When Asus, Noctua, and Nvidia team up, it’s clear the result won’t be a compromise. For those who find the RTX 5090 already too expensive, the GeForce RTX 5080 Noctua OC Edition offers a unique combination of a massive heatsink and three top-tier Noctua NF-A12x25 G2 120 mm fans. An air cooler capable of keeping a powerful 360 W card at low temperatures with the fans spinning at just under 700 RPM is something you rarely encounter.
Fan behavior and noise levels
Fan speed can be set from 30–100%, which corresponds to fan speeds in the range of approx. 530–1750 rpm (29.1–46.8 dBA). At the minimum speed, the sound level meter showed 29.1 dBA, which is below the lower limit of its measuring range.
With the BIOS in Performance (P) mode, once warmed up the fans ran across tests at 1100–1250 RPM, corresponding to measured noise of 33.4–36.7 dBA. The card isn’t exactly silent, but it’s still an excellent result because the TUF RTX 5080 posts similar noise in Quiet mode.
Q mode is much more interesting. In this mode, after the card warms up, fan speed across tests hovered around 670 RPM—already below what typical sound meters that start at 30 dBA can register. It’s certainly the quietest RTX 5080 I’ve tested and very likely the quietest air-cooled GeForce RTX 5080 on the market.
Cyberpunk 2077, RT Medium, 3840 × 2160 (P mode)
The first set of measurements comes from the Cyberpunk 2077 benchmark using the RT Medium preset at 3840×2160. It consists of eight loops. The final run is highlighted, and the average for the warmed-up card is calculated from that run.
The GPU is fully utilized in this test, and the card is pushed up against its power limit.
Once warmed up, the GPU clock ranges from 2782 to 2805 MHz, with an average of 2796 MHz.
The card’s power draw, based on the monitoring (light-green data series), sits just under 350 W. According to the specs, the power limit is 360 W. Last chart in this chapter indicates the boost is constrained by the power limit, which matches how all RTX 5080 cards I’ve tested behave.
The dark series in the chart shows total system power measured with a UT71E multimeter. Under load, the system averaged 466.1 W.
The average GPU temperature peaked at 62.2 °C over the measured interval, with memory reaching 66 °C. Given the card’s ~360 W power draw and measured noise of about 35–36 dBA, these are excellent results.
The massive heatsink has high thermal inertia, and the large fans ramp up gradually as well, so during the ~70-second load of a single loop the card doesn’t reach its peak temperature.
⠀
it seems even on Q mode it may get outside of the silent zone when pushed hard enough…
I know, I know, all the other cards are worse, but if even 3 of these fans struggle it means a lot about the power output of modern cards, doesn’t it?
What matters most is what a “quiet” graphics card means to each individual — it’s a pretty subjective thing.
At just under 700 RPM, it’s extremely quiet. At night, when everything is silent, you can hear it if you put your ear close to the card. From a typical distance of about a meter, even on an open test bench, I can’t hear it at all. But my hearing isn’t what it was twenty years ago.
However, considering the temperatures the card reaches, there’s still some headroom to lower the fan speed by adjusting the fan curve. Throughout the day, I’ll be running additional tests across the full fan-speed range in Cyberpunk and will add line charts of clock speeds and temperatures to the article.
Without V-Sync, the coil whine is still more noticeable under load than the fans themselves—even in Q-Mode—though it’s quieter than on other cards with the same performance. The tougher part might be keeping the CPU just as quiet and getting the heat out of the case without adding noise.
I haven’t precisely tuned G2 fans yet, but 700RPM would be somewhere in the range of upper limit, for G1 that was around 670, the Q profile seems to be willing to push above that if heated enough instead of limiting the heat to ensure fans won’t spin up, requiring manual tuning
which to be fair is expected of Noctua products as they don’t tune for absolute silence as general audience expects much less, but their high quality allows to achieve it with some extra effort
the coil whine though absolutely shouldn’t be there, that’s something most companies seem to neglect the most, I know it’s hard to control and there’s a ton of variance between specific items, but at this price point it’s unexcusable
I’d say even more: it’s unexcusable at any price point that isn’t explicitly cheapened cutting all costs for the last pennies, unfortunately… getting components and accessories that don’t suffer from that is nearly impossible, like… I’m not gonna try dozens of different units of a monitor to pick the one with least whine, right?
At least when it comes to temperatures shown in the monitoring, they’re fine—even at minimum fan speeds.
https://www.hwcooling.net/en/asus-geforce-rtx-5080-noctua-oc-ed-16gb-review-silent-monster/20/
Does GPU watercooling even make sense anymore? The graphics cards are becoming quieter and cooler with each new generation, even without deshrouding.
Yes, some cards are now quiet enough that the main source of noise is elsewhere. For example, in my test rig I often find the CPU cooler is louder, because I have the fan curve set to ramp up whenever any core goes over 80 °C.
It would be more appropriate to say that it depends on the circumstances. The key difference is that liquid cooling moves the GPU’s heat to a radiator mounted away from the card, making exhausting hot air easier when the radiator is properly positioned in the case. With an air cooler, you still need to figure out exhausting the heat from the case.
It’s a bit different from processors. With a CPU, the priority is cooling a small area—the CPU itself and the surrounding VRM (voltage regulation modules).
A good block on GPU usually reduces hotspots on the die and VRAM, but it can worsen cooling of the other components. But on a graphics card, there are many hot parts spread across the entire PCB. That’s why having fans directly over the hot zones usually works best. Full-cover blocks solve the problem of cooling components, but this is an expensive and complicated solution. If, instead, the manufacturer uses a cheaper AIO cooler on the GPU and a card-mounted fan to cool the other components, it usually isn’t dead silent. The pump can also be noisy unless the manufacturer keeps it under control.
Air cooling, in contrast, is simpler, cheaper, and generally more reliable. When paired with a large heatsink with enough surface area, it can be a very competitive option.
Personally, I prefer air cooling for its ease of use, consistent long-term performance, lower risk of failure and its reliability.
Thank you for the info.
I’m building a PC with Ryzen 9 9950X and RTX 5070 Ti. The CPU is cooled by a custom liquid loop with one 420mm and one 280mm radiator. I just can’t decide whether to include the GPU in the loop. The full cover waterblock by Bykski costs about $100.
Since a water cooled graphics card does not heat up other components, such as SSD, RAM, chipset, etc, I’ll probably go with the waterblock. My radiator space should be enough for that.
There’s actually a true specialist in our country who focuses on fully liquid-cooled builds. I’ll write to Gabo (from TichéPC) — you might get a response full of hands-on experience. 🙂