SuperCoolers Return: Zalman CNPS9700 LED and Scythe Infinity

Cooling solutions manufacturers cannot remain indifferent to the ever growing demand for high-efficiency solutions needed to deal with constantly increasing heat dissipation of central processors. This is when products known as “super-coolers” come into the game. Today we are going to introduce to you two solutions like that from Zalman and Scythe.

by Sergey Lepilov
12/11/2006 | 03:12 PM

It is in response to the constantly increasing heat dissipation of central processors that the manufacturers have developed high-efficiency air coolers, which have come to be known as “super-coolers”. The peak of the heat dissipation growth fell at the Pentium 4 and D processors, but the CPU manufacturers’ current orientation at creating power-economical processors doesn’t allow us to do without super-coolers because two “cool” cores heat up just as much as a single hot core does. Moreover, a processor can generate much more heat than usual if overclocked.


This makes a super-cooler still helpful and I can’t but check out two new models from Zalman and Scythe. The Korean company has released a new evolutionary model called CNPS9700 LED whereas the Japanese firm has announced a cooler called Infinity. In this review I will examine the design of these coolers, compare them with their market opponents and evaluate their noise level.

Zalman CNPS9700 LED

The Korea-headquartered Zalman is used to making new coolers larger than older ones as they did with the CNPS7000 series that was followed by the CNPS7700. Zalman plays the same trick again, replacing the CNPS9500 LED, released over a year ago, with the CNPS9700 LED model that boasts a much larger heat dissipation area.

The cooler’s design hasn’t changed otherwise and it still looks like a hedgehog curled up into a ball, only a larger one now. Anyway, this cooler is worthy of a closer inspection and study which I am going to give it right now.

Package and Accessories

The new cooler is packaged into a large box with a cut in the middle.

Inside the paper box there is a plastic one the cooler is firmly fixed in. Above it, there is a small flat box with accessories. Next to a list of features and specifications, you can find a mention of the new thermal grease Zalman ZM-STG1 on the package. Besides this grease, you’ll find everything necessary to install the cooler on every modern platform:

These are (from left to right and from top to bottom):

The new cooler’s accessories differ from those of the Zalman CNPS9500 LED: there are no clips to mount the cooler on Socket 478. Instead, there are fasteners for AMD’s K8 platform, including Socket AM2 processors. Socket 462 is already forgotten, and quite rightly so.


The new cooler is designed alike to its predecessor. It has three figure-of-eight heat pipes that carry thin (0.2mm) copper ribs:


It’s all the same as with the Zalman CNPS9500 LED, except for the dimensions of the heatsink and the size of the fan. The new cooler is taller and wider by 17mm and also longer by 5mm. The dissipation area has thus increased from 3698 sq. cm to 5490 sq. cm, or by an impressive 48.5%! Having become larger, the cooler has naturally become heavier. The Zalman CNPS9700 LED weighs 764g as opposed to 530g of the CNPS9500 LED model. The fastening mechanism of the new cooler is simple and reliable, so this rather heavy weight shouldn’t worry you at all.

Besides being larger (110mm in diameter), the fan of the new cooler has a wider speed range. Using the included Fan Mate 2 controller, you can vary the fan speed from 1250 to 2800rpm (as opposed to 1350-2600rpm with the CNPS9500 LED). The max level of noise is higher at 35dBA as a consequence. The fan still has that blue highlighting that used to please the eyes of every modder.

The six ends of the heat pipes are soldered up and gripped in between the plates of the cooler’s base with four screws.

The top plate, made from aluminum, has a jut for the retention clips. Take note of the groove in the center of that jut. It makes it easier to turn the clips if the clips are positioned along the pipes. The CNPS9500 didn’t have that groove. The three punched hollows on each side of the jut and the corresponding dents on the retention clips allow to fasten the cooler in the most optimal position and fix it reliably on the CPU’s heat-spreader.

The cooler’s base is perfectly finished and is absolutely flat.

Well, you could hardly expect anything else from Zalman.

Assembly and Installation

Installing this cooler is a simple and intuitive process whatever platform you have to deal with. If you mount your Zalman CNPS9700 LED on an LGA775 platform, you have to take the mainboard out of the system case to fasten the back-plate to its reverse side. Then you put the plastic mounting frame on the mainboard’s face side and attach it to the back-plate.

This photo also shows you Zalman’s new thermal interface, ZM-STG1 grease, on the CPU heat-spreader. Even using a special brush, I couldn’t apply it uniformly because it had a too thin, near liquid, consistence. Its efficiency is not very high, unfortunately, being 2°C inferior to the well-known Zalman CSL850 grease under peak load.

