GPU settings
Also called: power limit, core clock, memory clock, GPU tuning
GPU settings (power limit, core clock, memory clock) adjust a graphics card's power and frequencies to balance speed against power consumption.
A graphics card is tuned with three main parameters. The power limit sets the maximum power the card can draw, in watts or as a percentage of its stock value. The core clock controls the frequency of the graphics processor. The memory clock controls the frequency of the video memory. These settings are made with the driver's tools, dedicated software or, on a rig, from HiveOS.
The right setting depends on the workload. Some mining algorithms mainly stress the memory and benefit from a higher memory clock, while others mainly stress the core. In almost every case, lowering the power limit saves a lot of watts for a modest loss in performance: it is the first lever for improving efficiency. The same reasoning applies to artificial intelligence workloads, where a power-limited card often loses little speed.
Move in small steps and test stability after each change: an overly aggressive setting causes errors, driver crashes or reboots. Keep an eye on core and memory temperatures, as well as fan speed. On a product page, look at the card's reference power consumption, the amount of VRAM and the memory type: these determine how much tuning headroom you have.
Our GPU profitability calculator estimates what a graphics card rented out for artificial intelligence could bring in, based on the number of GPUs and your electricity price.