Keccak Acceleration Chip and Benchmarks

We’re excited to share a significant milestone in our journey to optimize Ethereum block proving: the integration of a Keccak Acceleration Chip with the Valida zkVM and compiler toolchain. This breakthrough dramatically reduces the proving time of Ethereum blocks, bringing us closer to our goal of supporting real-time block proving with Valida.

‍

TL;DR

• Fast Keccak Proving: We’ve optimized the Keccak permutation (Keccakf-1600), the core computational burden of Keccak hashing.
• New Chip Integration: A novel interaction between the Keccak acceleration chip and the Valida zkVM memory system uses pointers to addresses, enabling seamless communication and consistency.
• Significant Speedup: Initial results show impressive performance improvements:.

Initial Results: A Promising Start

We’re thrilled with the initial results of our Keccak acceleration efforts:

These results demonstrate the potential of our approach to significantly reduce the proving time of Ethereum blocks.

We proved execution of 500 Keccak hashes, each taking as input the output of the previous hash computation. We ran these proofs on a z1d.metal AWS instance, with 48 cores and 384 GiB RAM. We took a sample size of ten runs for each of the two groups. The reported results are the mean real time of each group: 183 seconds without the Keccak acceleration chip, and 23 seconds with the Keccak acceleration chip.

Why Keccak Matters in Ethereum Block Proving

Keccak-256 is a cornerstone of Ethereum’s cryptographic infrastructure. It’s used for:

• Generating addresses from public keys.
• Creating transaction hashes.
• Building the Merkle Patricia Trie.
• Verifying block headers and other critical components.

In Ethereum block proving, Keccak-256 hashes are computed repeatedly to validate block data and verify state transitions. According to the reth team, as of March 2024, around 80-90% of client time was spent on computing keccak. This makes Keccak hashing a major computational bottleneck for block proving. Optimizations made since then by the reth team reduce the number of Keccak hashes that need to be computed, while the Valida Keccak hash chip makes it faster to prove those computations on Valida. These measures together promise to dramatically reduce the time and energy spent on proving Keccak hash executions during block proving.‍

The Keccak Acceleration Chip

The Valida Keccak acceleration chip optimizes the computation of the Keccakf-1600 permutation, the most computationally intensive part of the Keccak hash function. Here’s how it works:

Key Features of the Chip

(1) Single-Core Keccakf-1600 Permutation:

• The chip computes all 24 rounds of the Keccakf permutation in a single setting, outputting the final post-image of the permutation.
• This eliminates the need for multiple iterations, significantly reducing computation time.

(2) Memory Integration:

• The preimage (input data) is stored in 50 sequential memory addresses, and the postimage (output data) is written to the next 50 sequential addresses.
• This eliminates the need for multiple iterations, significantly reducing computation time.

(3) Consistency via Memory Bus:

• A memory bus ensures consistency between the preimage and postimage as represented in the KeccakF chip’s trace and in the memory.

Integration with Valida zkVM

The Keccak acceleration chip is integrated with the Valida zkVM and compiler toolchain, enabling efficient block proving. A collaborative effort from our compiler and cryptography engineers. This integration is achieved through:

• Pointer Bus: Allows the CPU to pass the base address pointer to the KeccakF chip.
• Memory Bus: Ensures consistency between the KeccakF chip’s trace and the memory representation of the data.
• This tight integration minimizes overhead and maximizes performance, making Ethereum block proving faster and more efficient.

What’s Next

While we’re proud of our progress, there’s still more work to be done. Our next steps include:

• Full Integration: Completing the integration of the Keccak acceleration chip with the Valida zkVM.
• Further Optimizations: Exploring additional optimizations such as ECDSA signature acceleration to achieve even greater speedups.
• Hardware Acceleration: Leveraging hardware acceleration to push the boundaries of performance.

Read the full release notes here.