Hello everyone
Digital twins for fluid simulation is out and I’m trying to run it locally.
Following the github site(GitHub - NVIDIA-Omniverse-blueprints/digital-twins-for-fluid-simulation)
I’ve tried to visualize the car but only the UI is showing (car and background are not showing). Is it the expected result when machine spec is not enough to run it?
I know the spec on my machine isn’t good enough (RTX A5500) but I was wondering if whether the issue come from the machine spec or if I missed an important step when setting up digital twins.
Because I’m only trying to display it locally I’ve skipped the following parts:
Open Required Ports
Configuration for Clouds and VPNs
Does above parts has to be done even if running locally?
In the “open required ports” part, it says " This blueprint uses Omniverse Kit App Streaming and the @nvidia/omniverse-webrtc-streaming-library client library to stream the simulation to the client application.". However I’m not sure what to do exactly on this part.
hey @cnc-support
looks like you are on a roll with these blueprints, great to see!
You do need to complete the ‘Open Required Ports’ to get this to work - you are on Ubuntu, correct? Im not familiar with this workflow on this system so I asked chatGPT and it gave me this response
. Check if UFW is enabled
sudo ufw status
If it says inactive, you can enable it after configuring the rules:
sudo ufw enable
2. Allow Required Ports
sudo ufw allow 5273/tcp
sudo ufw allow 1024/udp
I recommend you reach out on their repo specifically for any more issues, good luck!
Hello
Thank you for your help.
You are correct, I’m on Ubuntu 22.04
I’ve checked ufw status and it was disabled.
After enabling the firewall and allow the required ports, I gave another try but nothing changed(only UI shown and the car/ environment doesn’t show up)
From the github, it says “at least 2x RTX ™ GPUs with at least 40GB of memory each, e.g., 2xL40S or 2xA6000” for hardware and my computer only has 1 RTX A5500 so I know my hardware doesn’t meet the requirement.
That’s why I wanted to know if the car/environment not showing was an expecting behavior when the hardware doesn’t meet the requirement or I just failed to configure my computer.
When you say to reach out on their repo, do you mean the github page?
The A5500 is a pretty great card, but it is two generations behind and with only 24GB ram. Some of these models require a massive amount of memory. We currently recommend either an A6000 or ADA A6000 both with 48GB ram
without insurance of any ROI, it’ll be hard to ask my company to buy such card.
I’ll ask about current state in the repository in order to know if current behavior is because of missing setting or if it is a normal state when hardware spec are not met.
Yes I am afraid it states quite clearly in the GitHub repo, “At least 2x RTX ™ GPUs with at least 40GB of memory each, e.g., 2xL40S or 2xA6000” If they are saying “40”, they really mean it. So 24 just won’t load the whole scene. However, it will load some of the scene. You can just prioritize what parts. If it is just the background, at least you have the CFD data working.
A really easy and cheap solution for you is to rent a dual A6000 machine on Amazon Cloud or something. You can rent it for the day, or the week and test it out. If it works, you know what you have to do.
Thank you for your explanation.
Because of the hard spec, I wasn’t sure if using a A5500, it would just take way more time than using 2*A6000 or if nothing is loaded as only the UI is showing, no part of the car mesh/ mesh for the particle system is being displayed.
I’ll check with my team the next step we can do.
Yes, ram is not for speed, but for workflow. It is to display all the geometry and data you need. If you NEED 40GB of data bandwidth, and only have 24GB, then something is going to have to be cut. Omniverse will go as far as it can and then just stop, and not display anything else.