Have a question, we have been working on an experience for a client and there are several contributors. We can all run the experience locally on windows but when remote-ing into the client’s machine and starting the experience (using the windows app for remote desktop unfortunately - client insists), touch designer starts to use 100% of the cpu and hangs the computer. We will be going in to the site physically to see what is going on, but wondering if there are any other techniques to see why we are seeing this behaviour on the remote machine and not locally?
Hi,
There is very little that can be done here and without project and access to the machine we are working blindly.
Do you have the complete specs? What is the GPU usage under load?
It is very common to see performance drops when using those remote sharing softwares, especially builtin or older techs. My preference is in Parsec which is performing well most of the time.
That being said, if the client is not flexible, your only solution is to make sure that this is not the actual issue.
What is the version of TouchDesigner you are using ? Additionally, follow these steps and see if you didn’t cover something already:
Make sure that all your drivers, including integrated graphic drivers, are up to date and that your machine is free of bloatware.
Make sure that TouchDesigner is fully up to date to the latest stable build available.
The Bulk Crap Uninstaller helps in identifying unwanted or unneeded software: https://www.bcuninstaller.com/
Find the latest driver for your GPU at Official Drivers | NVIDIA
or use Nvidia Geforce Experience Update Drivers & Optimal Playable SettingsYou can also make sure that your integrated GPU is running on the latest version available via Intel Driver Assistant.
You can use the AMD Adrenalin if you have an AMD Integrated GPU, in minimal install mode, via https://www.amd.com/en/products/software/adrenalin.html
If your machine comes with a proprietary power management software (From Asus, Dell / Alienware… etc), make sure that it’s setup properly to run TouchDesigner on the Nvidia / AMD DEDICATED GPU when plugged to power. Otherwise, it’s likely that TouchDesigner would run on the integrated graphics, hence it is important to also make sure your integrated graphics are up to date.
If the issue persists, try to generate a dump using Using WinDbg to Debug Crashes - Derivative
1 Like
I second Parsec here. It’s great for installations and remote monitoring. Highly recommended.
I’d also double check your Windows Power Management settings. Make sure you’re set to High Performance. We’ve noticed that if it isn’t, a lot of power is sacrificed in the name of energy efficiency.
OK, we were able to pin this down a little more… we confirmed latest drivers etc, we were loading a ton of .svg’s at startup and noticed on our local test machine (3090 intel i7) things where heavy but OK. Remotely on the installation machine (4090, ryzen 7 5800x) it seemed to get locked up when loading, so we throttled the load and used a timer to delay initializing one of the media pipe external tox’s (there are two running - 1 running directly off the camera and another running off of the spout cam). Once we throttled the loading and delayed the init of the second spout cam we seemed to be able to run it (sometimes) on the remote machine. We are getting occasional crashes still and managed to get this crash:
and crash file:
TouchDesignerCrash.2023.12120_1.dmp (1.2 MB)
I can confirm from the dmp file that it is indeed crash in a Transform SOP.
The crash is occurring during a simple memory operation which could mean that you’re running out of memory on the remote machine during loading. It might be that some part of the load is temporarily using significant resources, which could be why throttling it reduces the chance of the crash.
In the task manager, do you notice significant memory usage or memory usage increasing rapidly during the load on the remote machine? The column to inspect would be the Commit Size on the Details page.