![[H]ardOCP](/images/blank.gif){kind=small}
- Date:
- Friday , July 16, 2004
- Author:
- Steve Lynch
- Editor:
- Sean Quinn
- Google +1
GlobalWin Jefi Watercooling Kit
GlobalWin's latest foray into the world of watercooling is a unique open reservoir evaporative cooling system with a style that has to be seen to be believed.
Next you'll need to start setting up the Jefi unit. Attach the unmarked hose to the water pump and then place the pump into the bottom of the reservoir. Next fill the reservoir about 1/2 way and drop the heatsink and floatation carrier into the tank. Install the four threaded white standoffs to the bottom of the Jefi unit and line up the body of the Jefi over the heatsink and lower it onto the reservoir. Fill the reservoir up until the heatsink makes contact with the four white standoffs. If you have done this correctly the four white standoffs will fit between the fins of the heatsink. Add a little more water until the heatsink is partially submerged under water.
After the Jefi is assembled, take the remaining hose (marked with a purple arrow) and attach it to the fitting on the Jefi body. This is a compression fitting, so make sure you unscrew the compression ring from the valve, thread it onto the hose first, then install the hose and tighten the fitting down tight. Run the three lead fan wire to the fan control unit you installed earlier and then wrap both hoses and the wire with the included wire loom.
As stated earlier, the temperature probe is positioned so that it will be submerged into the coolant. (The manual says that the probe can be positioned outside the water for continuous operation but we found this to have little to no impact on system performance). Plug the pump into the wall outlet and cycle the water through the system for 12 to 24 hours to check for leaks. Once the system passes the leak test you are ready to fire up your system.
System Configuration & Testing Procedure:
We decided to see just what the Jefi was made of so we broke out an Athlon64 3000+ as well as a 2.8GHz Prescott. We know that the Jefi is not targeted towards the overclocking or performance crowd, but we wanted to see how well the Jefi would perform in an overclocked environment in addition to everyday cooling. For comparison we ran the Koolance Exos as a baseline. On the Koolance, we used the CPU300 block that supports both the socket 478 as well as the Athlon64. The fans on the Koolance were set to run in Mode 1 (Fans operate at 45%, speeding up to 100% when the temperature reach 113°F/45°C ) for all tests. The Jefi fan was set to the #1 position for full speed operation as well.
ABIT IC7-MAX3, Intel 2.8GHz Prescott clocked at 2.8GHz (14 * 200FSB) and 3.36GHz (14 x 240FSB), MSI K8T-NEO (VIA K8T800), Athlon64 3000+ clocked at 2GHz (10 * 200FSB) and 2.15GHz (10 x 215FSB), 1024MB Corsair XMS PC4000, VisionTek Radeon 9800 Pro, 80GB Seagate SATA/150 hard drive, Vantec 550w PSU.
Our Testing Methodology:
Using the on-die sensor of the Athlon64 3000+ and 2.8GHz Prescott, we are able to see temperature changes almost immediately as the load changed on the CPU. We used Prime95 and a UT2K3 16 Player Bot match to load the CPU. The CPU is operating at "100%" according to WinXP task manager, which is exactly what we need for temperature testing. Motherboard Monitor 5 was used to monitor temperatures during testing. Temperature readings are taken from the on-die temperature sensor on the Athlon64 3000+. Prime95 was run simultaneously with a UT2K3 CTF map running in the background with 16 bots going at it to get our temperatures as high as possible for a sustained period of time for testing. With the noted two applications running, we allowed the CPU to stay at 100% load for 30 minutes and then observed the temperatures and kept an eye on them from there. What we're looking for is where they peaked and leveled off. This is the temperature we recorded. These tests are repeated five times for the average temperature. The ambient temperature in the test area was 81°F. We kept the temperature level for the duration of the testing in order to make our test results as accurate as possible. The different ambient temperatures and other test variances will result in having test temperatures unique to this test, making comparison to other previous [H] testing impossible.
Test Results::
All test results are recorded in Fahrenheit. I'm including this link to an awesome Fahrenheit to Celsius Converting Tool for all of our non-USA readers. The newest version of the Google Toolbar will also do temperature conversion by simple typing it into the Google navigation box. By typing "48C in F" without quotes, Google will convert the temperatures for you.
In our first round of testing we see that the Jefi system bested the Koolance Intel numbers at idle and the Jefi Athlon64 came very close to the Koolance Athlon64 numbers as well. Overall a very impressive performance. Let's see how the Jefi does under load.
The Jefi simply blew my hair back with temps running neck and neck with the Koolance system under load. Remember, the Koolance system is running the CPU 300 water block and all three fans running at full speed. Hats off to the Jefi.
Overclocking was a whole different ballgame. Initially the Jefi appeared to perform well, but the longer the tests went, the more apparent that the Jefi wasn't going to handle the heat as well as the Koolance.
The Jefi just couldn't handle the heat load that these overclocked processors put out while under load. Idle conditions were one thing, but this kind of heat sent the Jefi temps soaring. Keep in mind these temperatures are respectable compared to an air cooled computer, they just cant keep up with the Koolance while overclocked. The Intel system reached 151°F during testing which caused the system to hang after an hour at that temperature. On the whole, the Jefi's performance was very impressive, cooling even the uber-hot Prescott like a champ under extended load conditions. While the Jefi couldn't hang with big overclocks for extended periods of time, the unit did handle mild overclocking just fine. When the Prescott was clocked down to 3.15GHz (14 x 225MHz) the system performed fine for hours on end. Something to keep in mind.