Vitalik Buterin proposes new ‘efficiency ratio’ to measure cryptographic performance
The efficiency ratio compares the encrypted computation time to the raw computation time, giving you a clearer picture of performance overhead.
Challenges remain due to heterogeneous operations, parallelization, and memory patterns
Ethereum is exploring new ways to measure and improve the performance of its cryptosystem.
Co-founder Vitalik Buterin is now calling for changes to the way developers evaluate cryptographic systems such as zero-knowledge proofs (ZK) and fully homomorphic encryption (FHE), with the aim of providing more meaningful metrics.
Buterin proposes a new efficiency indicator
Traditionally, cryptographic performance has been measured using a metric called “operations per second,” which is hardware dependent and can be misleading. Instead, Buterin suggests using an “efficiency ratio,” which is the ratio of the cryptographic computation time to the raw computation time.
I wish more ZK and FHE people would show the overhead as a ratio (in-cipher computation time vs. raw computation time) instead of just saying “I can perform N operations per second.”
It is more hardware independent and provides very useful numbers. In other words, how efficient is it?
— vitalik.eth (@VitalikButerin) October 18, 2025
He points out that this approach is less hardware dependent and clearly shows how much efficiency is lost by encrypting apps. It also makes it easier to estimate performance because developers already know how long the raw calculations will take.
Challenges in measuring cryptography
Vitalik acknowledges that this is difficult because the operations involved are heterogeneous, especially since the execution and proof steps can be different due to differences in parallelization (SIMD) and memory access patterns. Therefore, even ratios can be influenced to some extent by hardware.
Despite these limitations, he believes that the overhead factor is a useful and meaningful metric for evaluating cryptographic performance.
Cryptographer Lucas Helminger asked how to benchmark the overhead of fully homomorphic encryption (FHE) and multiparty computation (MPC), noting that they are more complex than zero-knowledge proofs. He also wondered what network assumptions and number of parties should be considered when calculating overhead.
Is there a way to benchmark the overhead against raw in a FHE/MPC setup? What network should I assume, number of parties, etc? It’s not as easy to derive as in ZK imo.
— Lucas Helminger (@luhelminger) October 18, 2025
Buterin explained that since FHE is mostly a single-party process, network considerations have little impact. Only small steps such as submitting input and performing threshold decoding are important, and these are negligible compared to the computation time.
Helminger pointed out that in real-world blockchain scenarios, there can be extra overhead when many nodes are involved. Buterin agreed, but said the raw runtime in a deployed configuration still provides the clearest picture.
Ethereum scaling breakthrough
Most recently, Brevis announced Pico Prism, a high-performance zkVM for real-time Ethereum block proofs. Using 64 RTX 5090 GPUs, we prove 99.6% of blocks within 12 seconds, taking an average of 6.9 seconds.
This breakthrough could increase Ethereum’s scalability by up to 100x and move us toward a future where anyone can validate the blockchain, even from their smartphones. Buterin also emphasized that this is a major advancement for ZK-EVM, which proves its speed and versatility.
Excited to see @brevis_zk’s Pico Prism enter the ZK-EVM demonstration arena.
Significant advancement in ZK-EVM demonstrating speed and versatility. https://t.co/nOeYt9YMvm
— vitalik.eth (@VitalikButerin) October 15, 2025
The future powered by Ethereum’s ambitious zk
Cryptocurrency investor Ryan Sean Adams emphasizes that Ethereum is on a fundamentally different path than other blockchains.
He sees this evolving into a zk-powered chain with layer 1 handling global DeFi with high throughput (10,000 TPS) and nodes light enough to run on a phone. The Layer 2 network handles everything else, including general-purpose applications like Base and Arbitrum and app chains like Writer.

