WaterChill CPU/VGA/Chipset Water-Cooling System

Testing

I tested the WaterChill system on an AMD Athlon XP-2400+ running at stock speeds and voltages.  This is the same reference system I use for all testing but this time I mounted it inside the modified Lian Li PC60 case.  The test computer components include:

• Epox 8RDA+ nForce2 motherboard with 3129 BIOS
• AMD Thoroughbred XP-2400+ @ 2,000 MHz and 1.65 Vcore
• 512 MB Corsair XMS PC3500C2 DDR RAM (3-2-2-2.0) 
• Antec TrueControl 550 watt PSU
• 80 GB Western Digital SE 7200 RPM HD, Plextor CD-R/W, FDD
• Gainward Geforce4 MX 440 AGP
• Windows XP Pro SP-1 (with all critical updates)

Measuring equipment used:

• CPU and GPU — Barnant Model 115 digital thermometer ((+/- 0.4ºC)
• Ambient air and Chipset — Fluke 52 II digital thermometer (+/- 0.2ºC)
• Water temperature — measured in reservoir with digital thermometer
• Calibration — ASTM-T mercury precision thermometer (+/- 0.1ºC)
• Extech Model 407736 sound level meter (accuracy +/- 1.5 dB)

The CPU was fully loaded by running the distributed computing application Folding@Home and the GPU was loaded by running continuous loops of 3DMark 2001 SE.  Before mounting, all the chips and waterblock bases were cleaned with isopropyl alcohol. A very thin coat of AS5 was applied to the CPU core and Arctic Silver Ceramique was used on the GPU and Chipset.  I used a calibrated thermocouple attached to the main processor core to record CPU temperatures and thermocouples attached to the backside of the video card and motherboard to obtain readings for the GPU and nForce2 North Bridge.

I found that for my particular computer, connecting the Chipset and VGA waterblocks in parallel with each other after the CPU waterblock produced slightly better results than running all three waterblocks in series.  This flow path arrangement is quite easy to implement thanks to the dual outlet Antarctica waterblock… 🙂

The ambient room air temperature remained a relatively constant 23.5°C ± 0.5°C during testing and the water temperature was measured in the reservoir.

The following data is presented for comparative purposes only.  Your actual results may be different depending on the variables unique to your system (CPU, video card, ambient air temperature, overclocking, case air flow, temperature monitoring, etc).

Looking at the system as a whole we see that the average water temperature is running 7.4ºC over the ambient air temperature with fans on low (7V) and 5.6ºC over ambient with the radiator fans on hi (12V).  This is a good indication that the radiator/fan combination is sized appropriately for this system.  Keep in mind though that I am not running particularly hot components in my test rig nor are they overclocked.  As the system heat load increases the average water temperature will go up (along with all the processor temperatures) unless a larger radiator and/or more airflow is used.

As mentioned before, I found that I could obtain slightly better CPU temperatures by connecting the VGA and Chipset waterblocks in parallel.  Neither the GPU nor the NB chipset is dumping much heat into the system so there was little affect on their temperatures but the increased flow rate helped cool the CPU an additional ~2ºC over the series flow path.

I’ll be the first to admit the GeForce4 MX GPU isn’t the hottest video processor on the block but it serves my test rig needs.  The Asetek VGA waterblock had no problem keeping it nice and cool.

The NVIDIA nForce2 NB chip on this motherboard is normally passively cooled with a large aluminum heatsink and no fan.  Notice how the NB temperatures closely follow the average water temperature — when the water temperature goes up, so does the NB temperature.

The Asetek WaterChill system did a very good job cooling the stock Athlon XP-2400+ processor, GeForce4 GPU and nForce2 North Bridge.  These numbers confirm that Asetek has done a good job of matching the capacities of the various components so that they perform well together in a balanced system.