Shulou(Shulou.com)06/02 Report--

Today, a new Dell R740 server with NVIDIA Telsa M10 graphics card is installed, and the VMware ESXi 6.0U3 system installed by the server real estate system operates normally, and the video card driver installation also shows normal. However, the invalid GPU configuration file was encountered when assigning the video card to the virtual machine through vCenter, as shown in the figure:

Using the NVIDIA-smi command also reported an error

First of all, in order to confirm whether there is a fault in the video card, I can configure the video card through the client and assign it to the VM, so it is determined that it is not the video card problem.

The video card driver is downloaded from the NVIDIA official website, and there is no problem with unloading and reinstalling, and the driver version is eliminated. According to the analysis and judgment, there is a problem with the BIOS setting of Dell R740. Try the following steps:

F2Murray-enter the BIOS setting

Check each configuration

The default value is 56TB, (ESXi6.X supports maximum 16TB memory, default value 56TB, there may be a problem) modify it to 12TB, save and restart to enter the system.

Re-enter vcenter to assign GPU, the configuration scheme takes effect, and GPU works normally

Fault handling explanation:

Although ESXi is a 64-bit system, memory addressing is limited to 44 bits, that is, the maximum 16TB memory is supported; official kb: https://kb.vmware.com/s/article/2087943?lang=en_US#q=2087943

Currently, the Dell R740 factory BIOS is configured with 56TB memory by default, so the GPU is not recognized correctly, so it is necessary to ensure that the PCI addressing of the NVIDIA GRID GPU is below the 44-bit limit for GPU to work properly.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.