Performance Data, Benchmarking and Analysis

We are going to start out with a quick snap shot of CPU-Z. This is our test bed configuration and we will demonstrate the performance and capabilities of this product using this system.  Our initial test will include complete default system settings.

The new Prescott platform is a real overclocking beast. These CPU's can really take a clock beating! For the first time ever, we are able to run this system @ 4.14GHz all day long! We play games, benchmarked and written this review using the system at this clock.  

Hardware Configuration:

• Cooler Master Stacker Model STC-T01-UW Case
  • Both Side Panels in use
  • 1 X 120mm Intake Fan
  • 1 X 120mm Exhaust fan
  • 2 X 80mm Exhaust Fan
  • 1 X 80mm Exhaust port (Top of Case)
• LANParty™ 925X-T2 Motherboard
  • Award BIOS Revision 2.1
  • CPU Voltage Identifier (VID): +1.328V
  • DDR2 Voltage Identifier (VID): + 1.800V
  • NB Voltage Identifier (VID): + 1.500V
  • Southbridge: 82801 (ICH6)
  • Chipset: i925X (Alderwood)
  • Bus: 1 x PCI-Express x16, 3 PCI-Express x1
• Intel® Pentium® 4 550 LGA 775 3.4GHz CPU
  • D0 Revision Stepping 4 CPU
  • 1024KB L2
  • Voltage Identifier (VID) +1.328V
  • 17 X Multiplier 
• Corsair DDR2 XMS2 Pro 5400C4Pro 675MHz
  • JEDEC standard 1.8V ± 0.1V
  • CAS# 4-4-4-12
• Antec Neo Power 480watt Modular PSU
  • Continuous Power 
• NVIDIA PCX 6800Ultra 256MB ( NV45 )
  • Engineer Sample BR2.1
  • PCI-Express
  • 256MB GDDR3
• NVIDIA ForceWare 70.90 Beta
  • Default Settings
• 2 x 120GB Maxtor SATA ICH6R RAID0
  • 7200 RPM
  • 8MB Cache
• Windows XP Service Pack 2 2600
  • Direct X 9.0c

Voltage Configuration:

• CPU Voltage : 1.33V
• RAM Voltage: 1.79V
• NB Voltage: 1.50V
• Aux Voltage : 1.79V
• +3.3V Voltage : 3.35V
• +5V Voltage : 5.25V
• +12V Voltage : 12.03V
• -12V Voltage : -1.83V
• -5V Voltage : -4.10V
• Standby Voltage : 4.97V
• Battery Voltage : 3.15V

Environment conditions:

• Ambient Room Temperatures:  26.1°C / 79.0°F

Hint:

• Radiator fan can be run at +5v, +7v or +12v for user convenience. At +12v this allows slightly better performance at the cost of sound increase. At +5v the fan is virtually silent but you will lose a very small amount of performance, which is barely a degree difference. Using a multi channel Rheostat will allow real time adjustment of the fan. We chose to run our tests using the fan at 7v For us that’s the sweet spot! Right down the middle! At this voltage all we could hear was the Power supply, and the hum of the hard drives.

Test & Comparison:

Stock Intel Retail Box LGA775 Heatsink Fan

• Board Temperature : 41.0°C / 105.8°F
• CPU Temperature : 44.0°C / 111.2°F
• Power / Aux Temperature : 42.0°C / 107.6°F td
• Ambient Room Temperatures:  26.1°C / 79.0°F

Swiftech H20-120™ REV. 3

• Board Temperature : 36.0°C / 96.8°F
• CPU Temperature : 36.0°C / 96.8°F
• Power / Aux Temperature : 31.0°C / 87.8°F td
• Ambient Room Temperatures:  26.1°C / 79.0°F

Using default voltages we over-clocked the FSB and took sample readings using Sandra 2005 at various intervals and ended up with these results. The system was left at idle for 20 minutes between each clock change and we took 3 readings 3 minutes apart to necessitate further accuracy with our sample readings. We ran the Radiator fan at 7 volts for convenience.

