Zama TFHE-rs v1.7 Achieves 1080 TPS, Surpassing Roadmap Targets

Zama's TFHE-rs v1.7 hits 1,080 TPS on 8 H100 GPUs, a 2.7x boost over prior versions, completing roadmap goals a year early.

Woofun AI reports that Zama's open-source homomorphic encryption library, TFHE-rs v1.7, reached a throughput of 1,080 transactions per second on a single node with 8 NVIDIA H100 GPUs in the ERC-7984 privacy transfer benchmark. This performance marks a 2.7-fold increase from the previous version's 400 TPS, while latency held steady at 50.9 milliseconds. Zama confirmed that these results fulfill earlier roadmap objectives more than a year ahead of schedule. The protocol noted that encryption operations are no longer the primary bottleneck for privacy transfers; instead, consensus mechanisms and block generation speeds now limit performance. ERC-7984 serves as a privacy-enhanced variant of ERC-20, ensuring that account balances and transaction amounts remain fully encrypted during transfers.