Introduction

Corsair first dove into water cooling in 2003 with the introduction of their Hydrocool unit ([M] review). It was a self-contained unit and seemed to have all the proper ingredients for a successful design, yet just two years later it's been phased out of Corsair's product line-up. Although it was fairly popular, perhaps Corsair engineers simply didn't like the idea of a stand-alone unit, which of recent, has been contraindicated in today's "smaller, better, faster..." ergonomic world. Today PC-users want sleek integrated designs with a minimum of fuss and maximum performance. This is not an easy order to fill. Corsair, not easily deterred by such technological challenges, may have found just the alternative. Based on my reading almost every Hydrocool review published and the data we'll be presenting today, I believe COOL may be a superior product.



In the Box – First Impressions

COOL arrived in perfect condition as do most Corsair products. I found the box to be hefty but was pleasantly surprised to find Corsair neatly packed everything into a single box.



After removing the box I placed the unit in front of the intended "target." The venerable yet crazy-making DFI LAN party nF4 Ultra motherboard was installed in Thermaltake's SHARK aluminum full-tower case. This case is specifically made to house water-cooling systems. Clear water on the Corsair box reminded me of my childhood when we got our first above-ground pool.



Opening the COOL box reveals two layers of foam with pre-formed compartments, each compartment securely holding a specific component. Placed side by side in the photo below, you can see all the included parts in their individual foam compartments. The packaging should withstand just about any shipping "incident" thrown at it. From the art-work on the box, to the installation manual and included CD, it becomes evident Corsair put their collective hearts into this project.



In the photo below we take a closer look at the included mounting hardware and other accoutrements which we'll cover later.



Next we move onto the "primary heat exchanger" a.k.a. water-block. COOL features nicely finished mounting brackets suitable for Pentium 4 Socket 775/Athlon XP or Socket 478/Athlon 64. For our purposes we'll be using the Socket-939 compatible hardware. The other accessories in the photo below are sitting in the radiator/fan cage assembly which will be discussed later.



In the photo below we see a close-up of the CPU-cooler with its pre-affixed (clamped) hose sections.



On the next page, we'll delve into a description of each piece of included hardware ->

At the heart of the COOL system is the LAING Delphi 12V DC pump. While the LAING is a "Low Flow" model, it is definitely a high-pressure part with an approximate 13-foot maximum head. The pump runs virtually silent.



Next the reservoir which fits into a 5.25" bay is made of plastic and features brass inserts for mounting screws, a simple yet important detail. Another utilitarian detail on the reservoir is its "hump back" where the fill-cap is located making bleeding your system much easier since air will naturally rise there.



As important as any other device in a water cooling system is the radiator/fan combo. Following the First Law of Thermodynamics the kinetic energy (heat) which has now entered the closed system via the water block, must then be expunged from the system. In water cooling this occurs within the radiator with the assistance of a fan. Of course it can also be done without a fan as in passive systems.



Mounting hardware supplied for the radiator/fan combo include a two-piece mounting cage Corsair has named this the RMK (Radiator Mounting Kit). One half of the RMK cage mounts to the back of the case with the second half affixed to the radiator/fan. Four screws fasten the cage halves together from top/bottom and left/right which allow easy access to the screws. The RMK assembly supports and extends the radiator out from the rear of the case approximately 2.5cm so as not to disturb nor inhibit air-flow.



Corsair instructions explicitly state the fan should be oriented to push air through the radiator. Beyond this their RMK provides mounting options limited only by your imagination or case type. Taking a closer look below I've assembled one half of the cage to the radiater/fan and mounted the "anchor" assembly to our Thermaltake SHARK case.



I tested several variations on manufacturer sugested mounting instructions, each over several days, across several weeks while recording the data. Although Corsair strongly suggests oreinting the fan to push air through the radiator, I found just slightly better results reversing the fan direction so air was pulled through the radiator. Below the fan is pulling air through the radiatior . I also reversed the radiator direction so there were no sharp bends in the hose and tested this variation as well comparing results. While following Corsair installation instructions verbatim may not have bent the hoses enough to restict waterflow, the data seemed to support better flow as is oriented below. I decided to ere on the side of caution.



