Is VirtualLink possible between 2 RTX A6000 of different brands?

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

If I use two RTX A6000 GPUs, one from nVidia, and the other one from PNY, will I be able to use VirtualLink? Is the exact location of the VirtualLink plug/bridge “standard” across different manufacturers?

Thanks for your insight.

2

Did you perhaps mean to ask about NVLink?

NVLink can be used to connect cards “together”. Virtuallink is a method for connecting an AR/VR headset to a GPU.

3

Are you sure the RTX A6000 has a VirtualLink connector? NVIDIA’s datasheet for the RTX A6000 does not mention support for VirtualLink. Pictures of the RTX A6000 show only four external connectors, which correspond to the four DisplayPort 1.4a connectors mentioned in the datasheet.

By contrast, the datasheet for the RTX 6000 does list a VirtualLink Connector, and pictures do show an additional USB-style connector next to the four DisplayPort connectors, matching the datasheet.

4

There is no virtuallink support on RTX A6000.

5

My apology, I meant NVLink!

Can I bridge two RTX A6000 using NVLink if they are from different brands?

The RTX A6000 from PNY is shown here:

The model from nVidia is here:

Both cards seem mostly identical and have NVLink connectors at the exact same location… it seems.

I have one RTX A6000 on the way (nVidia) and could use another card from PNY for NVLink expansion. Will it work?

6

All RTX A6000 cards share the same physical and electrical form-factors. It should be possible to physically connect the NVLink bridge between any 2 such cards.

I personally don’t recommend mixing cards from different sources, as there may be VBIOS differences that pertain to integration with a particular vendor’s workstation platform. However I doubt any of this would interfere with NVLink connectivity.

7

Thank you!