MyComputingArt

Introduction

As I wrote about the Scythe Ninja Plus rev. B I wanted to replace the AMD stock heatsink with a more efficient one. According to some reviews AMD stock heatsink is quite good (it cools fine and it's not noisy) but aftermarket heatsinks cool better and more quietly: this is especially true for heatsink with a 120 mm fan, that move much more air and more silently than the 70 mm fan coming with the AMD stock heatsink. Indeed, in a tower case the highest area of the motherboard isn't reached by the typical airflow (bottom front to back top) and hot air may rest in that area, unless the case has a top exhaust like the Antec P180, P182, P190); on tower shaped heatsinks that are positioned vertically it's often possible to mount a fan that pulls the air from that area and push it toward the back of the case through the heatsink fins, resulting in a better cooling of the motherboard.
My choice fell on the Noctua NH-U12F because:

according to some reviews its cooling performance are very good: Bjorn3d, socketzone, pcekspert, silentpcreview)
it's silent, thanks to a 120 mm fan, to be exact the SSO-bearing Noctua NF-S12-1200
there are no clearance issues with my motherboard, a MSI K9N Platinum, as reported on Noctua site and confirmed by a prompt answer of Noctua support to my mail: "There will be no problems with the capacitors, because msi never use some which are 30mm high. 30mm is the gap between the cooler base and the lowest cooling fin, with the cpu socket and cpu, there will be more than 35mm gap beween the lowest fin and the mainboard."
I wanted to change the Scythe Ninja that didn't fit in my mobo with a Scythe Infinity but the Infinity was not available
6-years warranty

Pictures

156.10 KB	197.35 KB	175.82 KB	123.02 KB	124.54 KB	108.47 KB	122.35 KB	111.34 KB
120.81 KB	131.29 KB	186.42 KB	96.41 KB	110.06 KB	116.93 KB	136.94 KB	146.11 KB

The box contents can be seen in the pictures:

the Noctua NH-U12F heatsink
the Noctua NF-S12-1200 fan
an English only manual
three clips with accessories and backplate to mount the heatsink on AMD AM2, AMD K8 and LGA 775 sockets
the "common parts":
- four adhesive antivibration strips and four steel clips to mount two fans on the NH-U12F
- the U.L.N.A. (Ultra Low Noise Adapter), simply a downvoltage fan cable
- a thermal paste tube
- four fastening brackets (two for standard mount and two for 90� rotated mount) with 8 screws
- two pressure springs
- a screwdriver

Noctua NF-S12-1200 performance, as measured by silentpcreview:

Voltage	Noise	RPM	CFM	Power
12V	25 dBA@1m	1250 RPM	63 CFM	1.32W
9V	21~22 dBA@1m	1020 RPM	52 CFM	1.00W
7V	~20 dBA@1m	840 RPM	42 CFM	0.79W
5V	<19 dBA@1m	640 RPM	30 CFM	0.60W
@25 CFM (4.3V)	<19 dBA@1m	570 RPM	25 CFM	0.53W

Making a flat surface

As can be seen in the first picture, there were some arcs along all the copper surface that would be in direct contact with the cpu. I first tried to remove them with cotton and a rubbing compound, as I did with the Scythe Ninja, but it didn't work, so I decided to lap it (it was my first time):

I attached a sheet of 800 grit sandpaper on a table
I wet the heatsink base with a few drops of extra-virgin olive oil
I placed the heatsink on the sandpaper and moved it over the sandpaper making shapes of 8, for about 20 minutes
Then I repeated the process with 1200 grit sandpaper
At the end I polished the base with cotton and a rubber compound

The result was an apparently flat but quite opaque surface, the Ninja's one was much more reflective.
Other people repeat the lapping procedure 4-6 times starting from 400 to 1200 or 2000 grit sandpaper but I had 800 and 1200 grit sandpaper only (see overclockersclub lapping guide, or an Italian lapping guide by microcip on hwupgrade forum).
Later I discovered that the arcs on the surface were there on purpose: according to Noctua NH-U12F FAQs the "micro-grooves" serve to disperse thick layers of high-viscosity thermal compounds to a uniform thin layer, because "too thick layers of thermal paste and air pockets drastically deteriorate heat transmission". It seems to me that those grooves could only help if an excessive quantity of thermal paste is used, but I'm not convinced they're useful if the thermal paste layer is thin.

Installing

107.76 KB	121.10 KB	121.53 KB	106.99 KB	115.79 KB	173.92 KB	173.86 KB	163.00 KB
173.31 KB	119.46 KB	115.00 KB	116.85 KB	128.75 KB	155.70 KB	154.59 KB	129.86 KB
132.15 KB	136.23 KB	152.13 KB	165.18 KB	156.34 KB	137.88 KB	144.52 KB	177.62 KB

The English-only manual is very clear and very well made, just as every manual should be: there's a section for each socket that explains step by step what to do, with clear pictures and a brief explanation for each one; I didn't have to think about what I was doing, I only had to mechanically follow the manual instructions.
The first 16 photos show step by step the backplate, mounting bracket and heatsink installation, the last 8 photos show the heatsink after installation.

