Temperatures
(*) Noise level is estimated. We changed the noise-testing methodology since those reviews.
Testing notes & interpretation- GPU temperatures listed here are based on GPU-Z measurements of the on-chip temperature sensor.
- We report these GPU temperatures under a constant load for ease of comparison, as well as an idle state most end users will experience often. This combination helps dictate cooling needs and provides context for how well the thermal solution performs.
- Please note that GPU temperature is contingent on a variety of factors. Some, including clock speed, voltage settings, cooler design, and production variances, are beyond the control of the end user. Others, such as ambient temperature, case design, and airflow pathway affecting the GPU, can be mitigated to certain extents.
- The data in the table above shows results for similar cards, achieved in identical conditions during previous TechPowerUp reviews.
Thermal Analysis
For this test, we first let the card sit idle to reach thermal equilibrium. Next, we start a constant 100% gaming load, recording several important parameters while the test is running. This shows you the thermal behavior of the card and how the fans ramp up as temperatures increase. Once temperatures are stable (no increase for two minutes), we stop the load and record how the card cools down over time.
We also ran the same test for the "quiet" BIOS:
Fan Noise
Noise Testing Details- The tested graphics card is installed in an open bench system. The system does not emit any noise on its own, all components are passively cooled.
- All cards are loaded with the same test: a custom-engineered game-like load in Unreal Engine 4 that always renders the same frame at 60 FPS to minimize variance.
- To ensure 100% loading of the GPU, the render complexity is set dynamically so that each card operates at peak performance without being CPU limited.
- We're not using Furmark because it creates too high a load that forces fans to run faster than what is encountered during typical gameplay.
- Noise results of other cards on this page are measurements of the respective reference design.
- We let the card heat up and wait for fan speed to no longer change. Depending on the fan-control algorithm, this can often take quite some time.
- Noise level measurements are conducted using a 1/2" calibrated measurement microphone that records onto the host PC using a digital audio processing pipeline. The microphone is recalibrated at least once a month.
- Measurement distance is 50 cm. Our previous noise-level measurements were at 1 m. We reduced the distance to be able to better distinguish acoustics of the very quiet graphics cards.
AMD finally introduced fan stop with their Radeon RX 6000 Series reference designs, so the RX 6900 XT turns off the fans in idle, browsing, and light gaming. Perfect!
Gaming noise levels are much higher than the AMD reference design, and also higher than the MSI Gaming X Trio. The Gigabyte Gaming Pro OC is still slightly louder than the ASRock OC Formula. While I can understand that the cooler has to work extra hard with the default "performance" BIOS, I feel like the "quiet" BIOS has lots of room for improvement. It runs lower clocks and voltages, which results in low temperatures, so there is plenty of headroom to quieten down the fans.