How CertiK Verifies a TPS Claim and What It Cannot Tell You

When a blockchain protocol claims it can process 40 million transactions per second, verifying the underlying arithmetic becomes essential.

Sui’s October 7 announcement presented that figure on a Basecamp stage in Singapore, and CertiK’s subsequent audit report placed a precise, verified figure alongside it. Comparing the initial marketing announcement with the final verification report reveals how independent auditors evaluate extreme throughput benchmarks.

This guide outlines how a security verifier like CertiK evaluates throughput results, using its TON benchmark audit as a foundational template and Sui’s tunnel performance test as a real-world implementation.

Why High TPS Claims Require Independent Verification

Throughput figures can easily be inflated under controlled conditions. A development team might select a simplified workload, calculate transactions using generous parameters, or present a single peak burst as standard network behavior. While these practices are common in marketing, unverified TPS figures remain speculative until an independent party recomputes the raw ledger metrics.

Sui’s initial announcement reported 40,614,180 TPS using its off-chain tunnel architecture at Sui Basecamp. The team also noted that CertiK would audit execution journals and publish formal findings. With CertiK’s verification report now public, we can examine how third-party technical verification works in practice.

Establishing Strict Definitions for Valid Transactions

Verification begins by defining what constitutes a single transaction, as this definition determines the final throughput total. In the Sui evaluation, CertiK counted a transaction only if it proved valid under protocol rules, contained every required authorization proof, and appeared exactly once within the canonical execution journal.

Furthermore, the audit restricted valid counts strictly to transactions committed inside the active test window. This approach filters out duplicate entries, unassigned user actions, and pre-test setup data. Consequently, if a throughput claim fails to define its underlying transaction unit, the figure carries little practical value.

Evaluating the Test Network Setup and Environment

CertiK’s TON write-up begins by documenting environmental parameters. The auditors verified the testnet infrastructure, which comprised 256 active nodes, specific software versions, and defined validator sets. Because performance on a test network does not guarantee identical results on mainnet, auditors must specify exact deployment parameters.

Similarly, the Sui evaluation documented its operational setup. The benchmark ran on mainnet on October 7 across 10 distinct test runs covering five workload types consisting of simple payments, flash loans, blackjack, tic-tac-toe, and battleship games. The report also noted that workloads launched at staggered intervals, with all 7,680 tunnels operating concurrently for a duration of 148.545 seconds.

Verifying Smart Contract Code Against Published Source Code

Matching deployed execution code to public repositories provides critical assurance during performance testing. In the TON audit, CertiK confirmed that stress-test smart contracts deployed on-chain matched the source code published by the foundation. The testing also utilized contracts with complex execution logic rather than basic balance transfers.

Without this step, a project could benchmark stripped-down programs and present them as full-featured application performance. Verifying deployed bytecode against open-source repositories prevents this discrepancy. In Sui’s benchmark, application complexity remained central, with simple payments accounting for approximately 78.98% of total recorded transactions.

Calculating Throughput Metrics From Precise Time Intervals

During the TON evaluation, CertiK collected transaction counts across one-minute intervals over a 25-minute period to derive overall TPS metrics. The peak minute recorded 6,806,474 total transactions, which translated to an audited rate of 104,715 TPS.

That result proved lower than dividing the total count by 60 seconds. The variance occurred because the network counter updated in five-second increments, prompting CertiK to measure the window as 65 seconds. Accounting for exact interval bounds produced a conservative, defensible figure.

For Sui, auditors utilized finer measurement increments. CertiK evaluated full UTC-aligned one-second intervals while discarding incomplete edge seconds. The peak throughput reached 40,682,328 transactions during the second starting at 08:46:17 UTC. Meanwhile, the sustained throughput reached a 60-second rolling average of 40,339,143.65 TPS, representing 2,420,348,619 total transactions. To measure concurrent tunnels, CertiK aggregated qualifying counts and divided by the shared duration rather than summing individual tunnel rates.

Reconciling On-Chain and Off-Chain Transaction Records

A thorough audit verifies that independent execution logs reconcile completely. CertiK confirmed that all 7,680 tunnel instances achieved VERIFIED status, with 15,241,348,297 in-window off-chain transactions matching full-history admission records exactly. Each tunnel’s logged count aligned with its on-chain declaration, and opening checkpoints matched closing parameters.

The underlying settlement architecture remains visible through these records. The benchmark tunnels opened and closed using 960 shared opening transactions and 960 shared closing transactions. As a result, 1,920 on-chain base layer transactions anchored more than 15.2 billion off-chain executions.

Comparing the Announcement Figures to the Final Audit Report

Sui’s initial announcement reported 40,614,180 TPS, whereas CertiK’s audited peak reached 40,682,328 TPS. This 68,000 transaction difference represents a minor 0.17% variance. Such a margin reflects the difference between real-time telemetry estimates and strict post-hoc recomputation.

In addition, the initial announcement stated that formal verification was ongoing. CertiK’s final report subsequently provided the missing audit metrics, establishing verified peak rates, sustained averages, total transaction volumes, and explicit calculation methodologies.

Understanding the Scope Limits of Off-Chain Audits

Despite these figures, understanding the boundaries of the audit methodology remains crucial. CertiK explicitly noted that tunnel TPS measures off-chain state transitions rather than Sui’s base-layer on-chain execution capacity. Consequently, dividing off-chain counts by base-layer settlement times does not represent native on-chain throughput.

Furthermore, several technical scope limitations apply:

• CertiK did not re-evaluate individual cryptographic signatures or execute a complete cryptographic replay for every off-chain receipt.
• The retained logs do not evaluate node clock drift, execution latency percentiles, or performance repeatability under varied network conditions.
• The sustained figure represents the single highest observed 60-second average rather than a guaranteed baseline for standard operations.
• The benchmark environment was configured and executed directly by the Sui development team for this specific campaign.

Similar scope boundaries applied to TON’s prior audit, which was evaluated on testnet infrastructure using a single-minute peak window. Independent throughput audits confirm performance within defined test parameters rather than guaranteeing general network capacity.

A Standard Checklist for Evaluating Performance Claims

1. Identify the exact definition of a transaction and confirm whether cryptographic authorization is verified.
2. Determine whether the benchmark ran on live mainnet or isolated testnet infrastructure.
3. Verify that deployed test smart contracts match public open-source code repositories.
4. Review measurement interval durations and verify the mathematical formulas used.
5. Distinguish peak single-second bursts from sustained rolling averages.
6. Differentiate off-chain channel capacity from native layer-one on-chain throughput.
7. Read the auditor’s explicit scope limitations and unverified parameters.

Evaluating independent audit reports requires examining both verified metrics and stated scope boundaries. Sui’s tunnel benchmark demonstrates significant off-chain execution capacity under controlled test conditions, provided the results are interpreted within their proper architectural context.

Disclosure: This is not trading or investment advice. Always do your research before buying any cryptocurrency or investing in any services. Follow us on X @nulltxnews