After you’ve installed the frame on the mainboard, you should secure the cooler on it. To do this, you have to take the special LGA775 clip, pass it through the pipes in the necessary direction and fasten on the frame with two screws. If the clip is perpendicular to the pipes, it is not easy to turn the screws around due to the enlarged dimensions of the cooler – you have to tilt your screwdriver. Well, this is not that difficult as it seems if you’ve got some dexterity. But if you want to avoid this inconvenience, you can position the clip in parallel to the pipes.

As for installing the cooler on Socket AM2/754/939/940, you don’t have to take the mainboard out of the system case to fasten a back-plate to it. The cooler is secured on the socket’s own plastic retention frame. You need to use the appropriate clip for that:

Pass it through the pipes at the bottom of the cooler, attach the L-shaped cap and hitch it on the prongs of the plastic frame. You will require a flat screwdriver to install or remove the cooler because the pressure force is very strong.

Here is how the installed cooler looks like inside a system case:

There are several positions of the retention clips possible, so you can orient the cooler on every platform it supports in the way you want.

You can download the cooler installation guide from the official website (a 2.95MB PDF file).

Scythe Infinity (SCINF-1000)

I have been looking forward to test this cooler because various news, announcements and previews have all been heating up my interest.

So, is it true that the Scythe Infinity provides unprecedented efficiency? Is it the ultimate super-cooler? Let’s check this out.

Package and Accessories

The tall box with the cooler is covered with inscriptions in two languages, with logotypes and names of key technologies this model implements.

The box is divided into three compartments. The top compartment contains a 120mm fan. Then goes the cooler’s heatsink, and the bottom compartment is occupied by a small flat box with accessories.

There is just the necessary minimum of accessories included:

The cooler’s Versatile Tool-Free Multiplatform System (VTMS) doesn’t require screws and nuts, so there are only three pairs of fasteners for Socket 478, Socket AM2/754/939/940, and LGA775. Besides that, there is an Installation Guide, a 1g pack of SilMORE thermal grease, and two wire brackets to secure the fan on the heatsink. It’s sad there’s only one couple of brackets here because two pairs would be most appropriate!


The Scythe Infinity is a very large cooler at 125x116x160mm. Its weight with the fan is 960g.

There seem to be nothing special about this design: an aluminum tower on five copper heat pipes whose ends are sealed with figured caps at the cooler’s top. This seems to be simple, but let’s take a closer look:


I didn’t try to count up, but there are very many aluminum plates here. They form three overlapping sections: a central and two side ones. Thus, the plates have the highest density around the pipes where the heat load is the highest whereas the middle and the sides of the cooler are less dense to reduce resistance to the airflow. But the cooler is large and the plates are placed no wider than 2mm apart in the densest spot, so the low-speed fan shipped with the cooler is going to be too weak to blow through this ribbing well enough.

The top plate of the Scythe Infinity is adorned with an embossed logo of the manufacturer.

The caps on the ends of the pipes look cool, too.

The cooler’s base is protected from scratches with a piece of film:

The copper plate in the cooler’s base is perfectly finished:

It is polished and absolutely flat. This plate is a mere 2.5mm thick and the pipes contact it with their bottom surface only (judging by the traces of solder on the edges, the contact is implemented by means of soldering). I think that Scythe lost something in terms of efficiency here. If they made grooves for the pipes in the base-plate, as is done in the Zalman CNPS9700 LED for example, the Infinity would keep the temperature a couple of degrees lower.  

The Scythe Infinity is equipped with a 120mm fan with seven curved blades.

The fan is rotating at about 1200rpm, creating 46.5CFM airflow and producing 23.5dBA of noise. It is exceptionally quiet in my subjective opinion.

Here are a few photographs of the Scythe Infinity in comparison with other well-known coolers:

Scythe Mine

Scythe Ninja

Tuniq Tower

Assembly and Installation

VTMS technology I mentioned above (Versatile Tool-Free Multiplatform System) doesn’t require you to use any tools or screws as you are mounting the cooler on the mainboard. You only have to insert the appropriate couple of clips into the Infinity’s bottom heatsink and fix the cooler on the CPU. It’s all simple and easy in the manual, but only outside the system case. If you are trying to do that in your PC case, you will find it highly awkward to hitch the fastening mechanism on Socket AM2/754/939/940 on the jags of the standard plastic frame or to press down on the heads of the “nails” of the LGA775 fastener because the heatsink’s bottom plates are too close, providing little room for your fingers. And it is next to impossible to turn or unhitch the LGA775 fastener. That’s why I advise you to mount the Scythe Infinity on your mainboard out of your system case in order to avoid misaligning the cooler.

