My understanding is that NVIDIA GPUs have supported some form of firmware protection for several generations. For example, Maxwell-generation GPUs appear to have introduced anti-tamper mechanisms, where the hardware started to enforce signature checks for firmware running on internal microcontrollers. Later, NVIDIA documentation and blog posts mention firmware authentication on Volta V100, and additional firmware security features such as encrypted firmware, firmware revocation, and fault-injection countermeasures on Turing/Ampere. Based on my understanding, these mechanisms seem to represent the main purpose of GPU Secure Boot: ensuring that only authenticated, untampered firmware is allowed to execute on the GPU’s internal processors during the boot and initialization process. However, from the user side, it does not seem straightforward to confirm whether GPU Secure Boot is actually enabled and enforced on a device.
My questions are: 1.For non-Confidential-Computing GPUs such as A100 or RTX GPUs, is GPU Secure Boot / firmware authentication enabled by default and enforced by hardware?
2.Is there any public interface, such as nvidia-smi, NVML, kernel logs, or driver APIs, that can be used to query whether GPU Secure Boot / firmware authentication is enabled?