EVGA CLC 120, 240, 280mm Liquid / Water CPU Cooler RGB Roundup Review: Page 6 of 8
Posted by Mark Taliaferro on Monday, October 9, 2017 - 8:00am
Test Equipment & Setup
We pondered what software to use for stressing the CPU 100% and OCCT and Prime 95 as well as Intel Burn just product an unrealistic amount of heat. Some of the CPU stress programs are designed specifically to generate heat which has no real bearing on real-life computing. We tested a host of software from Realbench to never heard of programs and finally landed on CPU-Z and the Stress CPU utility. It's easy, free and available to everyone and during test selection, it consistently kept 4 cores 8 threads at 100% and generated enough heat to convince us of the validity of the Stress Test.
Notice the Stress CPU Button, we used 4 Core 8 Thread testing
We went with a worst case scenario, no chassis fans, no ambient air movement. Lab temperature was kept at 72oF or 22oC throughout testing. We disabled Intel's Turbo feature and locked the CPU to a steady frequency and CPU voltage for each clock speed on the CPU. We went into BIOS and set OC profiles for 4.5GHz, 4.6GHz, 4.7GHz, 4.8GHz and 4.9GHz. We cut off at 4.9GHz as the CPU voltage required to reach 5GHz stable with a stress test load running for an hour were just unacceptable and liable to degrade the CPU. To sustain a 5GHz OC we were well over 1.42v and Asus recommends triple radiator cooling for 1.35v (which we exceeded). We fired up the test station and let it Idle for 30 minutes and recorded the Idle reading. We then ran the test at each CPU speed for an hour then recorded the reading.
Between tests, we shut the machine down for 30 minutes then went to the next CPU speed. Each test at each speed was repeated 3 times then we report the average of the 3 readings. Every test was run using the pre-set profile in BIOS without any changes to maintain test consistency.
We used the provided thermal material on the CLC coolers for testing. The thermal material which we recommended replacing earlier well we can eat our words on that one. We did each test with Arctic Ceramique and the readings didn't change to use the included pre-applied paste if you want it's pretty good. Thermal measurements were taken using RealTemp not EVGA Flow Control but it should be noticed Flow Control mirrored RealTemps readings but we wanted a per core breakdown of temperature readings.