Secrets of the D14 - Ch 4: Fans for Very- and Ultra-Quiet Operation (56k warning)

Secrets of the D14 – Chapter 4: Fans for Very Quiet and Ultra-Quiet Operation

Now we get back to comparing what fans do. In this chapter we will explore the effects of various combinations of fans in the D14. We will look at Very Quiet operation, with decibel readings in the upper teens to low twenties, and Ultra-Quiet operation, with decibel readings in the mid- to upper teens.

Contents:

4.0 Introduction

4.1 The setup

4.2 The Fans

4.3 Center Fan: Noctua NF-P14 with LNA

4.4 Center Fan: Noctua NF-P14 with ULNA

4.5 Center Fan: Scythe Kaze Maru 2 (a.k.a. Slip Stream 140) SM1425SL12L, nominal 800rpm

4.6 Miscellaneous fan setups

4.7 Final and Additional Thoughts


Related:

Chapter 1: Solo 140mm fans in the center position

Chapter 2: Fan Position

Chapter 3: In and Out Games

Chapter 5: Quiet Operation


Edited by ehume - 11/5/11 at 6:42pm

4.0 Introduction

The Noctua NH-D14 is an outstanding heatsink. What is most astonishing is that it does a lot of cooling with little noise. The stock setup, for example, will cool my 4GHz overclocked Intel i7 860, holding it down to under 51 degrees Celsius above ambient temperatures while making the equivalent of only 26 to 28 dB noise at 1 meter, depending on what direction you measure it from. Measured from the side it makes the equivalent of 26 dB sound Pressure Level (SPL). A SPL of 26 dB at 1m is quiet operation. But what can the D14 do when it is running even quieter?

I decided to explore the effects of the fan that comes with the D14, the Noctua NF-P14. This is a quiet fan, but it makes too much noise for what we are doing in this chapter. So I attached first one then the other of the “Ultra-Low Noise Adapters” that come with the heatsink. Now, when these resistor wires (that’s what they are) come with individual Noctua fans they are called a “Low Noise Adapter” (LNA – the adapter with the black socket) and an “Ultra-Low Noise Adapter” (ULNA - the adapter with the blue socket). When you put the ULNA on the 120mm NF-P12 push fan and the LNA on the 140mm NF-P14 center fan, both run around 900 rpm. But when you put the ULNA on the P14, the fan runs 750 to 800 rpm. The different rpm’s may make a difference in noise and cooling performance. We’ll have to test it and see.

Next, my preliminary studies with various 140mm fans in the center position of the cooler showed me that the Scythe Kaze Maru 2 (AKA Slip Stream 140) model with a rated speed of 800 rpm should do a good job working with other fans to cool the D14 while making very little noise. Can it operate in an ultra-quiet or very quiet manner and still cool effectively?

I also take a peek at the Thermalright TY-140 using the LNA and UNLA that come with the D14. Is this fan so much quieter than the P14 that it is worth buying just for its quietness? I ask this question because the TY-140 will be the star of the upcoming testing with PWM fans since it is the only PWM 140mm fan available to me ATM.

And last, I look at some miscellaneous setups to see if we can have very quiet or ultra-quiet operations that still allow us to cool a very hot chip.

Be sure to revisit the data over the next few weeks. I find that I learn more each time I return to a dataset. More jumps out at me. I’m finding the same to be true with this data. So do come back.




