ZisK, an open-source zero-knowledge virtual machine project, said on Aug. 18 that its v1.1.0-alpha prover finished a benchmark run in 9.62 seconds at the 99th percentile on four Nvidia RTX 5090 GPUs. That's under Ethereum's 10-second target for proving blocks, but the project hasn't released the full methodology, and the claim is provisional.
The numbers ZisK posted
ZisK's announcement claims a 9.62-second p99 across its test set, with 99.7% of tested Ethereum blocks finishing in under 10 seconds. The project also says the result carries 128-bit security and post-quantum resistance. The hardware: four RTX 5090s, which have a combined GPU-only power draw of 2.3 kW and cost $7,996 at launch. That's well under Ethereum's $100,000 equipment cap, but the benchmark doesn't say what the whole system pulled from the wall.
What Ethereum actually demands
Ethereum's working standard isn't just a latency number. The Ethereum Foundation paired the 10-second target with five other conditions: on-premises equipment costing no more than $100,000, power draw under 10 kW, fully open-source code, at least 128-bit security, and proofs no larger than 300 KiB without trusted setups. ZisK checks the open-source box — it's MIT or Apache 2.0 licensed — and claims 128-bit security. But there's no word on proof size, and the power figure is only the GPU rating, not the whole system.
The missing methodology
The benchmark lacks several critical details: the block range tested, the sample size, how the p99 was calculated, the timing boundary, the proof size, and measured whole-system power. That makes the result provisional. The Ethproofs API, which defines how proving time is measured, includes witness generation but excludes data fetching and proof-submission latency. Without knowing which boundary ZisK used, the 9.62-second figure isn't directly comparable to Ethereum's requirement.
How OpenVM's result compares
ZisK isn't the only project chasing this target. OpenVM's July production release reported a 9.8-second p99 on eight RTX 5090s across 7,200 Ethereum mainnet blocks starting at block 24,000,000. That result used 100-bit provable security and kept proofs under 300 kB. ZisK used half the GPUs and claims a faster p99, but also claims stronger security. Without the full benchmark data, it's hard to say which is the more meaningful result.
ZisK is open source, so anyone can run the code and check the numbers. The next step is for the project to publish the missing details — block range, sample size, timing boundary, proof size, and real power draw. Until then, the 9.62-second p99 is a headline, not a verified claim. Ethereum's bar is set, and ZisK hasn't shown it clears all of it.
