Neo SPCC publishes test bench results for the final N3 version, significant TPS increases for C # and Go nodes

Neo SPCC has published the results of its last benchmarking round for the N3 milestone release. The tests, which followed the same single and four-node setups as before, show that both the core team’s C # node and Neo SPCC’s Go node have seen significant throughput improvements since Preview4.

The hardware has remained constant since the last test round with a Ryzen 9 5950X with 64 GB of RAM and an SSD. However, the Neo-Bench tool itself received some tweaks to reduce its impact on CPU resources, which gradually began to affect the results.

The single-node test bench uses a block size of one second, which leads to a maximum possible TPS of 50,000. The results of the benchmark have shown significant increases in performance for both the C # and the Go nodes since Preview4.

Average TPS in single node tests (Source: Neo SPCC)

In Preview4, NeoGo averaged 10,300 TPS while C # hit 3,100. With the upgrades in the final version, the nodes reached new heights of 30,300 and 7,900 respectively.

The four-node tests use five second blocks, which results in a theoretical limit of 10,000 TPS at the maximum block size. In Preview4, the NeoGo and C # nodes averaged 1750 and 600 TPS, respectively.

Average TPS in the four-node test (Source: Neo SPCC)

With the final N3 version, both nodes have significantly improved throughput. The core client managed to peak above 5000 TPS, but problems processing blocks of 50,000 transactions (the entire mempool) caused a drop, bringing the final average to around 1000 and always relative performance since Preview4 still doubled.

In contrast, after a brief delay at the start of the test run, the NeoGo network sped up to almost the maximum 10,000 transactions per second, resulting in a stable average of 8,800 at the end of the test.

The full benchmarking report from Neo SPCC can be read at the link below:
https://neospcc.medium.com/benchmarking-neo-n3-final-4356af1ef4eb