PCIE peer-to-peer capabilities of RTX Ampere "GeForce" 3000 cards

[RoyInHouston]

Q1: Are the new RTX “Ampere” 3000-series cards capable of peer-to-peer / UVA over pcie?

The previous gen of RTX GeForce card (2000-series) could not peer over pcie but could peer over NVLink (maximum 2 cards)

Q2: As NVLink is only available on the 3090 card specifically in the 3000 series - is that a true data-link UVA type of link for CUDA or only for “SLI” type graphics linking?

For Reference Purposes:

Using multiple RTX 2080 Ti cards in parallel not possible?

I have done further investigations with caffe framework by looking at the source code and doing some experiments (I presume tensorflow will be same). The way caffe uses multiple GPU cards is by spreading batch between GPUs. That is if you have batch of 64 samples which you want to process in parallel on 4 cards than each card will be processing forward propagation stage and most of backward propagation for batch of 16 . Only tiny amount of computation (with very small amount of data) is done bet…

I believe the 2000-series cards offered NVLink for 2070 and above and only for 2-cards max. That seemed to link for UVA as well.