You are here

Scythe NCU-2000 Fanless CPU Cooler

Sept 19, 2004 by Mike Chin

Product
NCU-2000 CPU Cooler
Supplier
Scythe USA
Price
~US$60

The Scythe NCU-2000 CPU Cooler is an update of the NCU-1000 reviewed here almost a year ago. That fanless heatsink was (and still is) one of the most unusual we've seen; the NCU-2000 carries on the tradition. As with the review of the original, it's worthwhile starting with a picture.



NCU-2000 on left next to recent Scythe FCS-50 in use on a motherboard.

Yes, it's very large, and especially tall. Its dimensions are very similar to the original NCU-1000 pictured below.



The original NCU-1000.

The NCU-2000 is actually very slightly smaller than the older version: W108 x D83 x H140 mm compared to W108 x D86 x H143 mm. That's about 4 x 3.5 x 6 inches. It weighs considerably less: 505 grams compared to 615 grams for the NCU-1000. And it is designed for both socket 478 CPUs (P4) as well as A64 (754, 949, 940) without a fan. The NCU-1000 is for socket 478 only. Finally, despite its large size, the bottom portion of the HS is well within the permitted HS zone for both socket 478 and A64 motherboards, which should allow it to clear any nearby tall components on the PCB.

As with the NCU-1000 the new model is a heatpipe device. In the original, the heatpipe consisted of a circulating capillary tube which is visible at the top.



NCU-1000 circulating capillary tube.

In the NCU-2000, a flat heatpipe is employed. This was first seen in the e-Otonashi fanless EPIA-M cooling case and more recently in the FCS-50 heatsink. As in the FCS-50, a length of flat heatpipe is shaped into a rectangular loop with the ends overlapped and joined together. The fins are soldered to this heatpipe frame, and the copper base is clamped to the bottom. This drawing used to show the flat heatpipe design in the FCS-50 is worth reproducing here. Compare the FCS-50 drawing to the photo of the NCU-2000. The basic design concept is the same. One of the big differences is the spacing between the fins, with is fairly small with the FCS-50, but much greater with both the NCU-2000 and the older NCU-1000. This open spacing is necessary for convection-only cooling; there's no fan to force the air through tight spaces.

In both coolers, the flat heatpipe loop is the frame to which the fins and the base are attached.

Aside from the use of the flat heatpipe, the NCU-2000 differs from its predecessor in that for both the P4 and the A64, the stock mounting bracket is used. Given the relatively modest 505 gram weight of the HS, the standard HS retention bracket on P4 or A64 boards should have no problems. The earlier NCU-1000, in contrast, replaces the P4 retention bracket and backplate with a bolt-through-board solution. The change makes installation easier.

WHAT YOU GET

Here's the retail box and all of the contents, excluding the red and yellow tape measure.

The instruction sheet is in Japanese and English. It is detailed enough and well-illustrated to ensure correct installation. There is a separate sheet of safety precautions, which is quite extensive and detailed. As with the NCU-1000, there are explicit cautions about the toxicity of the HFC-134A coolant fluid in the heatpipe. The box, too, has a lot of text, including warnings and cautions about use and applicability, including the following:

"This product is designed for the usage such as e-mailing, Internet browsing, word processing and spreadsheet or an equivalent at an ambient temperature up to 25C. Please note that this product is NOT for conditions such as data encoding/decoding, using benchmark software, and/or other extraordinary conditions. The manufacturer accepts no liability for the usage of this product not complying with the warning listed above."

That's a pretty serious disclaimer, similar to the one that accompanies the NCU-1000. You get the sense that they're heeding the advise of the company's lawyer. The above disclaimer is combined with many other cautions on the box, including:

  • Ensure good ventilation for the PC case
  • Monitor temps on high speed CPUs
  • Don't use the HS on a horizontal motherboard
  • Check fit as it may not fit some motherboards
  • Use it only with a new motherboard

Whew! All this could make you nervous, but if you're even looking at this product, chances are pretty good that you are technically adventurous (at least with your PC) and you've got some understanding of thermal issues in PCs.

