Benchmarking t-rex, kawpowminer, nbminer and ttminer

This test was conducted by the MinerMore sysop Traysi over the course of exactly 24 hours from 2020-05-03 10:00:00 UTC until 2020-05-03 10:00:00. This is a followup from a previous test.

Hardware:

• GPUs: 4x EVGA 1080ti 11G-P4-6593-KR
• MB: SLI Z270
• CPU: Celeron G3930 @ 2.90GHz
• Power supply: Corsair HX Series, HX1000, 1000 Watt, 80+ Platinum

Other Details:

• Operating System: Ubuntu 16.04.3 TLS, NVIDIA 440.82, CUDA 10.2
• GPU details: 170W power, default clocks (no alterations to mem speed, clocks, etc).
• Pool: MinerMore RVN Testnet, Username traysi, Port 5305 static difficulty 0.01 per submit

• kawpowminer: 1.2.1+commit.c5a08914 (Open source, no dev fee)
  • kawpowminer -U --cu-devices $1 -P stratum+tcp://[email protected]:5305

• nbminer 29.1 (Proprietary with 2% dev fee)
  • nbminer -d $1 -a kawpow -o stratum+tcp://rvnt.minermore.com:5305 -u traysi.nbminer

• t-rex v0.15.2 rev.3c3012b1d6ae (Proprietary with 1% dev fee)
  • t-rex -d $1 -a kawpow -o stratum+tcp://rvnt.minermore.com:5305 -u traysi.trex -p x

• ttminer Version: 4.0.0 (Apr 21 2020 12:41:17) (Proprietary with 1% dev fee)
  • TT-Miner -d $1 -coin rvn -P stratum+tcp://[email protected]:5305

Methodology:

Each of the 4 miner software programs were assigned to a GPU in the system. Every 2 hours this assignment was changed. Starting at 10:00 the assignment was:

• GPU 0: t-rex
• GPU 1: kawpowminer
• GPU 2: nbminer
• GPU 3: ttminer

At 12:00 the miners were stopped and restarted with a new assignment of:

• GPU 0: ttminer
• GPU 1: t-rex
• GPU 2: kawpowminer
• GPU 3: nbminer

At 14:00 the miners were stopped and restarted with a new assignment of:

• GPU 0: nbminer
• GPU 1: ttminer
• GPU 2: t-rex
• GPU 3: kawpowminer

At 16:00 the miners were stopped and restarted with a new assignment of:

• GPU 0: kawpowminer
• GPU 1: nbminer
• GPU 2: ttminer
• GPU 3: t-rex

And finally at 18:00 the miners were restarted with their original assignment and the loop continued as described above. In this way, after 24 hours, each miner had mined for an equal amount of time on each GPU in the system. The data was collected on the pool side and the number of accepted shares is what counted towards performance. This means that when you look at the report below, these numbers reflect the final performance of the miner even after the dev fee is taken into account.

Conclusion:

t-rex performed the best in this test. Once again kawpowminer outperformed ttminer and nbminer, but the newly released t-rex somehow outperformed them all. The difference was substantial enough that at this time t-rex may be the best option to choose for use on nvidia cards at the time of the fork on May 6th. Future updates to miner software will be tested by Traysi but for now this is the best information we have.

It is worth noting that this test isolated each miner to a single GPU. The effectiveness of the miners when run on multiple GPUs is uncertain and a future test may be conducted with trex and kawpowminer each being given 2 GPUs to work with.

To learn how to mine Ravencoin after the fork on May 6th, visit this page.