If I remember correctly they stopped supporting NVLINK on the first generation of cards that had tensor cores (Turing). This is not an oversight on NVIDIA's pa

If I remember correctly they stopped supporting NVLINK on the first generation of cards that had tensor cores (Turing). This is not an oversight on NVIDIA's part, this is intentional product segmentation.