ASSEMBLY

Some assembly is required. The assembly is not difficult. Four bolts sandwich the bottom portion of the HS between the copper base and the mini-heatsink at the bottom. It ends up looking like this:

Configured for socket 478; on each side, a metal strip into which the screw threads acts as a spring tensioner.
Photo above shows HS set up for A64. There is a spring loaded bolt on each side.

The copper base plate can be turned 90 degrees from the position shown above. The instructions state that the wider side of the fins structure should face up/down. Which way the copper plate should be mounted depends on the CPU socket configuration of the motherboard. The point is that you want the holes in the heatsink to be facing up/down rather than side/side so that air will flow through them in natural convection.

The copper base is smoothly polished. Some fine machining lines are visible and can be felt with a fingernail. Lapping might make a difference but I think it would very small.



Base can be mounted rotated 90 degrees from the above.

Depending on which way the base is mounted, the mounting bolts end up under the side-extended fins. Access to the bolt then becomes impossible with a screwdriver, which is why Scythe includes a small wrench for the hex-headed bolts. I found out firsthand how tedious it is to install this HS with the base mounted the hard way.



Wrench, P4 clip with hex/Phillips head bolt, and P4 clip on one side HS on motherboard.

INSTALLATION ON TEST PLATFORM

SPCR's standard heatsink testing method of a Panaflo 80mm low speed fan is not applicable here. The design calls for no fan to be used. There's no facility to mount any fan, although it is easy to rig up.

Test platform

  • Intel P4-2.8A The Thermal Design Power (TDP) of this P4-2.8 (533 MHz bus) is 68.4 or 69.7W depending on the version. As the CPU is a sample without normal markings, it's not clear which version it is, so we'll round the number off to ~69W. The Maximum Power (MP), as calculated by CPUHeat & CPUMSR, is 79W.
  • Intel D845PEBT2 motherboard - Intel 845PE Chipset; on-die CPU thermal diode monitoring
  • nVidia GF400MX VGA card (AGP)
  • OCZ

    DDRAM PC-3700, 512 MB

  • Seagate Barracuda IV 40G 1-platter drive (in Smart Drive from Silicon Acoustics)
  • CoolMax Taurus CF-300 Fanless ATX PSU
  • Zalman Multi-Connector (ZM-MC1) and Fanmate1 voltage controller
  • Arctic Silver Ceramique Thermal Compound
  • Two-level plywood platform with foam damping feet. Motherboard on top; most other components below. Eases heatsink changes and setup.
  • CPUBurn processor stress software
  • Folding @ Home client software
  • Intel Active Monitor and Motherboard Monitor software to show CPU temperature

For the correct orientation of the fins on the Intel D845PEBT2 motherboard, the copper base had to be mounted the easy way, as shown below.



View from top edge of test motherboard. Arch of clip is the source of tension to keep HS tight against CPU.



When the clip is installed, the bolt simply fits into and pushes against the round niche at the top edge of the base.



Here it is, fully installed.

Before testing could be started, however, the motherboard had to be vertically positioned. This HS is designed to be used in tower case systems, remember? So, the entire test rig was carefully turned on its side as shown below. It is not quite perfectly vertical, but is close enough for our purposes.



Only precarious if jostled hard.

 

TEST RESULTS

No fans were anywhere near the CPU heatsink or motherboard. The ambient temperature was 23°C. The NCU-2000 made no noise whatsoever.

NCU-2000 Fanless Performance
Idle
General Apps
Folding
Max Temp
Max Temp Rise
°C/W

TDP
°C/W

MP
55
57~62
64
72
49
0.71
0.62
* All figures except °C/W in in °C.

TDP (Thermal Design Power of CPU) = 69W

MP (Maximum Power
of CPU) = 79W

Ambient temperature = 23°C

All the temperatures are pretty bad by fan-cooled HS standards, where we often record performance that's twice as good. However, it is important to keep in mind that the HS did not get any help from any fans anywhere near the system. In a real PC, you will almost invariably have a PSU with a fan or a back panel exhaust fan. Even spinning very slowly with low noise, a single fan will cause enough airflow through the open fins of the NCU-2000 to lower the CPU temperature.

The system did not crash at any time. Running [email protected] is a pretty good approximation of the typical highest load most desktop users are likely to reach. We have never been able to achieve the load of CPUBurn with any combination of actual applications. Running stress programs like CPUBurn is explicitly cautioned against in the NCU-2000 instructions, you will recall.

