Benchmarking the future of private computations
Description
In this deep-dive technical session, P2P.org presents benchmark results comparing two NVIDIA GPU architectures — H100 and L40S — using the open-source prover. The team explores two scaling strategies:
Assigning multiple GPUs to a single proof generation, and
Running multiple proofs in parallel, with one GPU each.
The findings? Parallel proof execution delivers significantly higher throughput on both platforms, with diminishing returns for multi-GPU single-proof setups.
Categories
Format: Conference
Topic: Digital Assets, Business
Distribution: in-person
Talk language: English
Ticket cost: Free access