Clearance

I had no clearance problems fitting the Noctua NH-U12F on the MSI K9N Platinum motherboard, and I bet I wouldn't have, after all the times I measured the area around the socket and checked the heatsink and fan measures before ordering it:

the heatsink doesn't invade space outside the motherboard, even if mounted vertically
when mounted vertically, the heatsink (without the fan) is slim enough not to invade the sky above the capacitors or RAM
the gap between the base and the lowest fin is 30 mm and becomes 35 mm above the motherboard floor because of the socket and cpu thickness; this way it's possible to mount one fan on each side of the heatsink
the only clearance problem I found is about inserting or removing the nearest RAM module while the fan is mounted on the heatsink, but the fan can be removed and mounted again quite easily and that would happen only when 4 RAM modules are used
as pointed out by Tony Whitmore in his blog and can be seen in this picture, on the back of the MSI K9N Platinum a resistor is not completely in the center square of the backplate, but neither with the MSI backplate it was totally uncovered; so I prudentially made two cuts in the plastic sheet to reduce contact with the resistor

Results

Configuration:

cpu: AMD X2 3800+ Windsor
mobo: MSI K9N Platinum, passively cooled (no fan)
ram: 2x1 GB Corsair DDR2 PC6400 800MHz
gpu: Sapphire Ati Radeon X1300 512MB ROHS (passively cooled)
psu: Seasonic S12-500
hard disks: Western Digital WD2500JS and WD1200JB, cooled by the front case fan, a Papst 4412 F/2GLL fan, and suspended
case: Antec SLK3000B with standard TriCool fan on the back set at low speed and a Papst 4412 F/2GLL fan behind the unused cards slots

Temperatures at idle have been measured after 1 hour of low core activity since I switched on the pc (cpu between 0 and 5% while editing files, core fan at 40%), temperatures at full have been measured after 10 minutes of avi encoding by Virtualdub (cpu at 65%).

Cooling

Temperatures at idle with stock AMD heatsink and 70 mm stock fan revving at 1170 rpm (40%), just before installing the Noctua NH-U12F (ambient temperature 24�):

cpu: 36�
motherboard: 34�, 39�, 39�
hard disks: 32�, 32�

The core fan speed is changed automatically from a minimum of 40% (1160-1170 rpm) to 100% (3300 rpm) by SpeedFan.

Temperatures with Noctua NH-U12F and Noctua NF-S12-1200 with U.L.N.A. revving at 975-980 rpm (100%), cpu at 65%, Tricool at medium speed and both Papst at maximum speed, ambient temperature 24�:

cpu: 42-44�
motherboard: 40�, 43-45�, 43-45�
hard disks: 32�, 31�

Temperatures with Noctua NH-U12F and Noctua NF-S12-1200 without U.L.N.A. revving at 1395 rpm (100%), cpu at 65%, Tricool at medium speed and both Papst at maximum speed, ambient temperature 24�:

cpu: 39-42�
motherboard: 39-41�, 43-45�, 42-43�
hard disks: 32�, 31�

It's important to notice that setting the Tricool at high speed the core temperature never rose over 40�, while at medium speed it reaches 42�: this indicates that maybe the airflow is not optimal because the hot air is not exhausted very fast and needs to be forced out increasing the exhaust fan's speed.
The thermal grease requires a few days to reach optimal efficiency, and unfortunately the next days the ambient temperature rose about 4� to 28�, so it was not possible to make a meaningful comparison. By the way, with Noctua NH-U12F and Noctua NF-S12-1200 without U.L.N.A. revving at 1395 rpm (100%), cpu at 95%, Tricool at low speed, front Papst at maximum speed and back Papst stopped (worst scenario), the motherboard temperature never exceeded 48� and cpu temperature never exceeded 46�.

A few days later, when thermal grease is supposed to have reached efficiency and when ambient temperature dropped to 24�, these are the temperatures at idle, with Noctua NH-U12F heatsink and Noctua NF-S12-1200 without U.L.N.A. revving at 40%, Tricool at low speed and both Papst at low speed:

cpu: 33�
motherboard: 37�, 39�, 38�
hard disks: 32�, 36�

Speedfan shows 0 rpm as the speed of the heatsink fan, but I checked that the NF-S12-1200 is revving; the minimum speed detected by Speedfan is 870 rpm at 50%.
The difference of the hard disks' temperatures (4� more for the second hard disk, unchanged temperature for the first) is very likely due to the fact that the bottom disk is now suspended at a slight lower height than before and gets slightly less air below it: it's suspended using cord and knots only, which I have to make every time I mount it, so the height at which it's suspended is very like to change a bit every time, and a slight difference in the quantity of air that passes below it is proven to affect the cooling results significantly.
The cpu temperature dropped by 3�, showing that Noctua NH-U12F cools better than the standard AMD heatsink, even with the doubt of a non-perfect lapped surface. Instead the temperatures of the motherboard are the same as before but one, which increased by 3�, but this temperature seems to be affected the most by the front fan, so it doesn't depend on the NH-U12F.

Silence

With all the fans revving at minimum speed and U.L.N.A. installed, a low noise can be heard only getting the ear at about 20 cm from the case: it's the Papst. Without U.L.N.A. and the NF-S12-1200 revving at full speed (1395 rpm), it can be heard a few meters far only during the night, when there is complete silence, but it doesn't disturb, being a low volume whoosh; early in the morning, birds twitter and cars wheels noise on a 100 meters far road (with the window open) cover it. Without U.L.N.A. and the NF-S12-1200 revving at low speed (40%, <1000 rpm) it's barely audible even during the night.
So, I decided to remove the U.L.N.A. because I don't want to limit the cooling performance at 1000 rpm maximum and even without it the system is nearly silent.