b3rs3rk

Members
  • Posts

    207
  • Joined

  • Last visited

  • Days Won

    1

Everything posted by b3rs3rk

  1. @antihero412 Yeah, this is a bug. Your Kepler Quadro doesn't support PCI RX/TX utilization and I'm not checking for it to be numeric before I try math functions against it. I'll push a fix. Thanks for the report.
  2. Can anyone using Emby and/or Jellyfin tell me what processes show up at the bottom of nvidia-smi -q -x when transcoding? For plex it is the 'Plex Transcoder' but I'm not sure what it is for the others. I am expanding on the plugin offering for Nvidia to show little icons of what apps are using the GPU hardware.
  3. You need to provide troubleshooting information as seen in my original post before we can attempt to help you.
  4. @ich777 and I think this should be fixed with the latest update. Let me know if it still doesn't work. Thanks!
  5. If you have the default options set it may not be saved to your config file. Try setting the Intel options to No and hit Apply to make sure it is saved. You can't save the defaults which is something I'm looking into. If that doesn't fix anything you need to give me the troubleshooting info from the first post of this thread.
  6. In your case, no. Once you use passthrough nvidia-smi will likely not see it anymore. Just test it. If after your gaming VM is started you can still poll stats by running nvidia-smi in the UnRAID console then I'm mistaken. In the general sense of multi-gpu I've considered doing a drop-down box like the NIC/network widget where I populate all graphics cards there so it can be dynamically selected. This would take quite a bit of work though and I'm not even sure it's possible because all the page code would have to be replaced for different vendor support. I dunno.
  7. Make some change in the GPUStat settings and apply it. The plugin will not load automatically anymore due to too many issues with people not changing the settings.
  8. You need to make some change (any change) on the settings page and apply it. You can then revert that change if you want the defaults. The new default is to have no vendor in your config file so that people don't install the plugin and never change any settings which are necessary for the Intel crowd. You have to change something for the apply button to show.
  9. I haven't seen anyone using it during transcoding operations. From Intel:
  10. @muwahhid Yeah, looks like I'm capturing everything useful from that so far. I have only seen less metrics than yours, but never more. Maybe Intel will expand this in the future.
  11. I've seen results from a UHD 630 I believe, but since you are the first to say you have a Xeon package instead of a desktop processor it can't hurt to look!
  12. iGPU support is now merged to master and plugin is updated. Thanks to everyone who tried it on the test branch. You can now revert to the CA version by removing my plugin altogether and reinstalling from CA for cleanliness.
  13. @Daniel Tildeman Your GPU is rather old and based on the nvidia-smi output you pasted you're correctly seeing the few metrics your card provides. For that card, you should only be monitoring memory utilization, fan usage, temperature and power state in the plugin. This should limit the amount of N/A's shown on the screen to a bare minimum and clean up the plugin a bit. Again, the plugin doesn't do miracle work so you're going to have to update your card to get better monitoring support. Sorry.
  14. New test plugin version available. Running Check for Updates on the plugin page should show it available for install. This will only show for the people that are currently using the iGPU test branch I posted before. Let me know how it goes. Thanks!
  15. Hmm. I'm thinking rc6 is some kind of reverse power state. Like the higher the percentage, the less power you're using. Might be something I can use for a power-based slider bar only using the inverse of it to plot.
  16. You have a very modern iGPU, could you send me the results of: timeout -k .500 .400 intel_gpu_top -J -s 250 I'd like to see what metrics your iGPU supports as most of my investigation includes older iGPUs. Thanks.
  17. I'm sorry, I'm trying to think of how to troubleshoot it since I can't reproduce it on either my production box (Dual Xeon) or my tester box (Atom D515). Both of them run the new code (and the new new code with Intel support) just fine with no spikes. You're not on the latest UnRAID version by the way. I'd install the newest release candidate rather than sit on a Beta version. There have been two release candidates delivered since the Beta version you are running was released. This plugin tests fine on 6.8.3 and 6.9.0-rc2.
  18. Did you select Intel in the settings page and apply it? You're getting the NVIDIA output on the dashboard.
  19. @smidley @madelectron any anyone else wanting to test Intel iGPU support: First install the Intel GPU Top plugin from @ich777 -- This is NOT the docker version on Community Apps. It has to be the plugin version by ich777. Then uninstall my current plugin and manually install this one - https://raw.githubusercontent.com/b3rs3rk/gpustat-unraid/dev-intel-test/gpustat.plg Make sure you go to settings and set it for Intel. If you have a valid iGPU, it should show up on the settings page. If not, message me the output of the following command run from the UnRAID console: lspci | grep VGA If your iGPU is detected, but you're not getting any data on the dashboard, go to the UnRAID console and type/copy paste the following and send me the result: cd /usr/local/emhttp/plugins/gpustat php ./gpustatus.php If you are getting some, but not all the data you'd like to see review the UnRAID console output of: intel_gpu_top If the metric I display on the dashboard is not showing up for you in intel_gpu_top your CPU/iGPU/Chipset doesn't support monitoring that. Recommend disabling that particular metric in your plugin settings. If the metric shows up in intel_gpu_top but still doesn't display on the dashboard, send me the output of: timeout -k .500 .400 intel_gpu_top -J -s 250 To revert back to my existing plugin, just uninstall the test version manually and reinstall from Community Apps. Enjoy!
  20. Leave the plugin un-installed and in the UnRAID console run: watch -n 1 nvidia-smi -q -x -i 0 Leave it running in the console and watch top in another console. If you're still getting CPU spikes, then it isn't my plugin.
  21. 2021-01-08 now released. Should fix update issues with UnRAID 6.9. Thanks @mlapaglia for the contribution.
  22. If you simply installed the graphics card and booted up, you're probably missing the step where you need to install the UnRAID-Nvidia plugin, install an nvidia kernel and reboot. Then you should have access to the nvidia-smi command from an UnRAID console. My plugin depends on that command to be available.
  23. I have a branch for it already. Once I've merged the changes from @mlapaglia into mine I can look at setting up a dev test for people like you who are willing to help out. Keep in mind, the Intel stats are starkly different compared to what Nvidia gives using their built-in tools.
  24. It looks great. I’ve got no front end Dev experience to speak of and I will excitedly review the PR work you have done when I return home from vacation. I have some other stuff I was working on that I will have to merge in but I’ll handle that later on. Thanks.