Eu encontrei em logs:
Jul 11 17:09:41 gpu-006 kernel: [ 7.902609] NVRM: This PCI I/O region assigned to your NVIDIA device is invalid:
Jul 11 17:09:41 gpu-006 kernel: [ 7.902609] NVRM: BAR1 is 0M @ 0x0 (PCI:0000:1a:00.0)
Jul 11 17:09:41 gpu-006 kernel: [ 7.902611] NVRM: The system BIOS may have misconfigured your GPU.
Jul 11 17:09:41 gpu-006 kernel: [ 7.902616] nvidia: probe of 0000:1a:00.0 failed with error -1
Então eu atualizei o BIOS do servidor. E voila tudo corrigido, ambos agora são detectados.