Optimization of fan load via core clocking of a triple fan GPU system

Throughout the decade, the graphics processing unit (GPU) has seen numerous innovative advancements. The processing capacity of a GPU may compete with that of an existing Central Processing Unit (CPU) when it comes to running high-profile software. Any electronics that pass current through it, on th...

Full description

Saved in:
Bibliographic Details
Main Author: Hanif, Haziq Haiqal
Format: Thesis
Language:English
English
Published: 2021
Subjects:
Online Access:http://eprints.utem.edu.my/id/eprint/26103/1/Optimization%20of%20fan%20load%20via%20core%20clocking%20of%20a%20triple%20fan%20GPU%20system.pdf
http://eprints.utem.edu.my/id/eprint/26103/2/Optimization%20of%20fan%20load%20via%20core%20clocking%20of%20a%20triple%20fan%20GPU%20system.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Throughout the decade, the graphics processing unit (GPU) has seen numerous innovative advancements. The processing capacity of a GPU may compete with that of an existing Central Processing Unit (CPU) when it comes to running high-profile software. Any electronics that pass current through it, on the other hand, will generate heat. The biggest issue with GPUs is the thermal issue. The best GPU performance requires the lowest fan speed and core temperature while maintaining the most efficiency (hashing power). The goal of this study was to use the Design-Expert software's optimization function to get the best response at a specific core clock and memory clock. The reaction was recorded when the GPU was receiving a consistent quantity of power. The GPU used is the ASUS TUF RTX3060 OC, which is a triple fan GPU. The link between GPU responses (fan speed, core temperature, hash rate) and clocking (core and memory) was discovered. Central Composite Design (CCD) came up with a single equation for each fan speed, core temperature, and hash rate as a result of this. The optimization method then proposes a number of new clock settings that outperform the existing setting. The best core and memory clock were chosen for the validation and confirmation phase. The validation results show that the expected response was accurate, with a margin of error of less than 2%.