Edited by ehume - 8/15/11 at 8:41am

4.1 The setup:

• CPU: Intel Core i7 860 overclocked to 4004MHz (22x182MHz), Vcore = 1.312 to 1.328v, HT and LLC enabled. This chip runs hot enough that a Megahalems needs a San Ace 9G1212H1011 to cool it down to 50c over ambient.
• Motherboard: Gigabyte GA-P55A-UD3P.
• RAM: 4x2GB GSkill Ripjaws DDR3-2000 running at 1456MHz (8x182MHz). BTW - Don’t get this RAM. Get low profile RAM. Low profile RAM will fit under any heatsink and under any fan. RAM with even mildly elevated RAM heatsinks like the classic Ripjaws can require you adjust the front push fan upward. Other RAM can have downright silly coxcombs. Don’t get them.
• Heatsink: Noctua NH-D14. Heatsink fans were powered directly from the PSU, each with an rpm line to one of the fan headers on the motherboard. Fan RPM measurements were read from the ET6 utility. These numbers aren’t terrible accurate; they tend to fluctuate by up to 3-4%. But you get an approximate notion of how fast the fans are going. The rpm relates to noise and cfm, and hence the TOA obtained.
• PSU: Seasonic X-750, enclosed by the wooden test bench that the motherboard sits on.
• Ambients: TEMPer USB digital thermometer - reads ambient temperatures from a fixed point 30cm (1 foot) in front of the front D14 fin stack. It measures the temperatures to the closest 0.1c six times a minute and outputs the results to a CSV file. Temps for each run are averaged.
• SPL: Tenma (house brand for MCM Electronics) 72-942 digital SPL meter - displays its readings to the nearest 0.1 dBA. I record the sound pressure level measurements to the nearest 0.5 dBA. The SPL meter specs say it is accurate to within 1.5 dBA at 30 dBA or above. The SPL meter does display readings under 30 dBA, but they get less reliable the further they are from the 30 dBA point. In any case, the readings were taken 10cm from the side of the D14, so no readings fell under 30 dBA. Results were correlated with 1m results by subtracting 20 dBA from the SPL readings. Since this correlation is not absolutely accurate, the readings must be taken as relative results, which only approximately correlate to SPL measurements actually taken at 1m. Remarkably, the SPL readings were consistent to within 1 dBA for any particular fan setup, usually within 0.5 dBA.
• Stress testing was done with OCCT v3.10, using the Linpack module, 90% memory, hyperthreaded. OCCT generates CSV files that record CPU temps once a second. Each run was 45 minutes long. Since Linpack “rests†periodically during each run, and because we are not interested in including resting temps in our calculations on how hot the chip gets, only the highest temps were harvested. This was done by rank ordering the readings from the first core of the i7 860, which runs the hottest of the four cores. The highest temps made a figurative plateau (the graph of the temps looks like a series of plateaus). I included in the average only those temps at and above the lowest temp where the duration was 10% of the run or longer. This method produced consistent results, with subsequent runs always within 1c of each other, usually within 0.5c, and often exactly the same. Re-checking confirmed result consistency.
• The entire assembly sat on a home-built wooden test bench. This allowed close approach to the heatsink for SPL measurements, quick fan changes and overall babysitting.
• Fan sets were normally a 140mm fan in the center position with a single 120mm in push or a pair of matching 120mm fans in push and pull at the front and back of the heatsink. Sometimes I ran a pair of 120mm fans with the center position empty or containing a shroud. After all, in this chapter we are trying to find the best, quietest cooling we can do with the D14.

