Determining if New Blackwell RTX Pro 6000 Server Card is a DOA unit?

Hello. We have received a new Blackwell RTX Pro 6000 Server Card that doesn’t appear to work, but we can’t be 100% sure if there is something we are missing in running it up.

It will show as a device when entering lspci command. Something along the lines of (currently away from the server in question and will have to re-install the card for the exact output)

xx:xx.x 3D controller: NVIDIA Corporation Device xxxx (rev a1)

No video output from the display ports. nvidia-smi doesn’t find the card after installing Nvidia open drivers 575.

We have put the card in another workstation that is running Windows 11 with latest Nvidia driver. No video signal from the RTX 6000. Will have to source an additional non-powered GPU to bring up a display to further diagnose under Windows.

We don’t have any of the same Blackwell RTX Pro 6000 units to compare against.

My feeling is we should expect the card to produce a signal from the display ports without the need for additional software, like the previous generation of Nvidia GPUs or any other GPU. Unsure is this new to Blackwell and we are missing something.

Thankyou

seems dead indeed, but if you want any Nvidia employee to have a look (not guaranteed of course, actually pretty low chances due to their understaffing), you should include a bug report file.

Cheers!

We’ve come to the same conclusion, shipping it back for a replacement.

What PCI-E x16 Gen you are using ? Had you updated the BIOS/UEFI Firmware before installing the card. Recent gen cards pose some compatibility issues due to backdated BIOS or PCI-E generational issues. Kindly check both.

Tested using latest bios in workstations released in 2024 - Dell Precision 7XXX (AMD) and Lenovo Thinkstation PX (Dual Xeon). PCI-E x16 Gen 5.

Its not dead, I have the same issue with 2 of my cards, they come locked into compute mode only, and you can only change it with the Display mode selector tool, and you cant get hold of it unless NVidia approve you. and I’ve wasted weeks to work out what the issue was.

So ill probably be sending mine back too if I cant get hold of the tool to be able to use the cards as intended.

If the card is in compute mode to change it to graphic (desktop) mode use nvidia-smi utility by running nvidia-smi -g 0 -dm 0 for the first GPU (dm options being 0 = WDDM mode, 1 = TCC mode).

Is there any output when running nvidia-smi -q -d COMPUTE?

According to NVIDIA Display Mode Selector Tool | NVIDIA Developer it should only be used if the default GPU configuration is insufficient for the desired use case

Unable to set driver model for GPU 00000000:41:00.0: Not Supported

same for the other card too

Our distributor confirmed the card was DOA.

It also says:

A system (workstation or server) needs to be specifically qualified for any non-default mode. It is critical to check with the system OEM to confirm that the desired display mode is supported. Misuse of the tool can permanently damage the GPU without the option to return to the default configuration.

oh, so now additionally to MIG support, Nvidia also locks display output on these cards? This is beyond pathetic…

I emailed PNY as nvidia dont seem to want to talk unless you have an enterprise account, anyway PNY to their credit sent me the tool, I enabled them with 8gb bar and then used smi to enable graphics, and would you believe it… works just fine. (i had to disable the CSM module in bios boot also)

I suspect a lot of distributors will also come to the conclusion the cards are dead when they test them, mine were in the exact same state as the op of this thread.

no dead, pls use this tool: NVIDIA Display Mode Selector Tool | NVIDIA Developer