I'm going to post some informal tests I've done on my own system, to compare the Arctic Cooling Freezer 64 Pro and the AVC 4-heatpipe top-down cooler. The latter came with my PIB (Processor In a Box)
Test Setup
- CPU: Dual Core Opteron 165 - 2700 MHz - 1.30V VCore
Mainboard: MSI K8N SLI Platinum (MS-7100 PCB rev 1.0) + Titan CUV2AB chipset cooler
Memory: Kingston HyperX 3200ULK2 (2x512 MB TCCD; 2-3-3-6 1T; 2.85V VDimm)
Video Card: Asus EN6600GT PCIex + Arctic Cooling NV Silencer 6 revision 2
Additional Cards: Sound Blaster X-Fi ExtremeMusic
Hard Drives: 2x Barracuda 7200.8 232 GB
Case: CM Stacker
PSU: Enermax NoiseTaker II 495AX
- Back fan: Cooler Master 120x25 (A12025-12CB-4KN-L1) PL12S12L, rated for about 42 CFM and 1200 RPM @ 12V.
Back fan: Jamicon 80x25 JF0825-1H, rated for about 38 CFM @ 3000 RPM - downvolted (5V)
Back fan: Coolink 80x25 SWiF 801 Basic, the specs aren't worth noting (completely untrue IMO) - downvolted (5V)
Front fan (hard disk cooling): Cooler Master 120x25, same as before but this one is downvolted at 5V.
The load applied to the machine was just its daily chores: two boinc work units, crunched in parallel; one from the CPDN and the other from QMC. Although Prime95 can generate a little more heat, the difference isn't that big actually. This system works 24/7 and the temperature readings were taken after several hours of continuous work to ensure the case has reached a stable working condition.
Both the coolers were cleaned with ArctiClean and then applied using a thin film of Arctic Cooling MX-1 Thermal Paste. The Freezer Pro was lapped because of some oxidation on the copper base (my fault); many people tried lapping the Freezer, but the opinions I found online generally agree with mine: while this can sound weird, it seems that lapping has no effect on cooling effectiveness of the Freezer.
On the other hand, the AVC was left untouched; this should not be a problem, because the MX-1 paste is designed to work well even with slightly rougher surfaces.
The MX-1 paste has a quite long curing time (AC states 200 hours): I could not wait that long, so I let one full day pass before making my test. I have to say I've never noticed any big improvements over time in my previous experiences with this paste, however I'll post an update if there's interest, and there is news.
Here are the photos of the copper bases, the difference in picture quality is due to different cameras being used, as a result the AVC base seems worse than it really is:
The Arctic Freezer was used with stock fan @ 100% (about 2200 rpm), while the AVC was tested with a Nidec TA275DC, which is capable of 6000 rpm max, but was throttled at about 3000 rpm. It is worth noting that the 70mm Nidec fan is slightly quieter than 92mm AC at the speeds they were tested.
For all measurements, I used the logging feature of Speedfan, so the temperatures are not accurate in absolute terms, but the difference has meaning. Ambient temperature is about 19 °C.
And now if you managed to delve through all this lenghty introduction it's time for the graphs:
Arctic Cooling Freezer 64 Pro (with lapped base)
AVC/AMD 4-heatpipe cooler
How to read the graphs:
All the lines are 10-period moving averages; the NBa reading comes from a "system" or "northbridge" sensor which is placed on the motherboard, I don't know exactly where. In both the graphs, NBa is constant and equal to 35 °C, which is good.
The salmon line (CPUa) is data coming from the motherboard CPU sensor. You will notice that both the coolers exhibit a quite stable CPU temperature, about 46 °C.
The green line (COREa) comes from the Opteron DTS; SpeedFan can only read the second core temperature on my system, which is usually slightly lower than the first core's, but within the range of variation for this sensor.
While both the coolers attain the same CPU temperature, the data from the DTS is different: the Freezer exhibits a higher core temperature (above the CPUa line), reaching 49 °C and averaging around 48 °C.
The AVC manages to keep the core under the CPUa temperature, with a peak of 46 °C and averaging around 45 °C.
One last note on the GPUa (light blue) and VGAa (blue) lines: while the card "ambient" sensor reads almost the same, the GPU core runs a bit hotter. This is easily explained: the AVC cooler is mounted with the fins parallel to the ground, however it has two slits on the top and the bottom, which are needed for the retention clip. A fast jet of warm air is exhausted through the bottom hole and hits the back of the graphics card, heating it up.
Conclusions
Under the conditions the coolers were tested, the AVC comes out as a winner, with a 3 °C better performance over the Freezer Pro.
It was quieter than the freezer, also having some headroom to cope with higher ambient temperatures: at about 5500 rpm (11V) it could push the core temperature down to 42 °C and reach 39 °C with some help from the 80mm case fans. The Nidec 70mm fan become noticeable over 3300 rpm, noisy at 4000, and unbearable at 5500
I can't really explain the difference between the Freezer and the AVC; while there was a difference in average VCore of about 0.010V between the two test runs, which could have adversely affect the Freezer performance, I don't think this can really be the reason.
The Freezer has received good reviews all around, and many tests (SPCR's own, too) showed it can achieve excellent cooling results, often on par with much more expensive coolers as the popular Scythe Ninja (+Nexus; see SPCR), the Zalman CNPS9500 (see Frostytech and Madshrimps), and even the Tuniq Tower 120 (+Papst: see Madshrimps)
It might be that my sample has somewhat less effective heatpipes than the samples these sites received.
I'm probably going to buy a new high-end CPU cooler soon, I'm comparing the Zalman 9700 (noisy), the Ninja Plus RevB and the Infinity, and the Noctua NH-U12F (which has uneven reviews). The Thermalright Ultra-120 and the Tuniq are hard to find in my country.
In that case, I'll repeat the test and will try to assess my Freezer performance against a new reference HSF.
That's all folks, I hope you find this post interesting