Test Setup

Test Platform
Processor Intel Core i5-6600K
Motherboard MSI B150A Gaming Pro (ATX)
Memory Crucial Ballistix Sport 8 GB 2400 MHz DDR4
Graphics Card XFX AMD Radeon R9 290X Double Dissipation Edition
Storage OCZ Vertex 460 120GB SSD
Cooling Corsair H75 Liquid CPU Cooler
Power Supply SilverStone Strider ST1000-P 1000 W Modular PSU
OS Windows 8.1 64-bit

The testing process I have adopted is as follows:

  • CPU Temperatures
    • Temps generated using a closed-loop liquid cooler (Corsair H75) in each enclosure
    • Temps measured at idle, load, and "stress". Load temps created using the video transcoder benchmark x264, with stress results from prime95 (large FFT torture test)
    • Temperatures defined as the hottest core as recorded at the 5-minute mark using HWMonitor software
    • A custom fan profile is used to provide linear results for both liquid and air cooling
  • GPU Temperatures
    • Load temps (XFX Radeon R9 290X) created using the Unigine Valley benchmark (Extreme HD preset), with the highest temp recorded after two successive benchmark runs
    • A custom fan profile is used to provide linear results
  • Noise Levels
    • Measured using a digital sound pressure meter positioned 24 inches from the front of the enclosure (system fully assembled with side panels in place)
    • A fixed speed for load noise will be used for consistency, based on observed average max fan speeds under load for both CPU and GPU

I employed the ASUS motherboard software MSI motherboard settings to create a custom fan profile for the CPU, and both the Deepcool Gabriel air cooler and Corsair H75 liquid cooler used this profile for their PWM fans.

The XFX R9 290X DD graphics card was also setup with a custom fan profile using MSI's Afterburner software, with a linear rise in fan speed beginning at 20%.

The reason for the custom fan profiles is simple: default fan profiles flatten out during certain temperature ranges, which results in uneven results between tests and enclosures as a given temp will often fall below the threshold to increase fan speed. In my quest to provide consistent results I didn't like the position where a single degree often determined whether the cooler was placed into a louder, but more effective, state.

Add to this the complication of testing without complete control over ambient temperature and you can understand why noise testing was done at a fixed fan speed. Ambient temps can produce very misleading results with regard to fan noise as a colder room requires less fan speed to keep the system cool, with the opposite true for warm rooms. The adjusted delta temps are the same in either case, but for a neutral look at noise output a fixed speed was chosen for both CPU and GPU noise testing.

Next we will look at temperature and noise results with the Air 740.

« PreviousNext »