Cysic founder and CEO, Leo Fan, revealed the behind-the-scenes challenges of blockchain networks: zero-knowledge or ZK proof systems are currently forced to compete with trillion-dollar AI data centers for GPU supplies. This scramble for hardware has a direct impact on crypto network computing costs.
The biggest impact is felt by layer-2 networks based on ZK-rollups. Competition for GPU supply has triggered a series of cost increases for generating proofs in their systems. According to L2Beat data, the ZK proof generation process accounts for a dominant share, namely 60% to 70% of the total transaction costs on ZK layer-2 networks.
Why Is ZK Not Suited for Regular GPUs?
The root of the problem lies in the different workload characteristics. Fan explained that the AI industry focuses on matrix calculations. Conversely, the ZK proving process relies heavily on multi-scalar multiplication operations and high-level computation diagram organization.
Due to this fundamental difference, ZK cryptographic algorithms become inefficient when run on regular GPUs. The performance of existing devices favors artificial intelligence computing needs, forcing crypto systems to bear inefficiencies that waste time and money.
Chasing the Ethereum Foundation’s Strict Standards
Addressing the limitations of standard hardware, Cysic built a proving machine named Venus. This new computing machine successfully reduced proving time by up to 9% compared to the ZisK version 0.16.1 system. This speed improvement was achieved by reducing data transfers and synchronization processes between CPU and GPU components.
As a performance comparison, the ZisK system is claimed to have the ability to generate Ethereum block proofs in 7.4 seconds using 24 GPU units. The ZisK system is also capable of running in real-time with a single configuration of one RTX 4090 graphics card.
This computing speed is required to meet the benchmark targets set by the Ethereum Foundation. The foundation sets real-time proving standards as the finality criteria: systems must be able to process proofs for at least 99% of mainnet blocks within a maximum window of 10 seconds. In addition to the time limit, the proof file size must not exceed 300 KiB and must consume a maximum of 10 kilowatts of power.
Separation of Computing Infrastructure
Relying on off-the-shelf GPUs means placing the fate of blockchain transaction costs in the midst of artificial intelligence budgets. Transitioning to specialized, non-GPU hardware is starting to emerge as an alternative route so that ZK technology can advance without being dragged down by the operational costs of other parties. As reported by crypto.news.
Read also: How Crypto Staking Works and Its Risks
Disclaimer: This article is for informational and educational purposes only, not financial advice. Cryptocurrency assets are highly volatile and carry significant risk. Always do your own research (DYOR) and never invest more than you can afford to lose.
