AMD Firepro W7000 vs Nvidia Quadro 4000 for vGPU sharing

[ACiD_GRiM]

I've built a systems lab to test out Horizon VDI and am trying to decide on a card for graphics acceleration. I'm currently using sVGA off a low end AMD GPU in each node, however I'd like to provide more power to these test VMs AND retain vMotion ability to facilitate putting one node in maintenance mode. Since vDGA satisfies my first requirement, but not the other, I'm looking for the vGPU model.

I know it's well publicized that vGPU works with NVIDIA Grid, however I don't want to spend $$$ on just one card when I need at least two. I'm looking at the firepro W7000 and Quadro 4000 cards which are within my budget and I believe vGPU will work with the Quadro 4000, however I'd like to hear if anyone has experience with the W7000 supporting vGPU, or at least can confirm that the Q4000 will support vGPU for sure.

[Linjo]

Only the Nvidia GRID K1/K2 works with "vGPU", its actually a Nvidia specific feature so do not expect this to ever work with other GPU brands.

On the other hand the others might come up with similar functionality.

// Linjo

[Alistar]

Hi there,

I do not think you are out of luck yet - I can't guarantee that Quadro 4000 will work at least as vSGA but the following is the output of nVidia Shared Management Interface nvidia-smi inside the ESXi host and it looks quite promising:

/etc/init.d # nvidia-smi -h

NVIDIA System Management Interface -- v340.32

NVSMI provides monitoring information for Tesla and select Quadro devices. The data is presented in either a plain text or an XML format, via stdout or a file.NVSMI also provides several management operations for changing the device state. Note that the functionality of NVSMI is exposed through the NVML C-based library. See the NVIDIA developer website for more information about NVML. Python and Perl wrappers to NVML are also available.  The output of NVSMI is not guaranteed to be backwards compatible; NVML and the bindings are backwards compatible.

Supported products:

- Full Support

    - All Tesla products, starting with the Fermi architecture

    - All Quadro products, starting with the Fermi architecture

    - All GRID products, starting with the Kepler architecture

- Limited Support

    - All Geforce products, starting with the Fermi architecture

Also, you can find the supported deviceIDs in its parsing file: 06dd  NVIDIAQuadro 4000.

So theoretically if you install the nVidia VIB in your hypervisor, and on the next reboot xorg and NVIDIA Module startup is successful, you might reach the Shared VGA functionality - although with much lower ratio of users:desktop that nVidia Grid promises :smileyhappy:

If you decide to do that, please keep us updated and good luck!

[Linjo]

Alistar not sure about that, nvidia-smi is just the config utility and is very separate from the vib and xorg.

The nvidia-smi is probably just a ported linux version and the gpu:s listed in the helpfile is probably a red herring.

// Linjo

[Alistar]

Nice catch there, sir. But nontheless I'd be pretty interested if it would really work with the nVidia adapters mentioned in the supported adapters file - afterall from the hardware standpoint the GRID GPUs are *just* Kepler multi-chip PCBs with the vGPU enabled by the software.

But that would all be all just a discussion topic and would not justify spending any money on any GPU except GRID just to see "if it works" :smileyhappy:

EDIT: It seems this post is more or less confiriming what I first thought: VMware Horizon View graphics - vSGA vs vDGA VMGuru