The 64°C seen with [email protected] is around the high borderline of safe P4 temps. This was with an ambient of 23°C, so add another 10-12°C to estimate in-case temperature, and we have 74-76°C, at which point thermal throttling will probably occur with most P4s. However, airflow created by case fans will certainly impact this temperature as well.

NCU-2000 VS. NCU-1000

How does the performance of the NCU-2000 compare with the NCU-1000?

The NCU-1000 was tested last year on a different platform, the most important differences being:

  • Intel P4-2.53 processor. TDP is 61.5W; Max Power is 72W. 71° C rated maximum junction temp.
  • Gigabyte GA-81RXP motherboard - Intel 845 chipset; on-die thermal diode monitoring.

Keep in mind that the temperate readout was not calibrated and temperature readings are not directly comparable to the results with the current test platform. In the NCU-1000 review, I wrote, "thermal overload... using CPUBurn occurred in little over 10 minutes; the program crashed. As cautioned by the manufacturer, the NCU-1000 is not designed for this kind of stress." In contrast, neither the CPUBurn program nor the system ever crashed during testing, and this is with a CPU that runs 7-8W hotter.



NCU-1000 on current test rig.

For confirmation, I pulled the NCU-1000 sample out of mothballs and installed it on the current P4-2.8 test platform. I let idle overnight then had a look at the temperatures the next morning. The results were surprisingly poor; enough so that I decided to reinstall the HS altogether and run further tests on the following day. The second set of tests were within 1°C of the first set. The NCU-1000 results are compared with those for the NCU-2000 in the table below.

NCU-2000 Vs. NCU-1000, fanless
HS
Idle
Folding
Max Temp
Max Temp Rise
°C/W

TDP
°C/W

MP
NCU-2000
55
64
72
49
0.71
0.62
NCU-1000
63
72
-
-
-
-
* All figures except °C/W in in °C.

TDP (Thermal Design Power of CPU) = 69W

MP (Maximum Power
of CPU) = 79W

Ambient temperature = 23°C

There was no point running CPUBurn on the NCU-1000; it would have crashed and/or the CPU retreat into thermal throttling. I wrote in the NCU-1000 review, "I am not in complete agreement about the "up to P4-2.8" (>68W) recommendation for this cooler" when used as intended, without a direct cooling fan. My original opinion is confirmed. It's also clear that the NCU-2000 is a significantly improved performer.

Interestingly, in all the verbiage supplied with the package, there is no mention of the highest speed CPU recommended for use with the NCU-2000. This is probably wise from a legal point of view, and perhaps also from a marketing point of view. The buyer who needs explicit, plain-as-day instructions may not be an ideal user of the NCU-2000.

PERFORMANCE WITH A FAN

There are no mounting points on the NCU-2000 for any fan. This is also true of the NCU-1000. However, mounting a fan directly to this HS is a piece of cake: A bit of wire or string through the fan mounting holes and hooked around the fins would work fine in a tower case. I decided this just had to be tried.



A Nexus 120mm fan covers most of the fins. For the test, it was just placed atop the HS.



SPCR reference Panaflo 80L was also tried.

NCU-2000 Performance w/ Fans (blowing up only)
HS
Idle
Folding
Max Temp
Max Temp Rise
°C/W

TDP
°C/W

MP
No Fan
55
64
72
49
0.71
0.62
Panaflo 80L-7V
43
49
56
33
0.48
0.42
Nexus 120-7V
40
44
49
27
0.39
0.34
Nexus 120-10.5V
39
42
47
23
0.33
0.29
NCU-1000 w/Nexus 120-7V
40
44
50
26
0.38
0.33
* All figures except °C/W in in °C.

TDP (Thermal Design Power of CPU) = 69W

MP (Maximum Power
of CPU) = 79W

Ambient temperature = 23°C

It's clear that even a small amount of direct airflow from a fan dramatically improves the cooling performance of the NCU-2000. And the NCU-1000, as well. In fact, a Nexus 120mm fan at 7V (which provides about 20 CFM at a virtually inaudible noise level), virtually erases the performance difference between the older and newer heatsinks. With the Nexus 120mm fan, the cooling performance jumps into higher ranks of Recommended Heatsinks.

