Prime RTX 5070 Deshroud: Noctua NF-A12x25 vs. NF-A12x25 G2

Many ASUS graphics cards let you remove the fan shroud without breaking the warranty seal. We did that last time with the ASUS Prime RTX 5070 and swapped its fans—noisier than usual at low speeds—for Arctic P12 Slim fans. However, standard 25 mm fans usually provide higher static pressure and can cool more quietly. Let’s see how the card performs with the top-of-the-line Noctua NF-A12x25 and its successor, the NF-A12x25 G2.

Before looking at results with the pair Noctua fans, we’ll start with noise measurements. In the first chart you can see how the card performed with the stock fans. Fan control lets you adjust their speed between 30–100 %, corresponding to 700–3200 rpm. Because one of the fans had a faint ticking sound, even at just 700 rpm the cooler registered 32.1 dBA. This pushed the overall noise of the card higher across the entire range.

In the next chart are the noise levels with two Noctua NF-A12x15 first-generation fans. When connected to the control panel, they start at 10 % PWM and top out at 98–100 % PWM. The chart shows the faster of the two fans; the other ran a few dozen rpm lower.

At lower speeds the noise curve is flat. That’s because the fans are quieter than the sound meter’s threshold (30 dBA). They’re effectively near-inaudible, and the meter’s own background noise is higher than the fans themselves.

 

We’ll check how the card’s behavior changes across the full fan speed range. For time reasons I used short Cyberpunk benchmark loops with the RT Medium profile at 2560 × 1440 resolution with DLAA.

Before the main batch, one warm-up loop with four runs at 45 % fan speed was executed. That part isn’t shown in the charts below, which begin with the first measurement at 25 % fan speed. Going lower isn’t useful, as with low rpm and GPU temperatures over 85 °C, the card’s regulation starts cutting GPU power sharply, lowering clocks and performance.

Fan speed was then raised in 5 % steps from 25 to 100 %. Each step included four benchmark loops. Gaps between runs were only a few seconds, so the system stayed heated during testing.

I’ll take values from the fourth loop, occasionally from the third if the last one had anomalies.

This chapter shows how the card behaved across the entire fan range. Here you’ll see simple figures; detailed values and per-step graphs will appear in interactive charts in the next chapter.

At the bottom of the tested range, 25 % fan speed, the GPU peaked at 86.3 °C with an average of 84.3 °C. At maximum speed, the average temperature dropped to 61.2 °C.

The next chart shows fan rpm in light blue. The fans were connected via a splitter to an external Corsair Commander Pro controller and managed through Fan Control. Each step was set manually to a fixed output.

At 25 % they spun at about 500 rpm, while at 100 % they hit 2135 rpm.

The purple curve is mainly for reference. It shows what fan speed the card’s BIOS would have set automatically. Since the fans weren’t connected to the card but controlled externally, it doesn’t match the real rpm, but it illustrates how aggressively the automatic control would ramp up at certain temperatures.

At minimum rpm, when the GPU exceeded 85 °C, the regulation (needlessly) pushed for maximum speed. As temperatures dropped, it scaled back and demanded lower fan output.

The following chart shows GPU temperatures (red) and overall CPU load (blue). As GPU fan speed increased, airflow through the case improved, which slightly lowered CPU temperature.

Apart from the first run, where the CPU load was unexpectedly higher, the average CPU temperature declined gradually from 60.2 °C to 56.3 °C.

Performance differences are not dramatic. With a cooler chip, clocks rise slightly and so does performance, but only by a few tenths of a frame per second.

This is clearer on GPU clock speeds than on fps. Early on, as the card began to overheat, clocks dropped more noticeably. Once cooling was sufficient, clock speeds rose more steadily. Between 30 % and 100 % fan speed the difference was about 100 MHz.

For verification we also have charts of power draw for the whole system and for the card itself, as reported by monitoring. In the first run, when GPU temperatures were already too high, you can see that GPU boost throttled the card’s performance. In later runs, the card clearly operated right at its power limits.

⠀