Confidential Computing on NVIDIA Hopper GPUs: A Performance Benchmark Study
It addresses performance concerns for users implementing confidential computing in AI applications, but is incremental as it benchmarks existing technology.
This study evaluated the performance impact of enabling Trusted Execution Environments (TEE) on NVIDIA Hopper GPUs for LLM inference, finding that the overhead is primarily due to CPU-GPU data transfer, with typical queries experiencing below 7% overhead and larger models showing nearly zero overhead.
This report evaluates the performance impact of enabling Trusted Execution Environments (TEE) on NVIDIA Hopper GPUs for large language model (LLM) inference tasks. We benchmark the overhead introduced by TEE mode across various LLMs and token lengths, with a particular focus on the bottleneck caused by CPU-GPU data transfers via PCIe. Our results indicate that while there is minimal computational overhead within the GPU, the overall performance penalty is primarily attributable to data transfer. For the majority of typical LLM queries, the overhead remains below 7%, with larger models and longer sequences experiencing nearly zero overhead.