Pith. sign in

REVIEW 4 cited by

Confidential Computing on NVIDIA Hopper GPUs: A Performance Benchmark Study

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2409.03992 v4 pith:NKN2DQRZ submitted 2024-09-06 cs.DC cs.AIcs.PF

Jianwei Zhu , Hang Yin , Peng Deng , Aline Almeida , Shunfan Zhou This is my paper · ORCID
classification cs.DCcs.AIcs.PF
keywords overheadperformancebenchmarkdatagpushoppernvidiaacross
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

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.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. The Serialized Bridge: Understanding and Recovering LLM Serving Performance under Blackwell GPU Confidential Computing

    cs.DC 2026-06 unverdicted novelty 7.0 of 10

    Under GPU-CC, LLM serving losses come from a serialized VM–GPU bridge, not compute; simple scheduling and loader changes recover most of the gap on Blackwell.

  2. Hardware Mechanisms to Dynamically Throttle AI Performance

    cs.AR 2026-07 conditional novelty 6.0 of 10

    Dynamic microarchitecture throttling of GPU memory resources can cut LLM inference performance by up to 80% with low hardware overhead, giving architects a continuous, hardware-enforced AI capability control.

  3. Performance of Confidential Computing GPUs

    cs.PF 2025-05 conditional novelty 5.0 of 10

    Confidential GPU inference under model swapping is significantly slower than non-confidential, and the gap is driven by model loading, not by inference compute.

  4. Benchmarking Confidential GPU Inference on NVIDIA H100 under Intel TDX

    cs.AI 2026-05 conditional novelty 4.0 of 10

    Confidential mode on an H100 under Intel TDX adds roughly 21-28% time-to-first-token overhead and 17-21% global token-throughput loss for two LLMs, and causes earlier saturation for the larger model.

Pith tools