Two RTX PRO 6000 Blackwells, One VM: NUMA, PCIe P2P and the Road to Tensor Parallelism
中文摘要
探讨在单虚拟机中使用两块 RTX PRO 6000 Blackwell GPU，利用 NUMA、PCIe P2P 及张量并行技术实现接近 DGX 的性能。
English Summary
Exploring how to use two RTX PRO 6000 Blackwell GPUs in one VM via NUMA, PCIe P2P, and tensor parallelism to approach NVIDIA DGX station performance.
Original Excerpt
In the first article, From One RTX PRO 6000 to Two: How Close Is My AI Server to an NVIDIA DGX Station?, Continue reading on Medium »