HiveOS tutorial: worker configuration, flight sheet, overclocking, etc.

Your rig is built and HiveOS is installed: now it has to mine, the cards need tuning and you want to monitor it remotely. In this 2021 video (in French), we go through the HiveOS features we consider essential, on a rig of 8 RTX 3070 cards mining Ravencoin: flight sheet, overclocking, remote actions, updates and the watchdog.

1. Reading the worker dashboard

In HiveOS, each rig is a "worker". Its page shows, for each card, the hashrate, temperature, fan speed, estimated power draw and overclocking settings. The card details also list each card's TDP and memory type, a key piece of information for overclocking.

2. Creating and applying a flight sheet

The flight sheet tells the rig what to mine and how. It comes down to four choices:

1. The coin: Ravencoin here;
2. The wallet: a Ravencoin address added in HiveOS. We use one from the Binance exchange;
3. The mining pool: for example 2Miners, or another widely used pool;
4. The mining software: we recommend T-Rex, which only supports NVIDIA cards but handles them very well.

Once the flight sheet is saved, apply it to the worker by clicking the rocket icon.

3. Tuning overclocking to the cards' memory

At first, the rig kept crashing without any clear error message. We had started from the settings of our non-LHR RTX 3060 Ti Founders Edition cards, with the memory overclocked by +2400 to +2600 MHz. But the RTX 3070 cards in this rig do not all have the same memory:

• Samsung: the Gigabyte, the only card in the rig with Samsung memory, reaches +1400 MHz on the memory and 28.5 MH/s for 137 W according to HiveOS;
• Hynix: this memory does not clock as high. The other cards top out at 27–27.5 MH/s and draw 20 to 30 W more.

Our advice: before overclocking, check the memory fitted on each card, as it can differ.

Global or per-card overclocking

Clicking the worker's overclocking counter lets you set the core clock (GPU frequency, either as a relative offset, plus or minus, or as an absolute value, for example 1300 MHz), the memory clock, the power limit and the fan speed as a percentage. A single value, for example 65% for the fans, applies to all cards (GPU 0 to GPU 7 here). To tune a single card, click its line in the worker.

For a starting point, the "Popular" tab shows, for a given card model and algorithm (RTX 3070 and Ravencoin in our example), the settings used by other users and their share as a percentage. Memory values range from +1800 to +2600 MHz: it mainly depends on the card's memory.

4. Managing the rig remotely

The worker menu groups the everyday actions:

• Power actions: reboot, shut down, or shut down and power back on after 30 seconds. That last option has already saved us when graphics cards were no longer detected after a simple reboot;
• Miner actions: view the miner log (you have to click refresh, it is not automatic), stop or restart the miner.

With Remote Access, then "Hive Shell Start", you open a terminal in your browser. A few useful commands:

• nvidia-info: details of the NVIDIA cards (power levels, clocks…);
• help: the main commands;
• nvidia-driver-update: updates the NVIDIA drivers (AMD drivers come with each HiveOS version);
• hive-replace: a full reinstall of HiveOS, with the list of stable, beta and older versions and their drivers;
• gpu-fans-find followed by a card number: spins up that card's fans so you can locate it in the rig.

5. Updating HiveOS

In the web interface, the versions menu shows the installed version and the new ones, with their changelog. At the time of filming, the new version brought a T-Rex release able to partially unlock LHR cards. Choose the version, tick "Force reboot after install", then click "Upgrade".

6. Enabling the watchdog and setting the fans

The watchdog monitors the rig and steps in on its own when something goes wrong. We enable it on almost all our rigs:

• restart the miner after 2 minutes if there is a problem, then reboot the whole rig if it persists a few minutes longer;
• reboot if a graphics card is no longer detected;
• reboot if the system load average reaches 4, a value to adjust to the number of CPU cores;
• reboot if the hashrate drops below a threshold: 200 MH/s for this rig.

Very handy if you go on holiday or are rarely on site: the watchdog restarts the rig automatically. As for cooling, by default HiveOS stops the miner if a card goes above 90 °C. An automatic mode lets you aim for target temperatures, but we prefer to set the fan speed ourselves.

What about today?

Since Ethereum moved to proof of stake in September 2022, income from graphics card mining has collapsed. The features covered here are still the core of HiveOS, but the cards and settings in this article date from 2021.

For building the rig itself, read our article How to build your own mining rig and find our other guides on the tutorials page. To mine today, browse our ASIC miners, or contact us for advice before you buy.