The rtx pro 6000 blackwell server edition spec sheet says it has 4 DP 2.1 ports, but the one I have doesn’t display anything.
Does anyone know anything about this?
- attaching a bug report file and/or providing at least some basic details will dramatically improve your chances for getting any help.
- assuming the driver was properly installed (can’t say without the above), make sure that you have the latest BIOS/UEFI firmware for your mobo installed.
- basic search on this forum would reveal that Blackwell series and especially rtx pro 6000 are known to cause problems, here is the most recent example.
If its anything like my cards, they ship with compute mode only enabled.
likely in compute mode
Yes, the datacenter GPUs (passively cooled cards) all come with their mode being displayOFF as default, which equals virtualizationON, and matches the bigger portion of usecases…
This mode also comes with pretty large BAR1 requirements to the system, so many systems NOT on our certified systems list (NVIDIA-Certified Systems — NVIDIA Certification Programs Documentation) will struggle to assign enough resources to one or even multiple of these datacenter GPUs…
So if your system supports the large BAR1 and correctly detects the card, it will be able to boot, load a driver (in Windows defaults to TCC mode), but still not be able to drive any physical outputs…
This tool (NVIDIA Display Mode Selector Tool | NVIDIA Developer) is required (in a system that correctly assigns all requested resources, so the tool can talk to the GPU…), to change the default to DisplayON, equals virtualizationOFF, to then be able to drive a physical display… please read all the details in the tools docu.
This change of defaults also changes the BAR1 size requested from SystemBIOS, which then would make more systems be able to fulfil such a request, and hence be able to correctly recognize and initialize the same GPU…
Datacenter GPUs in DisplayON mode, baremetal, now can be changed from its default TCC/MCDM mode to WDDM, using nvidia-smi.exe and then can finally drive any attached display as well…
In case you want to run the datacenter GPU on a hypervisor, but pass-through the full GPU to a VM, and drive displays there, we still require you to buy a vGPU license to get into WDDM mode, since you run this on a hypervisor, which requires our testing and certification…
So, to summarize: datacenter GPUs like RTX PRO 6000 Server edition, come in a DisplayOFF default, need displaymodeselector tool to change that default, so they CAN drive a display. in Windows, the driver will default to TCC/MCDM mode, but it needs WDDM to drive a display, which needs a license, when run on any hypervisor…
The ‘easy’ way for such a config is the actively cooled twin of the product, RTX PRO Workstation edition, which comes with DisplayON by default, and also defaults to WDDM right away…
I hope this explains a few things, late, but better late then never ;-)
-Frank
Hi Frank,
Thanks for the input, I’m interested in the point about the vGPU license. I’ve been trying for 7 months now to get video output from my workstation graphics card (Pro 6000 MaxQ) in a VM (Proxmox host with ubuntu VM). Am I correct in understanding that this license restriction also applies to the workstation graphics cards? The card is set to the factory default “GRAPHICS” in the Display Mode Selector tool (which should correspond to DisplayON, virtualizationOFF if I understand it correct) and video output works on the host. However, as soon as I do a full pcie passthrough of the entire GPU, I no longer get any video output in the VM, but detection of the card and compute works fine in the vm.
We have Windows Server 2025 without virtualization, how can we change this display ON mode from the RTX PRO 6000 Blackwell Server edition?