Ultimately I settled on removing the RMK assembly, mounting the fan/raditor externally, reversing radiator direction thereby minimizing flow restriction and followed Corsair's recommended fan direction. In this respect I took advantage of the cooler ambeint room air, eliminated sharp or restrictive hose angles and in keeping with Corsair's suggestion to push air through the radiator, cooled mosfetts near the socket as well. In atypical internal mounted watercooling systems where the radiator/fan combo is located at the rear of the case, the fan usually serve's a dual purpose. First to cool the water, and then to act as the case exhaust. I strongly advise eschewing this formula wherever possible as pre-heated case air (while still much cooler then the CPU temp) is often much higher in temperature compared to the ambeint room air temp. While the settled upon method in this case (pun intended) may seem to fly in the face of convention, attempting to create a vaccum in Thermaltake's SHARK tower would have been in vein due to it's "honeycomb" side-panel. Since air can flow freely through the honeycomb panel located just beneath the CPU area, I settled on the option below.



The LAING pump utilized by Corsair is an excellent choice due to its extremely high pressure capabilities. At approximately 13-head feet and 400GPH, the unit is a wonder in the world of water cooling, especially given its miniscule size. The LAING Delphi 12V DC is a virtually silent operator.



Looking at the pump from another angle we see the included PCI-slot plate which has holes for the tubing to pass through and a quick connect leading to the radiator fan. To reiterate, this pump operates virtually silent, and its unobtrusive size allows plenty of component elbow room.



Time for testing ->

Working our way up to the reservoir, Corsair specifies in the installation manual this should be placed at the highest point in the system. I installed the reservoir in the second highest 5.52" bay which made filling and bleeding a breeze.



Onto the waterblock. The design is very similar to Swiftech's MCW6000 and with a good reason, as Corsair is working together with Swiftech on the "COOL" water cooling kit. It's not that Corsair doesn't have the wherewithal to begin manufacturing water blocks from scratch; it's simply a nonsensical approach. Out-sourcing is much more common in the technology industry then many others due to the varying levels of specialization. In fact Corsair's choice indicates they've carefully researched options available which not only offered decent performance, it does so at an affordable price. While there may be better performers out there the MCW6000 most likely represented the best ROI (Return on Investment).



I was unable to find actual internal photos of the MCW6000, or the 240-pin version used by Corsair, however; I've borrowed the diagram below from Swiftech's site. The block is described as an extruded pin design, featuring 281-pins with the entire block made from C110 copper. Copper offers superior heat transfer (or conduction) properties over many alternate metals. Do note that the Corsair "Cool" product website states 240-pins instead of Swiftech's 281. Is this a typo? Or did they get an adapted version? Some believe the lower pin count will increase flow rate thereby improving COOL's perfrmance over that of it's counterpart the MCW6000. Aspects (less flow restriction) was most likely more condusive to COOL's accompanying components, and their related placement in the system.



Corsair clearly states in their installation manual the block should be positioned with the outlet situated above the inlet. For thermal transfer paste I used Arctic Silver Ceramique, placing a rice-sized grain on the center of the A64 3500 hetaspreader. I chose Ceramique over Arctic Silver AS5 due to it's shorter set time, and due to the fact AS5 tends "adgere" to many CPUs. I've found the bond strong enough to pull CPU's from thweir socket except of course the LGA-775. This is due to catalysts used with the silver particles, and is indicative of it's excellent ability to infuse or fill every striation, and micro-pore in both CPU and heatsink/waterblock surfaces. The result is an air-tight seal which ultimately forms a bond. There's always a price to pay when using the very best, Unfortunately the combined surface area of the A64 3500+ and Corsair COOL owuld be pre-disposed for just this situation to occur. I allow the natural compression of tightening the mounting screws to set the paste as spreading can incorperate air into the mixture.



The next photo shows the system installed, and side panel removed from the SHARK case. I ran the system at LOAD alternating to IDLE, and shutting it down over several days to ensure the paste set properly. I'd also like to thank a reader whom brought to my attention I had orignally included the wrong photo's as I experimented with juxtiposing waterblock orientation placing the outlet below inlet. Thank you.



And finally with the radiator/fan in the correct (Corsair recommended) positon. As I stated earlier I tested the system several ways, each configuration for at least 48-hours to be as thorough as possible.