Edited by ehume - 8/15/11 at 8:43am

4.2 The Fans:

• AeroCool Shark 12cm (16 blades, fluid dynamic bearing), with and without its 7v adapter. Nominal 1500 rpm (12v) and 800 rpm (7v).
• Coolink SWiF2 1201 (11 blades, hydro dynamic bearings) nominal 1200 rpm; also used with a fan controller set to 800 rpm to emulate a SWiF2 1200. With a grounded SWiF2 120P (see next) I had the equivalent of a pair of SWiF 1200’s, which are nominally 800 rpm fans.
• Coolink SWiF2 120P (11 blades, hydro dynamic bearings) with the PWM line grounded. This causes the fan to run at its preset minimum. Coolink specs list the minimum as 800 rpm, so the SWiF2 120P with grounded PWM line emulated a SWiF2 1200, effectively giving me two Coolink 800 rpm fans to test.
• Gelid Silent 12 (seven blades, hydro dynamic bearings) nominal 1000 rpm.
• Nexus Real Silent D12SL-12 (seven blades, sleeve bearing) nominal 1000 rpm. As you might deduce from the model number, this is a fan that Yate Loon makes for Nexus. In order to emulate a second Nexus, I took a Yate Loon D12SL-12 and used a fan controller to drop its speed to 1000 rpm. This allows us to see more or less what two Nexus Real Silent fans would do.
• Nidec Gentle Typhoon AP-11 (nine blades, ball bearings) nominal 500 rpm - two
• Nidec Gentle Typhoon AP-12 (nine blades, ball bearings) nominal 800 rpm - two
• Nidec Gentle Typhoon AP-13 (nine blades, ball bearings) nominal 1150 rpm - two
• Nidec Gentle Typhoon AP-14 (nine blades, ball bearings) nominal 1450 rpm - four, @12v and 5v
• Nidec Gentle Typhoon AP-15 (nine blades, ball bearings) nominal 1850 rpm - two, @5v
• Noctua NF-P12 (nine blades, “SSO-Bearingâ€) nominal 1300 rpm - two
• Noctua NF-P14 (140mm, nine blades, “SSO-Bearingâ€) nominal 1200 rpm - two
• Two sets of LNA and ULNA adapters for the Noctua fans
• Sanyo Denki San Ace Silent 9S1212L401 (seven blades, ball bearings) nominal 1500 rpm - six
• Scythe Kaze Maru (140mm, 11 blades, sleeve bearing; abbr KM-1900) nominal 1900 rpm - two, @5v
• Scythe Kaze Maru 2 (a.k.a Slip Stream 140; 140mm, nine blades, sleeve bearing; abbr KM2-800) nominal 800 rpm
• Scythe Kaze Maru 2 (a.k.a Slip Stream 140; 140mm, nine blades, sleeve bearing; abbr KM2-1200) nominal 1200 rpm - two. Used with a fan controller to emulate a second KM2-800.
• Scythe Slip Stream (nine blades, sleeve bearing; abbr SS-500) nominal 500 rpm - two
• Scythe Slip Stream (nine blades, sleeve bearing; abbr SS-800) nominal 800 rpm - two
• Scythe Slip Stream (nine blades, sleeve bearing; abbr SS-1200) nominal 1200 rpm - two
• Scythe Slip Stream (nine blades, sleeve bearing; abbr SS-1600) nominal 1600 rpm - two
• Scythe Slip Stream (nine blades, sleeve bearing; abbr SS-1900) nominal 1900 rpm - two, @5v
• Scythe S-Flex E (seven blades, fluid dynamic bearing) nominal 1200 rpm
• Silverstone FN121 (nine blades, sleeve bearing) nominal 1200 rpm
• Thermalright TY-140 (140mm, seven blades, liquid state bearing) nominal 900 - 1300 rpm - two
• Yate Loon D12SL-12 (seven blades, sleeve bearing) nominal 1350 rpm
• Yate Loon D12SL-124B (seven blades, sleeve bearing) nominal 1350 rpm.
• The two Yate Loon fans do not have the same blade shape: the black D12SL-12 has blades with the same curve on their leading and trailing edges; the clear D12SL-124B LED fan has trailing edges that are nearly straight. But their similarity in rpm’s allowed me to try them as a pair. Feel free to disregard the results if you like.
• Zalman ZM-F3 (seven blades, sleeve bearing) with resistor wires, nominal 900 rpm - two

A note on the ULNA and LNA: one of each comes with the D14. When used as directed, they cause both fans to run around 900 rpm. But of course there are other possibilities, and we will explore them. For example, you could use a ULNA or an LNA on a 140mm fan other than the P14, as long as the other fan has a low enough current draw. If OTOH you wanted to use two NF-P12’s, you would buy a second one. And each P12 sold at retail comes with its own set – one ULNA and one LNA. This gives us the opportunity of using two P12’s, each with a ULNA or LNA. So we’ll look at those possibilities as well.

Reading the Charts. Each line of each chart contains text and bars. The text elements are: Push Fan name + Center Fan name + Pull Fan name - push fan rpm | center fan rpm | pull fan rpm. The bars: blue is for Sound Pressure Level (SPL) in dBA and peach is for Temperature Over Ambient (TOA) in degrees Celsius.

