Dashboard Improvement 1: GPU Widget

[dszidi]

Greetings community members! We have had a lot of feedback from users requesting a GPU widget. We've heard you loud and clear, and would like to present to you some wireframes.

Wireframes Disclaimer:
Designs in blue are wireframes that are meant to be low fidelity sketches. They will not be pixel perfect and may not be as visually precise as the final product in the presentation. It may also make use of placeholders for things like images and graphs. Please focus on the content and treat them as drafts.

About the Designs

The following are designs for what the widget will contain for Nvidia and AMD graphics cards. Due to some backend limitations there will be no Intel support at first until some dependencies get support for those devices. Also due to dependencies being used, not all the same stats are available between Nvidia and AMD. With that said, please take a look at the designs and give us your feedback.

Thanks!

[GMAzrael]

Looks excellent to me!

[xtj7]

A few thoughts:

• I assume with multiple GPUs we simply have one widget per GPU, even if these are multiple Nvidia GPUs for example?
• Has there been thought about a mobile viewport? I realise that probably isn't the highest priority, but checking stats on mobile for troubleshooting can be quite helpful. So especially the dashboards should work well with mobile too.
• Historical data would be a nice touch as well. Maybe configurable with a drop in resolution, similarly to how homeassistant is handling historical data.
• Custom alerts set for reaching a certain GPU usage, VRAM usage or encoder/decoder usage - if my CCTV isn't recording properly because of decoding/encoding limitations, I want to know :) And this could be visualised in the UI as well (for example: make the metric red once it is past the custom alert, so you can see at a single glance where you are above the threshold)
• Custom label or description per GPU: I have one GPU dedicated for my NVR and one GPU for everything else. Knowing which one's stats I'm looking at would be very helpful

[dszidi]

I assume with multiple GPUs we simply have one widget per GPU, even if these are multiple Nvidia GPUs for example?

Correct. If you have three Nvidia cards on your system you will have three GPU widgets on the dashboard

Historical data would be a nice touch as well. Maybe configurable with a drop in resolution, similarly to how homeassistant is handling historical data.

We treat the dashboard as the place to get information about the current state of the system, and the reporting page as the place to find historical charts and graphs. Typically we keep an icon button on most of our dashboard widgets linking to the relevant reporting page. You can expect to see that on these GPU widgets as well. I will update the designs to reflect that.

Custom alerts set for reaching a certain GPU usage, VRAM usage or encoder/decoder usage - if my CCTV isn't recording properly because of decoding/encoding limitations, I want to know :) And this could be visualised in the UI as well (for example: make the metric red once it is past the custom alert, so you can see at a single glance where you are above the threshold)

This sounds interesting. We'll take it under consideration.

Custom label or description per GPU: I have one GPU dedicated for my NVR and one GPU for everything else. Knowing which one's stats I'm looking at would be very helpful

This is a great suggestion. I'll investigate how we can incorporate that into our existing designs and update the initial post.

[Sophist-UK]

Nice.

[ElectronicBattle]

This is an exciting development for me, seeing as (1) I've just fitted a graphics card for use in TN Scale official apps and (2) I'm very much a beginner, having been running TN Scale for only a couple of years as a hobbyist with no programming background so frequently out of my depth!

[panzerscope]

For someone who is looking for a GPU widget on TN Scale, this looks encouraging, although one would have to decide what “GPU Usage” is actually monitoring on the widget. Would it take into account overall usage, or just that of the 3D Engine, or Media Engine for things like transcoding?

The widgets as designed on the old topic you linked look pretty, but IMHO not granular enough. For example The Intel_GPU_Top (for Intel GPU's) gives you a good breakdown of what part of the GPU is being used and by what process, such as the below:

I know right now that Intel GPU support is out of scope due to other dependency requirements (hope you guys can solve that) but I would think that having more granular visibility for Nvidia and AMD GPU's would also be helpful.

P.S The last comment on this was July, has there been any progress on this in general?

[boimouseorange]

[Sophist-UK]

This seems to be a good thread for me to jump on due to my current issue.

No - this is NOT a good thread for you to jump on.

1. It is (generally speaking) considered inappropriate to hi-jack one thread to discuss something different.
2. This is a discussion about something completely different, so not many people will latch on to your issue.
3. The correct place for this sort of question is in the TrueNAS forums.