Our test system today is based on the Socket-754 Chaintech ZNF3-250, and Athlon-64 3400. The Hydor L-30 feeds a 1/2"system. The Black Ice Xtreme radiator with 2x120mm/90CFM Sunon's in push/pull is also located outside the case. Since the CF-1 Rev.2 is based on the 1/8" ID Euro standard, adapters are connected approximately 6cm off the Cool-Cases CF-1 connector's, as shown below;





Test System Spec's

Cooling-system: Pictured below my test rig in reference to the AC, and the location of cooling hardware. I was able to maintain a constant ambient room temp of 20C/68F (+/-2C) running an EmersonQuiteCool 7K-BTU air conditioner. This is located in a window approximately 1-meter from the test system, the curtain directs air-flow towards my monitor, so as not to artificially effect the system temp. The double heater core pictured was not used during these tests.





CPU - Athlon64 3400 CAAOC 0402SPMW
Motherboard - Chaintech ZNF3-250 (reviewed here by Hexus)
RAM - Corsair Twin-X1024 3200XLPRO (2-5-2-2), OCZ Technology EL PC4200 1024MB (2.5-6-3-3), OCZ Premiere PC4200 1024MB (2.5-7-3-3)
VGA - Sapphire Radeon X800Pro (Catalyst 4.6)
HDD - Maxtor Diamond Max Plus 9 (SATA150 120GB)
PSU - OCZ Technology Power Stream 420
Case - TTGI TT-201T3 5x80mm fan brushed aluminum.
Software - Windows XP SP1, SiSoftware Sandra , ScienceMark 2.0, FurtureMark 3DMark2001SE, 3DMark03, Hexus PiFast Challenge, CPU-Cool 9.0, AquaMark 3, H'oda's WCPUID , CPUID's CPU-Z .

The pump used during this test was the Hydor L-30 (320GPH non-Compact version).



First up we'll test the CF-1 Rev.2. The block mounted fairly easily on the Chaintech ZNF3-250, although I dislike the instability of just two through board screws for Socket-754. My girlfriend found it quite funny I had inadvertently placed the 754-mounting bracket upside down. After filling, bleeding, and leak testing the system externally as usual, I went to mount the block and realized the Socket-754 top-plate had to be juxtaposed. This involved detaching both 1/2" hose adapter's, some bleeding, and some crying. Always check your mount prior to mounting the board in the case;





The photo below is the CF-1 Rev.2 with all test hardware ready to go (or going);



Below a macro shot of the CF-1 Rev.2 mounted on the Chaintech (correctly). When I'd first ran this system I tried the AMD stock heatsink included with the Retail 3400. The resulting temps using that unit were 50C to 55C on-die operational temps, with ambient room temp being 21C/71F. Since overclocking is very difficult with the ZNF3-250 due to its lack of multiplier adjustment, most tests were done at, or just above default speeds, and voltages.





In the photo above, the outlet (return) is located on the top. Recently I was critiqued for my hypothesis; positioning the block so the outlet was on the bottom might improve flow based on gravity (albeit minute). A few H20 Enthusiasts countered; in a closed (pressurized) system, this would have no effect. Furthermore, a close friend suggested; due to the water being heated and the fact hot water rises, locating the outlet on top may be the optimal position. I'll test this hypothesis in the future. I digress.

Unfortunately Chaintech has only one BIOS version dated 01/06/04. Six months later they still haven't released a BIOS providing their flagship Socket-754 (nForce-250) mobo with multiplier options? Therefore the highest 3DMark03 stable frequency I could maintain was 235FSB. This required the maximum 1.70Vcore to achieve. I switched between CPU-Cool, Digi-Doc, and HMonitor (Hardware Sensors Monitor), verifying their accuracy using an external TTF-103 thermistor pictured below.