100% CPU Load Temperatures:

Seti @ Home Work Unit - LGA775 Heatsink Fan

Board Temperature : 54.4°C / 129.9°F
CPU Temperature : 56.5°C / 133.7°F
Power / Aux Temperature : 48.0°C / 118.4°F td
Ambient Room Temperatures:  26.1°C / 79.0°F

Seti @ Home Work Unit - Swiftech H20-120™ REV. 3

• Board Temperature : 40.0°C / 104.0°F
• CPU Temperature : 47.0°C / 116.6°F
• Power / Aux Temperature : 42.0°C / 107.6°F td
• Ambient Room Temperatures:  26.1°C / 79.0°F

For this test we are using the popular data crunching application SETI @ Home. This is believe it or not a good application to force the CPU into 100% load. We allowed SETI to run for one hour under each clock and we again took three sample readings about 45 minutes into the packet to ensure temps where stabilized.

Real world results are individual anomalies. Synthetic results are marginally inaccurate, however if you mix the two together and use common sense and carefully study your systems performance using this kit, you will come to the conclusion this system rocks!  While we can provide these results using our test bed under our environment and conditions, the results will vary from system to system. There is no way to accurately say these results will work for everyone, but the fact remains this is one superb cooling kit!

Comparing our results to Swiftech’s published data:

We wanted to see if Swiftech’s published performance specs jived with our own findings. Below is an excerpt of their web site information:

“To characterize the kits' performance with different configurations, each kit was installed in a bare chassis with a heat die simulator and then tested in an environmental chamber where the radiator fan inlet air temperature was held at 25.00°C ±0.02. The thermal resistance of the system could then be characterized by the difference between the fan's inlet air temperature and the heat die, divided by the heat load in Watts; such ratio is commonly referred to as the "C/W".

Users comparing their or other systems and data are cautioned that such comparability is tightly dependent upon how accurately the heat load (Watts) is characterized.  

Kit Model	Radiator/Fan configuration (*)	Heat load Watts (**)	Earlier version H20-120 with F&B kit System °C/W (***)	H20-120 Rev 3 with Reservoir System °C/W (***)	Fan Noise, dBA (per specs, and est.****)
H20-120™	1 radiator	115 Watts	0.230	0.220	34
H20-120™	2 radiators in series	115 Watts	0.200	0.194	~37

* Testing performed with the fan(s) shipped with the kit
** actual Watts measured at the heater of die simulator
*** Noting that "C/W's are specific to an individual test bench (due to the lack of standard test methods); the system C/Ws described are those actually determined with a specific water-cooling kit under the conditions described above.
**** Fan noise estimate based on: "If the speed of a fan is reduced by 20%, the decibel level will reduce by 5 dB.” and "addition of a second equal source increases the sound level by ~3dBA".

To project the processor temperature at full load using Swiftech’s C/W ratio we need first to establish the power dissipation of our Pentium 4 550 processor – Intel sSpec number for our processor is SL7PZ, and its maximum heat dissipation is given at 84 Watts.

So, at stock speed, Swiftech’s projected temperature would be:

84 Watts x 0.220 (their published C/W ratio) = 18.5 °C
 
To this number, we would then add the ambient air temperature, which was in our case: 26.1 °C  =  44.6 °C

If we now look back at our test data, we see that the CPU probe reported  47 °C. We then see that there is a 2.4 °C (5%) difference between the measured temperature and the temperature calculated mathematically. Not too bad I’ll say, knowing that there are many variables that could slightly affect our own results.

AquaMark Score: 75175

This test was performed using NVIDIA ForceWare 70.90 drivers, installing coolbits, overclocking the GeForce 6800 Ultra PCX to 456MHz Core / 1220MHz Memory. We put the FSB at 17 x 244 = 4.148GHz. Using Mushkin DDR2PC4200 DDR2 1 GB Dual Channel at 4-4-4-12.