Cooling & Noise
Cooling Solution & Backplate
By removing the cooler's shroud, we get a better look at its two shells: an outer one made of light metal and the inner one of red ABS. The fans are mounted to the plastic cover, but not acoustically decoupled.
Although PowerColor coats the inside of its backplate in black so that thermal energy can be absorbed more easily, there are also no large surfaces to cover sensitive areas of the board and hinder passive ventilation.
All of the real cooling happens up front, so there's nothing more to say about the Red Devil's back side.
|Cooling System Overview|
|Type of Cooler||Air cooling|
|GPU Cooling||Nickel-plated copper heat sink|
|Cooling Fins||Aluminum, vertical alignment|
Narrow configuration, not inclined
|Heat Pipes||3x 6mm + 2x 8mm |
Nickel-plated copper composite material
|VRM Cooling||Four phases via an extra VRM sink in the cooler|
Three phases via GPU sink
|RAM Cooling||Cooling of HBM2 modules via heat pipe|
|Fans||3x 9cm fans (9.2cm opening), Nine blades|
No semi-passive fan control
No cooling function
In addition to the sink for some of the voltage converters, two 8mm and three 6mm pipes made of nickel-plated composite material are responsible for transporting thermal energy from the GPU to different parts of the finned cooler.
The GPU heat sink has a rather rough surface, yet performs its task at least as well as the polished plate on Asus' ROG Strix Radeon RX Vega 64.
Fan Curves & Noise
Despite PowerColor's claims to the contrary, our measurements show that there is no semi-passive fan mode. Like AMD, the company keeps its fans spinning continuously. A 33 to 34% PWM reading translates to 1100 to 1200 RPM with the GPU idling between 25-38°C. There doesn't seem to be an extra controller chip on the board that'd stop the fan below a certain threshold.
After a period of intensive cooling during the warm-up phase, PowerColor's fans slow down quite a bit as temperatures stabilize. It's also obvious that the card is acutely sensitive to operating in a closed case versus an open bench table. In order to keep thermals as constant as possible, the fans always spin quite a bit faster inside of a chassis, even after the temperature target is reached.
Little changes during our stress test. While the cooler and its three fans work well, they also have to move a lot more air over the sink in our closed case.
This card does have enough cooling headroom to dial the fans down a bit. You can even raise the temperature target to 70°C, allowing the volume-optimized thermal solution to operate more quietly.
|Fan RPM & Noise Measurements|
|Fan RPM, Open Test Bench, Maximum||1861 RPM (Peak)|
|Fan RPM, Open Test Bench, Average||1496 RPM (Warmed up)|
|Fan RPM, Closed Case, Maximum||2561 RPM (Peak)|
|Fan RPM, Closed Case, Average||1617 RPM (Warmed up)|
|Noise (Air) Range||32.2 (Minimum) to 51.4 dB(A)|
|Noise (Air) Average||34.2 dB(A) (Warmed up, Open bench table)|
38.3 dB(A) (Warmed up, Closed case simulation)
|Noise (Air) Idle||32.2 dB(A)|
|Noise Characteristics / Subjective Impressions||Low-frequency bearing noises|
Some motor noises below 1 Hz
Moderate air and turbulence noises
Slight voltage converter noises
The 34.2 dB(A) we measured in an open test bench is undoubtedly what other reviewers find so praiseworthy. However, if the fans are set to the speeds observed in a closed case, we blow right past 38 dB(A). Here's what that looks like in the form of a high-resolution frequency spectrum analysis:
With PowerColor's Red Devil RX Vega 64 installed, you'll want to pay close attention to airflow inside of your chassis. It's easier to exhaust waste heat with quiet case fans than to optimize the graphics card for a different target temperature. Or you could do both and win twice. At least that way you can keep noise output under 35 dB(A).
MORE: Best Graphics Cards
MORE: All Graphics Content