As concerns optimal orientation of the cooler on the CPU socket, on Socket AM2/754/939/940 platforms this will depend only on the holes in the mainboard and on the orientation of the retention frame. It is desired that these holes were placed in perpendicular to the rear panel of the case so that you could orient the cooler’s heatsink in the most optimal way and achieve maximum efficiency. Alas, the holes are in parallel to the rear panel on my ABIT AN8 SLI mainboard, so the Infinity’s heatsink can only be installed in one position, while its fan in either of two positions:


As you can see, neither position of the fan can be considered good. In the photo on the left the fan receives air from the hot graphics card (I use a GeForce 7950 GX2 in my tests, so this would mean heating up rather than cooling the CPU heatsink because the scorching-hot air from the overclocked card’s blower would be exhausted upwards, right to the blades of the Infinity fan). In the photo on the right, the position of the fan is inefficient as is mentioned by the manufacturer in the user manual and is also confirmed by my own tests on an open testbed with an Intel processor (the difference is 3-4°C under peak CPU load). There is not much space in front of the fan due to the large dimensions of the cooler (its width, in this particular case).

As for installing this cooler on an LGA775 platform, I should warn all the owners of mainboards from the popular ASUS P5B series that the cooler can be mounted on them in one position only (fortunately, it is the correct position). If you turn the cooler around so that its bottom heatsink was in parallel to the rear panel of the system case, the cooler’s base lies right on the coils of the CPU power circuit and you can’t close the latches:

Considering that the Mine model is fastened in the same way, you may have such problems with it, too.

Winding up this section of the review, I want to add that the fans were fastened to the Scythe Infinity heatsink in such a way that their top was on the same level with the top aluminum plate of the heatsink. That is, the fans were always blowing at the cooler’s bottom heatsink.


The following table lists the specifications of the coolers I have described above and of their opponents.

Testbed and Methods

Besides the described coolers, I also tested a Thermaltake Big Typhoon (with its native 120mm fan rotating at ~1320rpm) and a Scythe Ninja (it was tested with the same fans as the Infinity) for the sake of comparison.

I checked out the flatness of the base of each cooler by the trace of thermal grease on a piece of glass. Each cooler had a flat sole. I used a 120mm Thermaltake fan (Thunderblade A1926 model) with a speed controller from Coolink (the speed could be adjusted from 1100rpm to 2000rpm, creating airflow of 42 to 78CFM) to test the Scythe Infinity with two fans at 1200rpm and to test the same cooler, the Scythe Ninja, and the Thermaltake Big Typhoon at max fan speed (2000rpm).

The tested was assembled out of the following components:

The Chaintech GeForce 7950 GX2 was overclocked from its default frequencies to 570/1580MHz to make the test conditions somewhat harder because the air from the graphics card’s cooler remained in the system case. There was no sense in that on an open testbed and the graphics card was working at its default frequencies then.

The tests were performed in Windows XP Professional Edition Service Pack 2. I installed Nvidia nForce version 6.82 and Intel Chipset Drivers version, DirectX 9.0c (dated August 2006), and ForceWare 91.47.

SpeedFan version 4.30 was used to monitor the temperatures and fan speeds on the AMD platform. The temperature of the Intel Core 2 Duo E6300 was monitored with S&M 1.8.1 and, additionally, with Core Temp Beta (the latter program would report a 1.5°C higher temperature of each core than the former).

The CPU was heated up by running the FPU test from S&M at 100% load for 15 minutes. I also simulated a Game test mode by running 3DMark06’s Firefly Forest test with 16x anisotropic filtering and without full-screen antialiasing for 19 times.

The temperature was read from the sensor integrated into the CPU. The mainboards’ automatic fan speed management was disabled for the time of the tests. The thermal throttling of the Intel Core 2 Duo processor was controlled with RightMark CPU Clock Utility version 2.15.

The coolers were tested in a closed system case (which was equipped with two 120mm system fans for intake and exhaust and one 120mm fan on the side panel) and on an open testbed.

At least two test cycles were performed for each cooler in each test mode. I waited for 25-30 minutes for the temperature to stabilize during each test cycle. The stabilization period was two times shorter on an open testbed. The maximum temperature in the two test cycles was considered as the final result (if the difference was not bigger than 1°C). Despite the stabilization period, the results of the second test cycle were generally higher by 0.5-1°C.

The ambient temperature was monitored by means of an electric thermometer and remained at 18°C when I was testing the coolers on the AMD Athlon 64 and Intel Core 2 Duo platforms and 24-24.5°C during the tests on the Intel Pentium D platform (it took me a few weeks to complete this test session as I had to give up it for a while to do other tests).

Each diagram shows the results of my tests in two modes (in a system case and on an open testbed). The coolers are also divided into two groups: quiet mode and max fan speed.

Thermal Performance

AMD Athlon 64 Platform

The Athlon 64 3000+ processor was overclocked from its default 1800MHz to 2800MHz (a frequency growth of 55.6%) with a voltage increase to 1.65V.

CPU-Z doesn’t tell the voltage I’ve set up in the mainboard’s BIOS correctly here as well as for the other CPUs in this review.