MOUNTED IN THE WRONG DIRECTION, WITH FAN

Given the powerful role a fan has on the cooling performance of the NCU-1000, I decided it was time to find out what would happen if the HS was mounted incorrectly (turned 90 degrees so that air flows best through the HS in a horizontal plane) but with a fan blowing towards the rear of the case (if the motherboard was installed in a case).

As I mentioned earlier, screwing the inside bolt under the fins is truly tedious. The photo below explains just why: There's simply no room to move.



Installing the HS with the fins hanging over the bolt means the supplied wrench has to be used.

.

NCU-2000 installed facing the wrong way, with Nexus 120mm fan sucking air through it towards the rear of the case.

The fan was propped on two small boxes placed atop the VGA card; don't try this at home kids!

After this test was run, I kept the HS and fan mounted the same way but put the wood platform back on its feet, thus placing the motherboard on the horizontal plane, as it would be when mounted in a HTPC horizontal case. The results are shown below.

NCU-2000 installed facing wrong way w/ Nexus 120mm fan at 7V
Motherboard orientation
Idle
Folding
Max Temp
Max Temp Rise
°C/W

TDP
°C/W

MP
Vertical
41
45
50
28
0.40
0.35
Horizontal
41
45
49
27
0.39
0.34
HS mounted correctly, motherboard vertical*
40
44
49
27
0.39
0.34
* Data from previous table, with Nexus 120mm fan at 7V

All figures except °C/W in in °C.


TDP (Thermal Design Power of CPU) = 69W

MP (Maximum Power
of CPU) = 79W

Ambient temperature = 23°C

As you can see, the performance is virtually unchanged compared to when the HS is mounted facing up/down as recommended -- when there is a fan pushing some air through the fins. In both cases, unrestricted movement of the airflow through the fins and away from the HS is required to achieve this performance.

CONCLUSIONS

The manufacturer's application notes make it abundantly clear that the NCU-2000, when run without any direct cooling fan as intended, should not be used to cool a CPU stress tested with a constant 100% load. Despite this caution, the sample cooled a P4-2.8 adequately to keep the test system from crashing with half an hour of CPUBurn. This was in open air with an ambient temperature of just 23°C. Inside a case with temps at least 10°C higher, the NCU-2000 might not have been able to keep the CPU adequately cool. On the other hand, if the case has one or more fans blowing air near or around the CPU area, it might perform just as well if not better. In the fanless tests I conducted, the NCU-2000 had no help from any sidestream airflow at all; there was no fan producing any kind of airflow that could have helped cooling in any way.

Folding @ Home is a good approximation of the highest loads desktop PC users can expect to reach. With the convection-only test conditions, the NCU-2000 does just barely well enough to keep on the safe side of cool. Again, how well it does inside a real case will depend much on both ambient temperature as well as airflow in the case.

When used exactly as directed -- web surfing, e-mail, word processing, general office apps and playing music or DVDs -- the NCU-2000 will probably do a fine job of keeping a P4-2.8 cool enough in a good but moderate airflow case typical of silent PCs. (Unrestricted case airflow is an important objective for all PC silencers, but those who use the Zen NCU-2000 need to pay particular attention to good case airflow.) Compared to the NCU-1000, the new model is lighter, cools better, and fits both socket 478 and A64 platforms: All of these are positive improvements.

As with the NCU-1000, a 120mm fan power supply makes good sense. Whether the CPU heat passing through the PSU will affect its fan speed will depend on the particular PSU, the heat of your CPU, case cooling and ambient temperature.

With direct forced airflow, there's little doubt that the NCU-2000 can handle >3.0G CPUs. It would no longer be fanless, but that is a somewhat dubious distinction. Given its size and the proximity of the CPU to the PSU fan and/or case exhaust fan, there will almost always be some forced airflow across its fins in any typical system.

The Scythe NCU-2000 Cooler is a significant improvement over the original fanless NCU-1000. It's certainly worth trying for the more sophisticated silent computing enthusiast.

Our thanks to Scythe for the NCU-2000 sample and their kind support.

* * *

Comment on this article in our Forums.

Sections: 

Google

www SPCR