/tags/gpu/ Recent content in Gpu on Confidential Containers Hugo en-us /docs/examples/nvidia-gpu-examples/ Mon, 01 Jan 0001 00:00:00 +0000 /docs/examples/nvidia-gpu-examples/ <p>These examples show how to request NVIDIA passthrough devices in pod specs.</p> <p>They require that you have deployed CoCo following the <a href="https://docs.nvidia.com/datacenter/cloud-native/confidential-containers/latest/">NVIDIA Confidential Containers Reference Architecture</a> which documents supported component versions and the passthrough modes used below.</p> <p>In brief: NVIDIA Hopper, NVIDIA Blackwell, and NVIDIA RTX Pro 6000 all support Single-GPU passthrough (SPT). Hopper and Blackwell additionally support Multi-GPU passthrough (MPT). Protected PCIe (PPCIE) mode is unique to Hopper multi-gpu usages. The following sections are example pod fragments aligned to each case.</p>