Test Setup and Methodology

Test System
CPU	Athlon64 3500+ Winchester .09-micron process (2.2GHz/512k L2 cache) 1.365V default Vcore (Retail)
Mainboard	DFI LanParty nF4 UTDXG (BIOS v.310)
Memory	Corsair XPERT 3200 1024MB DC(2x512mb) CL 3-2-2-5
Graphics	BFG OC 6800GT PCI-ex (370MHz Core/1000MHz Memory) nForce VGA drivers 6.6.9.3
Power Supply	Thermaltake Silent Purepower 680APD
Operating System	Windows XP SP2
CASE	Thermaltake SHARKAluminum Full Tower

Temperatures were measured using Smart Guardian on-board software which reads the A64 internal thermal diode out-put, as well as all DFI motherboard temp related diodes/thermistors. For more precise CPU measurements I re-mounted the water-block several times inserting a thermistor between the base plate and CPU surface. I placed a thermistor fed into a TTGI USA Fan Master SF-609 as close to the center of the A64 IHS (Integrated Heat Spreader) as possible, using the best (highest temp) out of three mountings. To measure water-temps a Cooper Atkins 1246-01 calibrate able thermometer was placed at the center of the water flow, so as not to be effected by ambient temp near the hose edges. Finally a Fluke-187 digital-multi-meter (accuracy +/- 1.0% + 1.0C) was used wherever feasible to measure ambient temps, and corroborate the accuracy of other temps.

To push our A64 3500+ to 100% LOAD, I used the CPU stress utility appropriately named S&M. The utility has many useful features such as an extensive voltage and temp monitoring GUI which records maximum and minimum values during its operation. I ran S&M several times through its 19-minute cycle, although I've discovered just 60-seconds would result in the maximum temp so long as the system had been running throughout the day. Regardless I ran the test through a single full 19-minute run, began a second to circumvent any AMD thermal throttling features even though they were disbaled in the DFI Lanparty nF4 UT BIOS v.310.

For our first screenshot the processor is running IDLE under default voltage/frequency settings: 200FSB/2210MHz at 1.365Vcore.



In our next screenshot the processor was pushed to LOAD and once again under default settings: 200FSB/2.2GHz at 1.375Vcore.



In our final section we determine how far we can overclock the A64 3500+ with Corsair COOL extirpating the increased heat ->

Overclock test

In our overclocking tests I continued to raise the FSB and stress the processor until S&M either caused the system to freeze or reboot. The amount of heat Corsair's COOL dissipated allowed me to push the frequency beyond 2.7GHz, while concurrently raising Vcore to 1.550 + 113% or 1.72Vcore.

In the screenshot below the system is running IDLE overclocked at 250FSB/2750MHz under 1.72 vcore.



For our final screenshot once again running overclocked at 250FSB/2750MHz under full LOAD at 1.72Vcore.




Test Summary



I was pleasantly surprised with the overclocking head-room afforded me using the Corsair COOL H20 system. The kit allowed me to run stable up to 2750MHz at 1.72 vcore. Taking into account the Althon64 is a completely different animal from the Pentium 4 where CPU architecture is concerned; a 550MHz overclock is most promising.

Any P4, from the Socket 478 Northwood to LGA775 Prescott would be an exceptionally rare poor performer if it were to max out at just 550MHz above default speed. Yet one must remember Intel's seemingly unlimited headroom is actually indicative of its controversial pipeline depth and speed vs AMD's "Brainiac" engineering approach. This is the P4's Achilles Heel as pointed out in Van's Hardware articles.

Today we've experienced some excellent performance and as the screen-shot below indicates, there's even more headroom. I chose not to push the processor further due to my fear of damaging mosfets and as yet unresolved issues specific to the DFI LAN party nF4 UT.



Temps at this speed are obviously conducive to further overclocking; in fact the only issue holding me back was HDD file corruption which prevented me from entering the Windows XP splash screen. There was no problem booting and posting beyond 270FSB were it not for the repeated file corruption which can be circumvented running Windows 2000 and/or installing an ATA HDD and adjusting PIO modes. That, however; is another article all together.


Conclusion

Corsair's COOL offers excellent water cooling performance, virtually noiseless operation, reliability, and outstanding value. I must commend the memory maker on their selection of water cooling components which come together to create a cohesive H20 system. This product is another recent release from Corsair such as their XPERT series memory exemplifying their desire to meet the need's of the budding overclocker as well as anyone wanting to improve the performance of their PC. What impressed me most about COOL is the price, Cool can be found through PriceGrabber under 200.00 USD. Given the value I can find no Con's in my assessment of this product. I must confess this is one case in which the semantics behind the moniker "COOL" are well deserved! A Must Have recommendation is in order.

I'd like to thank Anna K. and Robert P. at Corsair for their cooperation.

Questions/Comments: forum thread