TOA is obtained by subtracting the Ambient temp from the temp of the first core, the one that always runs the hottest on this chip.
Edited by ehume - 8/15/11 at 8:45am

4.3 Center Fan: Noctua NF-P14 with LNA

If you are trying to save a little money after you have bought your D14 you might try exploring the resistor wires that Noctua included in your kit. With the D14 Noctua calls both adapters Ultra Low Noise Adapters (ULNA), but when these appear with individual fans, the one with the black socket is called a Low Noise Adapter (LNA) and the one with the blue socket is called an Ultra Low Noise Adapter (ULNA). In the discussion and charts that follow I will refer to these adapters as LNA and ULNA. First up will be the P14 with a LNA.

The data will be presented first alphabetically by the push fan’s name, then by SPL, then by TOA. It basically shows you what you can do with your own D14 kit and some 120mm fans:







Comments:

As you will see when you look at the results for other fans and fan settings, this set of results is dominated by the noise of the P14 in the middle. When you look at results with the Kaze Maru 2 and the P14 with a ULNA you see a steady increase of noise from very low levels. But with the P14 on a LNA the system never gets quieter than the equivalent of 18 dBA, staying there in a sort of flood plain (the converse of a plateau) across many fan combinations, and hardly moving upward. Now 18 dB is very quiet, on the edge of ultra-quiet, but that is the quietest a set of fans will get if you have a P14 with a LNA in the center position.

For my money, the best combination of cooling and quiet in this group is to flank the P14+LNA with a pair of Slip Stream nominal 800 rpm fans. At 19 dBA, they are very quiet, and 51.9c TOA is excellent. Look at the second chart in this group. You see a march of cooling performances, and then you see a drop in TOA. That’s the P14+LNA with a pair of SS-800’s. You can get a better TOA of 51.1c by putting a pair of SS-1900’s on 5 Volts, and have a SPL of 20 dBA - still very quiet. Although I didn’t test temps, the SS-1600’s ran about 30 rpm slower than the SS-1900’s at 5v. It also made 20 dBA of noise. Still, the P14+LNA with a pair of SS-800’s looks to be the best of this bunch.

If you look at the third chart, where fan combinations are arranged in order of cooling, you can also see how the SPL’s relate to the TOA’s. If you’re wondering at the limited range of cooling here, it is because There Ain’t No Free Lunch ™, no magic fans. What we have instead, I believe, is an optimal speed for each shape of blade, each incidence (angle of blade). But if you’re asking for very quiet operation, your range of cooling will be limited.

I did a run with stock fans, for reference (TOA = 50.6c, SPL = 27dBA). That is a superb TOA, and 27 dB is quiet. What I find remarkable is that fan setups could so closely approach the stock cooling and make enough less noise that you can readily hear the difference (see the third chart). This is at least in part due to an interesting phenomenon I discovered: the stock fan makes noise you can hear through the front and back faces of the heatsink. More interesting: you can muffle that noise by putting a working fan at the front and the back of the cooler. This is why three fans are not significantly louder than two fans.




Edited by ehume - 8/15/11 at 8:57am

4.4 Center Fan: Noctua NF-P14 with ULNA


The data is first presented alphabetically by the push fan’s name, then by SPL, then by TOA. It basically shows you what you can do with your own D14 kit and some 120mm fans:






Comments:

With the P14 running slower, the noise is more balanced between center and end fans. You can see this in particular with the Nidec Gentle Typhoon AP-14. With the P14 on the LNA, we had a TOA of 51.9c and a SPL of 21 dBA with one AP-14, a TOA of 50.9c and a SPL of 23 dBA with two. With the P14 on the ULNA, we had a TOA of 52.4c and a SPL of 20 dBA with one AP-14, a TOA of 51.1c and a SPL of 21.5 dBA with two. So with two AP-14’s, going from the LNA to the ULNA left the TOA almost unchanged while dropping the noise 1.5 dBA. That’s not much, it’s true. But it does point to a more balanced operation.