Here are the results of the test:

The excellent performance of the Thermaltake Big Typhoon relative to the other coolers is due to its high efficiency inside the system case thanks to the 120mm system fan on a side panel that is driving a stream of fresh air right into the Big Typhoon’s fan. This system fan helped the other coolers as well, yet the Big Typhoon benefited from it the most of all.

The Big Typhoon is also in the lead when cooling the Athlon 64 on an open testbed. The Scythe Infinity is about 3°C better than the Scythe Ninja in a closed system case and 1.5°C better on an open testbed. Note that the Scythe Infinity is very effective inside a system case if you equip it with two 1200rpm fans for intake and exhaust. It performs better then than with a single fan rotating at 2000rpm.

Alas, the Zalman CNPS9700 LED could only sport its blue highlighting, but not good performance. This cooler is obviously far from efficient in this test. It equals the Scythe Infinity at maximum speed and on an open testbed, but the level of noise is too high then.

Perhaps a single-core processor is not enough for the new coolers to show their full potential? Let’s check them out with an Intel Core 2 Duo.

Intel Core 2 Duo Platform

The dual-core processor Intel Core 2 Duo E6300 stepping B2 was overclocked from its default 1866MHz to 3450MHz (a frequency growth of 84.9%) with a voltage increase to 1.5V.

Here are the results of the coolers in this test:

The Scythe Ninja failed this test as it couldn’t cool the overclocked Allendale core in a system case. The system would freeze and the temperature value shown in the diagram is the last reported one. That’s rather strange as the cooler’s base was without any defects. The plates of the heatsinks should have been cooled quite well: the Scythe Infinity with much denser ribbing shows 3.5°C higher efficiency with the same fan.

As for the Infinity, you can note once again that this cooler is more efficient with two low-speed and quiet fans than with a single 2000rpm fan. Moreover, the Scythe Infinity with two fans becomes the winner of the test on an open testbed, i.e. when the cooler efficiency does not depend on the system fans and the size of the system case.

The Zalman CNPS9700 LED does well at max speed and inside a system case, but I don’t think you are going to use this operation mode. It is too loud. Our sample of the cooler may be defective or the heat dissipation of the overclocked processor is still too low for it. The platform with an Intel Pentium D is going to show us if it is so.

Intel Pentium D Platform

Notwithstanding its hotness, i.e. high heat dissipation, the Intel Pentium D processor was overclocked to 3900MHz (a frequency growth of 39.3%) with a voltage increase to 1.4875V.

By the time I had to test the coolers on the Intel Pentium D I had run out of Zalman CSL850 grease. So, I tested the coolers on this platform using Arctic Silver 5 thermal grease.

Here are the diagrams:

This sample of the processor would begin to skip clock cycles (throttling mode) after reaching a temperature of 81.5-82°C. The Scythe Ninja got near this, yet managed to keep the CPU in max performance mode. The coolers from Scythe with a single 120mm fan didn’t do well in a closed system case, although the Infinity is a couple of degrees better than the Ninja.

But note how efficient the Scythe Infinity becomes if you install two low-speed fans on it for intake and exhaust – the temperature is 11°C lower then! This confirms the fact that the cooler’s plates cannot be cooled well by only one fan.

The Zalman CNPS9700 LED shows its worth finally. On an open testbed it is as efficient in the quiet test mode as the Scythe Infineon with its native fan is and outperforms the Thermaltake Big Typhoon and the Scythe Ninja by 3.5° and 2°C, respectively. The latter model looks better than on the two previous platforms.

The Zalman CNPS9700 LED is the most efficient cooler on an open testbed in max speed mode.


There’s a lot of argument on hardware forums all over the Web about the best air cooler for CPUs. A lot of votes, evidence and refutations, links to reviews and test results – you can find all of that there. More often than not such discussions degenerate into a flame war sooner or later.

But when you come to think of it, there is a different super-cooler for each of us that would meet our particular requirements and match the specific CPU, mainboard, system case, room temperature, etc. I guess there is currently no cooler that would satisfy each and every one of us. It is people who don’t or won’t understand this engage into this purposeless argument.

Yes, there is indeed a caste of super-coolers that differ from others in delivering high cooling efficiency at low noise, but there is generally very little difference between products within the super-cooler category. Replacing one super-cooler with another may help you achieve a higher CPU frequency growth, but not as much higher as to justify the cost of the upgrade.

As for the two heroes of this review, the new cooler from Zalman should be appreciated by owners of CPUs with not just high, but extremely high heat dissipation such as overclocked Pentium D (especially the early revisions of this CPU). I think installing the Zalman CNPS9700 on such processors can help you squeeze an extra 100-200MHz from them.

Talking about the Scythe Infinity, I think it would be better if the box with this cooler contained two fans or at least two fasteners for them. Overclockers would have got a leader among air coolers then.