I verified the accuracy of this Thermistor to within +/-2C by wrapping it in plastic, and submerging it in an ice-bath. I mounted the thermistor as close to the CPU core as possible without adversely affecting the block pressure. To determine the discrepancy (if any) between CPU thermal diode, and an externally mounted thermistor. I remounted the block several times recording the difference. To measure room ambient temps I used a Cooper 1236-32 professional grade thermometer calibrated in an ice-bath.

While seeking the ideal thermistor location I used Arctic Silver Bill A Goop thermal paste, kindly supplied by Arctic Silver (thank you Nevin). The advantage of this paste, is its ability to set rapidly for testing purposes. Once the ideal thermistor location was found I replaced the Bill-A-Goop with Arctic Silver AS5 applying the paste in a rice size grain on the center of the IHS (do not spread).

I compared the accuracy of Digi-Doc using the thermistor (as stated above), and compared this to the accuracy of other utilities such as Podien's CPU-Cool which will be used in the first few screenshot's. Temp-1 measures the CPU's internal thermal diode, Temp-2 measures the Chaintech ZNF3-250 thermal device or "System" temp. Our first screenshot below shows temps at default speeds (11x200FSB = 2200MHz), under IDLE conditions. Cool-Cases CF1 Rev.2 provides the cooling.





To test LOAD conditions I used Seti@Home V.3.08. I believe Seti it's just as adept at raising system temps as CPU-Burn. Undoubtedly some will argue this point, however; if you check CPU-Burn's Homepage you'll find their own graphs indicate CPU-Burn may reach higher peaks, however; Seti@Home overtakes CPU-Burn under many circumstances. I tested CPU-Burn finding Seti to provide higher temps. I ran Seti until a maximum temperature was maintained, usually after just 45-minutes. Our next screenshot represents CF-1 performance under LOAD at 200FSB or 2209MHz.





These initial results bode well for Cool-Cases CF1 Rev.2. There was only a 6C increase in temps during the LOAD test. The next screenshot is a measurement of temps with the system overclocked to 220FSB for 2420MHz (default Vcore). This was the overclock ceiling running the A64 3400 at default Vcore.





Here the CF-1 Rev.2 does very well with temps only increasing one degree centigrade under LOAD.
Now onto initial test results for the Oregon water block maker, Danger Den.

It's certainly nice to utilize a single outlet barb which is a departure from Danger Den's RBX twin-outlet barb.



As I stated earlier room temp was maintained at a constant 20C/68F (+/- 2C) over several days of testing. Even with the side-panels removed from the TTGI-USA TT-201T3 case, the system temp climbed as high as 31C/86F. This leads me to believe the ZNF3-250 most likely measures power circuitry feeding the CPU, as opposed to a simple thermistor mounted on the motherboard surface. The screenshot below indicates temps at default speed/voltage (11x200FSB=2209MHz) TDX providing the cooling.




All things being equal (including ambient temp) Cool-Cases CF-1 Rev.2 and the TDX are practically equal in their ability to cool the Athlon64 3400 at IDLE. Next we will look at TDX temps under LOAD, still running at default speeds;



Once again these blocks are very close in their ability to keep things cool. While the Cool-Cases CF-1 Rev.2 does have a slight advantage. In the next screenshot the A64 3400 is running at 225FSB, for a frequency of 2475Mhz. This required raising Vcore to almost 1.7V, 1.65V was stable, however; during the most intensive 3D benchmarks the system would occasionally reboot at 1.65Vcore. Again TDX provides the cooling.



I've included a 3DMark03 screenshot for those interested in the performance of the Sapphire X800Pro, and Corsair's new 2-5-2-2 Twin-X1024 3200XLPRO, featuring "Plug and Frag" technology which allows DDR400 to run as high as DDR500. Below the system is overclocked to 235FSB (2584MHz), cooled by the TDX. Vcore at 1.65, and latencies at CAS 2-8-4-4 (still holding CAS-2).



Onto a comparison graph and our conclusion ->