After waiting 2-3 weeks, I have finally received my new graphics card, Palit GTX 1070 Super Jetstream, today, which I installed in my main rig to play triple-A games in 1440p at 60Hz. I'm far from a serious gamer and I tend to be more or less satisfied with 50+ fps in 1080p mode but, at the same time, I also highly prefer (actually, crave) visual fidelity and eye-pleasing textures, thus I opted for GTX 1070 which I believe suffices for 50-60 fps gaming in 1440p with a caveat of sacrificing less impactful (and almost indiscernible, subjectively speaking) settings such as "ultra shadowing", "msaa 16x", "pcss", and whatnots.
* Testbed (?)
The chassis is SilverStone TJ08-E and the CPU is i5-6600K. The front intake fan is Noctua NF-A14 PWM and the rear exhaust fan is Noctua NF-S12A PWM. The CPU temperature is below 50 Celsius all the time and the speed of the two CPU fans (Noctua NF-F12 PWM) on Noctua NH-U12S is within the range of 300-600 rpm.
Before diving into the experiment part, let me try to do minimalistic justice to the LED of the graphics card, which can arguably be changed to 16.8 million colors.
* First experiment with intake fan (Noctua NF-A14) @ 1100 rpm
I ran a triple-A game for about 20 minutes in SilverStone TJ08-E. It is worth noting that TJ08-E has rather poor airflow, particularly for the graphics card because the "hot air" from the graphics card gradually accumulates right above the graphics card whose air flow is utterly obstructed by the PSU. You can take the following result as some sort of the "near" worst case scenario. The exhuast fan Noctua NF-S12A is almost non-existent and I actually forget the typical rpm of this fan. I suppose the figures are around 300-500 rpm. However, The intake fan Noctua NF-A14 is rotating at a reasonably high number, i.e., 1100 rpm., at which point the fan is quite noisy.
I do not have a liking to manual overclocking and thus did not overclock this card (I do not overclock even my i5-6600K, which I went for because it was cheaper than i5-6600 non-K at the time of purchase). The core clock stayed exactly at 1961.5MHz for the best part of the entire experiment. The temperature was around 70 Celsius where the two fans were rotating at 1085 rpms on the average. I could not make the graphics card reach the power consumption of 100%. I suppose that the power limit is set to a generous figure so that it can possibly bottleneck the performance only on very rare occassions. The average power consumption was around 87.5%. Most importantly, these rpm numbers are absolutely outstanding as compared with other graphics card in GTX 1070 market. Althnough I can't recall the figures, I think the rpms of Asus Strix and MSI Gaming are around 1300-1600 rpms. For what it's worth, the two fans on my Asus GTX 970 Strix OC rotate at 1550-1650 rpms in full load. As far as my experience goes, Palit GTX 1070 Super Jetstream is THE quietest graphics card I have ever used and I firmly believe that it is indeed so among all AIB GTX 1070 cards available as of now. For your information, I'm unnecessarily sensitive to all sorts of noise partly because I cannot concentrate on working for a longer duration in noisy environments.
While there are several other versions of GTX 1070 which are even more aggressively overclocked, Palit GTX 1070 Super Jetstream has rather high base clock 1632 MHz (1633 MHz in GPU-Z), and the average clock in full load is almost 2GHz (1961.5 MHz). In this light, I dare to conclude that this card strikes a perfect balance between performance and silence (and cost) although the definition of silence varies greatly depending on your acoustic expectations.
* Additional Notes/Observations
1. I forgot to mention the ambient temperature, which was about 24-25 Celsius. I think it's close to 25.
2. I also conducted a similar experiment without the side panel of SilverStone TJ08-E. Though I did not record numbers, the average rpm was slightly below 1000 rpm, and the temperature was 1-2 degrees lower. "Less than 1000 rpm" result indeed corroborates the review of Hexus.net where a tiny bit superior version called "Palit GTX 1070 Gamerock Premium" was reviewed. I personally think the likes of these review results distort our understanding of actual performance because most people have closed chassis, resulting in much poorer ventilation.
* Fan curve
I tinkered with its two fans and summarized the fan speed in the following graph.
Although I must admit that my experience with fans is relatively limited as compared with other experts in this forum because I have used so far only Noctua fans and a handful of Arctic fans, I have to say that the fans are really solid, sturdy and decent. Surely, at 100%, the fans are quite noisy but the overall acoustic characteristics are rather reticent (I know it's a very subjective expression but ...) and do not contain high frequency components. I could not hear any clicking noise, either. As compared with other fans on my past graphics card, these fans are definitely superior to them in terms of high-frequency component and clicking noise, or the lack thereof.