Overall, the noise is a lot lower – down to 14 dBA. But the temps are higher. The 500 rpm Gentle Typhoon AP-11’s failed here. They let the cpu temp go over 85c, at which point OCCT called a halt to the run.

And the pair of SS-800’s are again the stars of the set. Making only 16.5 dBA of noise, they kept the cpu core at 51.9c above ambient. More: adding the second SS-800 caused a temperature drop of 2.3c and cost only one dB in noise. But this only makes sense. As we drop the speed of the center fan to zero, the end fans will become more important, so two vs. one end fan will make more of a difference.

Here I tested all-Noctua setups without and with Low Noise and Ultra Low Noise Adapters. If you have a second P12, you’re going to have a second set of LNA and ULNA. Going on that assumption, I used two of each in various setups. Of course, I also cheated at one point and used three ULNA’s. I hope you’ll forgive me.
Edited by ehume - 8/15/11 at 9:03am

4.5 Center Fan: Scythe Kaze Maru 2 (a.k.a. Slip Stream 140) SM1425SL12L, nominal 800rpm

It has always been a great puzzle to me why Scythe never made a Kaze Maru at this speed. The Kaze Maru was an 11-bladed 140mm fan, and is no longer made. It was available with 500 rpm, 1200 rpm and 1900 rpm nominal speeds. The Kaze Maru 2 series comes in 500, 800, 1200 and 1700 rpm nominal speeds. The 800 rpm model seems to inhabit a sweet spot, where it produces a useful airflow while making almost no noise. In the center position of a D14 it recorded its lowest noise with a pair of 500-rpm Slip Streams, giving an equivalent SPL of 12.5 dBA while cooling to under 57c TOA. To put this in perspective, this is a cooling as well as a Hyper 212+ when it is screaming its heart out while you literally can’t hear the D14 from a meter away.

At the other end (best cooling), among the fans I tested with the KM2-800 the best results came with a pair of AP-14’s that made the equivalent of 20.5 dBA SPL while cooling the hottest cpu core to 50.7c TOA. By comparison, with this apparatus the stock fans on the D14 cool the hottest core to 50.6c TOA while making 26-27 dBA noise. So, two AP-14’s and a KM2-800 take you from quiet to very quiet without losing any cooling. At these sound levels, 6 dB is quite a drop.

Time to look at the data. As usual, I’ve arranged it first by the name of the push fan; then in order of SPL, quietest to loudest; and finally by TOA, from coolest to warmest. Here’s the data:







Comments:

Based on work with the somewhat more powerful P14 with ULNA and LNA, I know that with the Kaze Maru 2 800 rpm in the center, Gentle Typhoon AP-11 will fail, so I didn’t test it.

When you add a second fan you tend to get another degree of cooling. An exception is when I went from a single 11-bladed Coolink SWiF2 at about 800 rpm to two of them. Got a drop of 2.8c. Even allowing for some error, there is at least a drop of 2c there.

Looking at other 800 rpm fans – the nine-bladed Slip Stream and Gentle Typhoons come to mind – one gets the impression that fans at about that speed seem to cool very well with this heatsink while being very quiet about it.

A single Zalman ZM-F3 on its resistor wire also cools very well. Yet adding a second of these adds 4 dB to the noise and you gain not quite a degree in TOA. I wonder if it has to do with the fan having seven blades.
Edited by ehume - 8/15/11 at 9:09am

4.6 Miscellaneous fan setups

There is another 140mm fan that has great potential – the Thermalright TY-140. At its nominal speed of 1300 rpm it is about the best 140mm fan sold. Since the D14 comes with a LNA and an ULNA these could be used with a TY-140.

Alas! What a bleak prospect! Another long string of testing! Look at each line of each chart. Each one represents 45 minutes of run time, and another 15 minutes of processing. That’s an hour per line. And do them again with the TY-140? Too much.

