Temperatures And Noise
Does the new cooler do the trick? It most certainly does. The modified GeForce GTX Titan Black runs at 73 degrees Celsius, or 10 degrees Celsius lower than Nvidia's reference model. Again, that's effective enough to keep the card from hitting its thermal limit, even under full load. I even tried bringing the modified temperature down further using a manual fan profile under Gigabyte’s OC Guru software.
In an open build, the noise level rises quickly and noticeably, so the extra cooling isn't worth the tradeoff. But I also felt it necessary to simulate a worst-case thermal situation in a cramped case. So, I used the same chassis Thomas reviewed in AIO DRN-STN Review: A Gaming All-In-One With A 120 Hz Display, filled to the brim with case fans spinning slowly.
The VRM temperatures were measured with a probe between the GPU and thermal tape, and the board temperatures on the back were gathered using an infrared thermometer. The measurements marked with (TH) are those that used my custom fan profile with elevated rotational speeds.
|Header Cell - Column 0||Bench Table(Open Installation)||Corsair Obsidian 900D||AIO Drone Station(Case Fans at Minimum)|
|GPU||73 °C||75 °C||82 °C|
|GPU (TH)||62 °C||63 °C||70 °C|
|VRM||87 °C||89 °C||100 °C|
|VRM (TH)||81 °C||84 °C||91 °C|
|Board||85 °C||86 °C||96 °C|
|Board (TH)||81 °C||83 °C||91 °C|
I measure graphics card noise with a calibrated high-quality studio microphone (supercardioid) that is placed perpendicular to the center of the board at a distance of 50 cm. The combination of this distance and the microphone’s very strong directionality represents a compromise between minimizing the noise generated by fan turbulence and minimizing the impact of ambient noise, which can never be completely avoided.
Sound dampening can help with ambient noise, but it can never be completely eliminated. This is why we’re performing these measurements predominantly at night.
As long as you own a suitable PC case that’s not too packed and offers decent airflow, then Gigabyte's WindForce cooler should perform well. It reaches a maximum of 1900 RPM under full load, which keeps the card surprisingly quiet.
Our manually-created profile is quite a bit louder, of course. But the noise level is still only a little worse than Nvidia's reference cooler, even as our custom fan ramp keeps the card 20 degrees Celsius cooler. To be fair, the only reason you'd want to try something similar would be if you were using a chassis with terrible airflow. Otherwise, there's no need to tinker.
The fan curves are only part of the take, though. Here's the numerical data in the form of a handy table directly comparing the different configurations:
|Header Cell - Column 0||Idle||Gaming||GPGPU|
|Nvidia's Reference Cooler||31.2 dB(A)||41.7 dB(A)||43.5 dB(A)|
|Gigabyte's WindForce Cooler, Stock||30.4 dB(A)||35.4 dB(A)||35.7 dB(A)|
|Gigabyte's WindForce Cooler, TH Profile||30.4 dB(A)||42.1 dB(A)||44.9 dB(A)|
The modified cooler is definitely successful, given that the GeForce GTX 780 Ti WindForce OC, which performs around the same level, produces just over 40 dB(A) with its older cooler and non-optimized fans.