The maxmimum fan speed is 2500 rpm and the miminium fan speed is approximately 500 rpm which is achieved at 18-20%. I also noticed that the fan blades are very sharp, in fact, "dangerously sharp", which I suspect, contributes to higher air pressure onto the heatsink/cooler and its overall acoustic performance.
Once again, subjectively speaking, although the two fans are clearly audible all the time (as I said eariler, all fans in my main computer are rotating at 300-600 rpm), they become impressively quiet below 1000 rpm, which I think is a sort of personal threshold for me, above which I can notice the existence of the two fans, below which I cannot perceive the fan noise unless I strive to hear them.
* Yet another observation
I tried to play some other triple-A games (by which I mean that they are very demaning in graphical terms). Under my preferred setting where the intake fan Noctua A14 rotates at about 700-800 rpm, I found out that the typical fan speed of the graphics card does not often exceed 1000 rpm simply because:
(i) most games do not consume more than 80-85% of the maximum power limit;
(ii) the very changeable nature of gameplay (such as looking up maps and roaming around) makes it very unlikely that the graphics card takes such a long-term sustained stress.
Thus, you can indeed expect to experience "less than 1000 rpm" fan speed with this graphics card during normal gaming sessions.
* Second experiment with intake fan (Noctua NF-A14) @ 1100 rpm and GPU fans @ 1501 rpm
A few benchmark sites are reporting that the fans of MSI Gaming X rotate at as low as 1330 rpm:
http://hexus.net/tech/reviews/graphics/ ... x/?page=10
However, I suspect that the fan speed in real scenarios with closed chassis is close to 1500-1600 rpms. Thus, under the exactly same set of paramters used in the first expeirment, I conducted another experiment with the fixed GPU fan speed of 1501 rpm as shown in the following screenshot of GPU-Z. Note that the ambient temperature in this experiment is 26 Celsius which is slightly higher than 25 Celsius in the first experiment.
This is an astoundingly excellent performance! The average temperature was 62.3 Celsius where the average fan speed was 1501 rpm and the GPU load was 98%. This result clearly demonstrates that this graphics card has far superior heatsink/cooler to any other graphics card in the GTX 1070 market.
* Coil whine
This part of the review is frankly worthless because the coil whine depends on the place/factory/country where the graphics card was manufactured. As far as my specific sample of Palit GTX 1070 Super Jetstream is concerned, I highly suspect that there is no trace of coil whine whatsoever. The trouble here is that my PSU, Corsair RM650, usually generates a very faint (practically, inaudible) trace of coil whine under load, thus I cannot tell with utter certainty which coil while is produced by the graphics card or Corsair RM650.
On the other hand, I did clearly hear noticeable coil while when I did not limit the fps, which was hitting 500-2000 fps. As far as I can gather, I suppose the coil whine with unlimited fps is rather a common phenomenon observed across all Nvidia graphcs cards. In normal scenarios where the fps is limited by the refresh rate of your monitor, you will not be able to hear any coil whine from this graphics card.
* Subtle Differences On GPU Fans Between Palit and Gainward Cards
The general consensus on the differences between Palit and Gainward GTX 1070/1080 cards is that they are almost the same, apart from the "cosmetic differences" such as the fan shrouds and "specification differences" regarding base/boost clocks across different versions of Palit and Gainward cards. Particularly, when it comes to thermal/acoustic designs, as far as I can gather from various pictures/videos and my observations on my own Palit GTX 1070 card, they use exactly the same cooler/heatsink for all Palit and Gainward GTX 1070/1080 cards.
However, as for the fans on the aforementioned identical heatsink, as observed by Lodestar in the following thread, there are subtle yet intriguing differences:
viewtopic.php?p=602138#p602138
In a nutshell, the fan blades in Palit cards take a curly shape as shown in the following picture (my card):
In comparison, the fan blades in Gainward cards are straight, the ramification of this difference is unclear at the moment. As of now, all I can say is that the fan blades of Palit cards are ruthlessly sharp.
Secondly, as pointed out by Lodestar, the dual GPU fans on Palit cards are indeed "counter-rotating", implying that the left fan rotates in the counterclockwise direction while the right fan rotates in the clockwise direction. Palit claim that this special feature is supposed "to reduce airflow conflicts and effectively improve overall cooling performance", which I think makes some sense and might contribute to the improved cooling and acoustic performance. However, I presume that the overall impact is minimal. As compared with counter-rotating fans in Palit cards, the GPU fans in Gainward cards rotate in the same direction. Thus they do not have this feature.
How low will the fans go rpm-wise? Have you tried using a fan utility to trade off gaming temps for fan rpm?
At any rate, I suppose that ~1000 rpm is a reasonably quiet speed for almost all people during gaming sessions.