So I took a different tack: I set up pairs of fans and tested them with empty space in the center; with a shroud; with the TY-140+ULNA; with the TY-140+LNA; with the TY-140 at full speed. This allowed me not only to compare the performance of the TY-140 with a few representative fan setups, but also allowed me to look at the effects of progressively increasing assistance in the middle.

Then I looked at a few random setups. This is, after all, a very flexible system. It would take a long, long time to test all the possible combinations. I’ll settle for a few very quiet and ultra-quiet cooling solutions.

BTW - I had been using a modded (cut up) TY-140. Since the TY-140 has an irregular shape, 140x160mm, this allowed me to more easily use the fan with the 140mm width aligned with the width of the heatsink. OK, the fan was 160mm tall, but it didn’t stick much above the tips of the heatpipes.




I discovered to my chagrin, however, that the Noctua clips would not go around a TY-140 that had not been cut up. Interesting, OTOH, was that with the TY-140 oriented so its 160mm dimension was horizontal, the fan clips engaged readily with the heatsink fanclip slots. All you needed was the screwhole mod:




When I compared temps I twice found the wide position produce 0.3c warmer TOA’s than the tall position, which is close enough. So I switched. You will see the tall position marked as TY-140t, the wide position unmarked. In future testing I will use only the wide position.


OK. Now the data:


Comments:

As much as I love my Gentle Typhoons, it is clear to me that the Slip Streams perform slightly better on the D14. Here I compared the SS-800 and SS-1200 against the 800 rpm AP-12 and the 1450 rpm AP-14. The Slip streams cooled slightly better at a cost of only one to 1.5 dBA. Pretty good for cheap fans. OTOH, the GT’s held their own, and they are ball bearing fans so should last a lot longer, even if you keep the SS’s lubricated (see my sig).

Among the Gentle Typhoons the AP-14 performed the best but bumped up against the top of very quiet operation. A nice compromise is the 1150 rpm AP-13. For only a rise in TOA of 1.7c you get 3.5 dB less noise. That’s down at the top of ultra-quiet operation. The data is not on the chart, and I didn’t test the TOA’s, but with a TY-140 in the center, the SPL was 18 dBA with the ULNA, 19 dBA with the LNA and 22 dBA at 12v. Worth testing later, I think.

As usual, the San Ace “Silent” 9S1212L401 matched the AP-14 and the SS-1200 for performance and noise. Its specs put it on par with the AP-15, but it performs more like the AP-14, which is nice in terms of very quiet operation. When I move on to quiet operation I’ll be sure to test three 9S1212L401’s since I have a bunch.

In the SS-800 section you can see the effect of going to TY-140 from TY-140t. I think 0.3c is not too high a price to pay to get to use the TY-140.

It has come to my attention that some people will need to use this cooler in center-pull configuration rather than the stock push-center fans so they can use RAM with tall heatsinks. I also learned that one can operate the D14 with a 140mm fan in the pull position. So I tried a couple pairs of very quiet 140mm fans. As usual the KM-1900 is fairly loud for the output it produces, but the pair of KM2-800’s give decent cooling in ultra-quiet operation. This is definitely an option when you want your cooler to be essentially silent. But really, modern RAM can and does operate at 1.5v and below. Even at 1.65v you don’t need high profile heatsinks on your RAM. It’s just a marketing gimmick. Get low profile RAM.
Edited by ehume - 8/15/11 at 9:10am

4.7 Final and Additional Thoughts

I was disappointed that the Nexus D12SL-12 (SPCR’s favorite fan) didn’t make a better showing. I was surprised to see the Slip Streams do as well as they did. I wasn’t surprised at the performance of the Gentle Typhoons, in particular the AP-14. I just wish it was available as a PWM fan.

What it all shows is that despite their not giving us PWM fans, the ULNA and LNA Noctua included with the D14 make this cooler supremely flexible. Yet this heatsink is better and more flexible than the fans that come with it.

I’ll add more thoughts as I – or some readers – think of them.
Edited by ehume - 8/15/11 at 8:52am