>The software was built to allow customers to track a chip's overall computing performance - a common practice among companies that buy fleets of processors for large data centers - and would use the time delay in communicating with servers run by Nvidia to give a sense of the chip's location on par with what other internet-based services can provide, according to an Nvidia official.
What will prevent the GPU from disabling itself when it can't reach to the servers for attestation? NVIDIA's GPUs have encrypted firmware, signing keys baked into the silicon and a RISC-V processor on-board. It can ask the driver to attest, and can stop working when the attestation time outs.
Also, remember, the driver's kernel module is gutted and moved to the card itself. Hence while GL drivers and other parts are also closed source, driver core is on the card and sitting there encrypted. Also, this feature uses the secure enclave on the card, dubbed confidential computing capability.
NVIDIA's cards are computers inside computers now, and they'll not stop just because their packets are dropped by a pesky iptables rule.
just a firewall will do