Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
CPU Thermal results without power limits
Without power limits enforced on Intel’s i7-14700K, the CPU will hit its peak temperature (TJ Max) and thermally throttle with even the strongest of air coolers. When the CPU reaches its peak temperature, I’ve measured the CPU package power to determine the maximum wattage cooled to best compare their performance.
The general exception to this comes with the strongest AIOs on the market, which can keep Intel’s i7-14700K under TJ Max. This is no small task as many liquid coolers, including Arctic’s Liquid Freezer III 360, fail this test and as a result the CPU will throttle to a certain extent.
Noctua’s NH-D15 G2 does well here, cooling an average of 240W during the course of testing. That’s a notable improvement of 10W cooling capacity compared to the original NH-D15.
When I tested the noise levels of the cooler, there was a bit of variance. While the chart above lists 44.6 dBA, it would be more accurate to say that it oscillated between 44.2 and 44.6 dBA.
Observant readers may notice that the chart starts at 35dBA, and might be worried that this starting point may cause the results to be deceptive. There are a few reasons that this is not something to worry about. First, noise level measurements are logarithmic. A person with typical hearing will perceive a doubling of noise for every 10 dBA increase, but the impact can vary from person to person.
Subjectively, I feel that a 5dBA increase feels like twice as loud – but I should add the disclaimer that I suffer from conditions that can make my ears more sensitive than others. 35dBA was chosen as the “zero” because it is the lowest level my noise meter can accurately measure.
To give you a better idea of how dBA readings correspond to noise levels in real life, I invite you to watch the video linked below. One note though: The author’s comparison for a refrigerator of 43 dB is based on older, louder models. Newer refrigerators may not operate nearly as loudly (depending on the model) - I recently purchased a Frigidaire refrigerator which has such a low noise level, I wasn’t sure it was actually operating when I first turned it on!
Thermal results with noise normalized to 38.9 dBA
Finding the right balance between fan noise levels and cooling performance is important. While running fans at full speed can improve cooling capacity to some extent, the benefits are limited and many users prefer a quieter system. With this noise-normalized test, I’ve set noise levels to 38.9 dba. This level of noise is a low volume level, but slightly audible to most people.
Both Noctua’s NH-G2 and G1 had somewhat underwhelming noise-normalized results, trailing Thermalright’s soon-to-be-released Peerless Assassin 140 by 13-16W.
165W CPU + 275W GPU results
I’ve recently updated my testing methodology to better represent the scenarios a user might actually encounter in real-life use. The first of these new tests is designed to emulate a CPU-intensive game with a full GPU load. In this test, there was a fairly significant 2C improvement compared to the original NH-D15.
Because the CPU averages 79C during the course of testing, the noise of the fans is no different than with a full workload, because my motherboard’s fan curve engages full fan speeds at 80C, and a 1C difference doesn’t make any appreciable noise drop. This is a scenario where, in theory, a quality AIO should run much quieter than an air cooler.
Observant readers may notice a change in how our temperatures were reported. Previously I reported temperatures in “Delta over Ambient,” which is often considered a more accurate and scientific way to report thermal results.
However, I’ve found reporting temperatures in this manner often confuses some readers. I’ve had many comments wondering why the temperatures a user experienced with a particular product were 20 degrees C or more higher compared to what was reported in our charts, and as a result had to explain what delta over ambient means to them.
I feel that reporting temperatures in delta over ambient temperature is most important when there is significant variance in the base temperature. But because I strictly regulate the ambient temperature, I don’t feel that it is necessary to report temperatures in delta over ambient form - any results obtained that are outside of an ambient temperature of 22.7-23.3 C are considered invalid.
110W CPU + 275W GPU results
120W Cinebench results
With my previous testing methodology, I used to record the temperatures and noise measurements using a 125W CPU-only test and I would emphasize how temperatures do not matter in this scenario.
As such, outside of exceptionally good or bad thermal performance, for this section I’m only going to report noise levels – because that’s the only data that matters for this test.
Noctua’s noise level here wasn’t particularly loud, at only 40.9 dBA, but it’s louder than competitors – many coolers will run at 38.2 dBA or less in this scenario.
Conclusion
Noctua’s NH-D15 G2 is a marked improvement over the original NH-D15, with CPU temperatures up to 3C cooler compared to its predecessor. Noctua also almost certainly has the best customer support in the PC cooling industry. However, since the launch of the D15, multiple companies have released excellent air coolers that cost much less, making the high $149.95 price of Noctua’s cooler way too high to justify. Given the competition from Thermalright and others, no one should consider paying such a high price for an air cooler.
Nvidia, AMD, and Intel all invest in light-based communication networks powering next-gen chips — Ayar Labs gets $155 million in funding
Fanless Ryzen 9 9950X seemingly outperforms the same liquid-cooled chip in Cinebench R23 — flagship Zen 5 CPU hit 95ºC at full load
AMD to release the Krackan at CES — CPU rivals Intel Lunar Lake in new benchmarks, Ryzen AI 7 350 takes on the Core Ultra 7 258V
