By using AWS re:Post, you agree to the Terms of Use

Nvidia Driver R515 does not work on P3 instance types


I have loaded the Nvidia drivers on a CentOS 7 AMI from Nvidia's RPM repositories. In the past this was loading the R510 drivers and they were working on P3 instances with V100 GPUs, but earlier this month they updated the repository to R515 and those drivers are not recognizing the V100 GPUs in the P3 instances. I have some on-premise nodes with a different model of V100 and the R515 drivers work there so the problem appears to only be with AWS instances. I have tried opening a bug report with Nvidia, but have not gotten a response yet. Can anybody else reproduce this issue or have I messed up something? If this is indeed an issue with the new driver then any tips on getting Nvidia to fix it? Or is it a problem specific to AWS instances that AWS needs to fix?