We’re running into a the RmInitAdapter problem specifically with the 3080 Ti. On our system, we’ve been able to successfully deploy the 3080 with 470.86 on Ubuntu 18.04 (5.4.0-91).
On a fresh install of Ubuntu 18.04, we still run into the RmInitAdapter problem. This problem is reproducible across the same model of computer with different 3080 Tis. It is only resolvable by using a 3080. We’ve tested with the following different drivers and have run into the RmInitAdapter problem for them.
470.86
470.94
495.46
The attached log file is for version 495.46 on Ubuntu 18.04 (5.4.0-91) after a fresh install of Ubuntu.
I’m using a 600W supply at a bench setup, which I figure might not be sufficient enough for any actual processing, but I believe should be enough to get the device up and running and detectable by the drivers. In production, with the same computer, we have more than a sufficient power supply and we still run into the same error messages. We use an external power supply, hence why we are able to test with one wattage and deploy to production with another wattage.
It’s run off of a vehicle’s battery, so it should be more than enough to power a GPU. Some alternators can source > 1000W, so it should be sufficient. I don’t have specs on the alternators we use, but it’s pretty industrial.
I’ve tested two different 3080 Ti’s on the same machine along with a 3080. Both 3080 Tis seem to run into the RmInitAdapter problem, but the 3080 is fine.
I contacted the manufacturer and they told me there has been no BIOS update to the mainboard.
Sounds like a general board incompatibility. According to the logs, the 3080ti is mounted in a x8 slot but the board also seems to have a x16 slot, currently housing some asix serial controller. Is it possible to use the x16 slot for the gpu?