeah, I feel your pain. This drove me nuts for over a week, and nothing I googled or asked on any of the forums helped. Like you, I also tried manually adding devices to xorg.conf and found my changes overwritten on reboot. Clearly there is a script operating outside your control, probably there as a safeguard. Xorg is an antiquated mess, and one of the easiest things people break. If you want to try making your changes stick, you can use sudo chattr +i /path/to/your/config before you reboot, which makes your changes immutable. Not even you as root can alter them without first removing that flag. However, there's a chance you will completely screw yourself, so you might want to make a backup image of your drive before trying this.

    In the end, I solved the issue another way, by doing a full remove/purge of the fglrx drivers, followed by reinstall using the GUI amd installer (automatic install). This worked on both Ubuntu 14.04 and 15.10, with the latest driver release (in my case, for r9 200 series gpus). No more error, I get temps on all devices by way of aticonfig --odgt --adapter=all

    Of course, after installing the drivers and before rebooting, create your new xorg.conf with sudo aticonfig --initial -f --adapter=all

    As far as the sequence for installing the drivers, just download the appropriate zip file from amd, right click>extract here, open the folder, right-click the .run file, choose properties>permissions, make sure the option to allow running executable file as a program is selected. Then go to the panel menu, edit>preferences>behavior tab and make sure 'run executable text files when the are opened' is selected. Go back to the .run and run it. You may not see the graphical environment appear immediately. Wait for it. You'll get a graphical popup asking for your sudo password. Oh, and before you do any of this, I'm assuming you have gksu. If not, install it, then follow the steps above.

    If it all works and you get to 'installation complete' be sure to click to see log. It will tell you if there were missing dependencies.

    **EDIT: as I think I mentioned elsewhere, my original driver install was manual, building the packages myself. This worked, except for the annoying error when querying gpu sensors for temps.

    can you please provide the link of the amd gui driver that you mentioned here.. thanks
  • blitzcoolblitzcool Member Posts: 23
    the reason why I'm asking is I can only download the deb packages from their website..
  • blaBLAblaBLA Member Posts: 10
    the problem of xorg.con getting overwritten is probably caused by https://bugs.launchpad.net/ubuntu/+source/ubuntu-drivers-common/+bug/1310489
    solution in the comments, "gpu-manager is responsible for overwriting your X config file. Commenting out the "start on" lines at the top of gpu-manager's upstart script located at /etc/init/gpu-manager.conf will temporarily disable gpu-manager and prevent xorg.conf from being changed."
  • tbuitendyktbuitendyk Member Posts: 1
    blaBLA said:

    the problem of xorg.con getting overwritten is probably caused by https://bugs.launchpad.net/ubuntu/+source/ubuntu-drivers-common/+bug/1310489
    solution in the comments, "gpu-manager is responsible for overwriting your X config file. Commenting out the "start on" lines at the top of gpu-manager's upstart script located at /etc/init/gpu-manager.conf will temporarily disable gpu-manager and prevent xorg.conf from being changed."

    Thank you very much blaBLA for this tip. I wasted a day or more messing around downloading / building / installing / removing various ATI driver packages on a previous Ubuntu rig (and reinstalling corrupted Ubuntu installations) before finding this comment of yours. THIS FIXED MY ISSUES IMMEDIATELY ON A NEW RIG WITHOUT ANY OF THAT MESSING AROUND. CARDS ARE DETECTED AND FULLY CONFIGURABLE NOW.

    (Signed-up here just to pass on this comment -- thx again.)

    t.
Sign In or Register to comment.