Grin Mining Software Performance Update #1

Matthieu Riou
BlockCypher Blog
Published in
2 min readMar 15, 2019
Flash, ahhah

The Grinmint team will be providing news on mining software when we see important updates or releases. This is to help our miners maximize their hashrates.

Currently, the fastest and most stable mining software for Grin on most Nvidia 6GB+ GPUs is bminer; the latest version of their software is 15.3.1. For Nvidia cards with only 4GB or AMD cards see minerbabe, which we will talk about in a later post.

We regularly upgrade our test miners and we’ve had a number of issues with performance and stability regressions since bminer version 14.3.1. To date, 14.3.1 was the fastest and most stable version. With 15.3.1, we have noticed an average an 8% increase in hashrate and fidelity*. These are the benchmarks we’ve seen for 15.3.1 on these cards:

  • GTX 1080 ti: ~7.3 G/s
  • RTX 2080: ~8.2 G/s

Remember that exact G/s is calculated on valid shares submitted to the mining pool. Your mining software will often show you a higher rate than what your actual hashrate is. We briefly explain the hashrate calculation in our FAQ and we’ll be providing more detail on why this calculation is accurate.

There are a few caveats with bminer version 15.3.1:

  1. Nvidia driver versions older than 410 have shown to frequently crash with this latest release. If you’re still on 390 (which was the longest supported release), it’s time to update your Nvidia driver.
  2. In some Linux environments, we’ve seen hard drive crashes on the 1080ti. Make sure to always test a few miners in your environment before fully upgrading all of them.

Make sure to upgrade to 15.3.1. When new versions provide additional improvements, we will let you know. Subscribe or follow us on Twitter to get new updates!

* Fidelity is the ratio of solutions per graph found by the miner compared to the theoretical ratio. On average, a miner should find one valid solution every 42 graphs. Mining software that finds all solutions for the graphs it analyzes should have a fidelity of 1.

--

--

Matthieu Riou
BlockCypher Blog

Founder @blockcypher. Founded @mmztweet. Unreasonable love for double ristretto and coding.