Testbed Configuration and Testing Methodology
Three new VGA coolers were tested inside a system case configured as follows:
- Mainboard: DFI LANPARTY DK X48-T2RS (Intel X48), LGA 775, BIOS 08/29/2008
- CPU: Intel Core 2 Extreme QX9650 (3.0GHz, 1.25V, 2x6MB L2 cache, 4x333MHz FSB, Yorkfield, C0)
- Thermal interface: Arctic Silver 5 (for all coolers)
- Graphics card: HIS Radeon HD 4870 GDDR5 512MB / 256bit, 750/3600MHz
- 2 x 1024MB DDR2 Corsair Dominator TWIN2X2048-9136C5D (1142MHz / 5-5-5-18 / 2.1V);
- 2 x 1024MB DDR2 CSX DIABLO CSXO-XAC-1200-2GB-KIT (1200MHz / 5-5-5-16 / 2.4V).
- Disk subsystem: Western Digital VelociRaptor (SATA-II, 300GB storage capacity, 10000RPM, 16MB cache, NCQ)
- HDD silencer and cooler: Scythe Quiet Drive 3.5”
- Optical drive: Samsung SH-S183L SATA-II DVD RAM & DVD±R/RW & CD±RW
- System case: ASUS ASCOT 6AR2-B Black&Silver (ATX) with 120mm ~960RPM Scythe Slip Stream 120 fans for air intake and exhaust (the fans are installed on silicon spindles) and the same fan at ~960RPM on the side panel
- Control and monitoring panel: Zalman ZM-MFC2
- Power supply: Thermaltake Toughpower 1500W PSU (W0218) with a default 130mm fan
To increase the general power consumption of the testbed we overclocked our quad-core processor from the nominal frequency to 3.9GHz. The Vcore was increased to 1.55V. DDRII memory was working at 990MHz frequency and 2.10V voltage.
The testing programs were installed under Windows Vista Ultimate Edition x86 SP1. We used DirectX 9.0c libraries (from August) and Catalyst 8.9 drivers. Radeon HD 4870 was warmed up by two 10-run cycles of Firefly Forest benchmark from the synthetic 3DMark 2006 suite in 1920x1200 resolution with activated x16 anisotropic filtering but without FSAA. Our preliminary testing showed that the GPU heated up less with enabled full-screen antialiasing (which is quite logical as it is loaded less) and other components temperature didn’t change.
We monitored the graphics cards temperatures using RivaTuner v2.11 utility (created by Aleksey Nikolaichuk aka Unwinder). We performed at least two test cycles for each cooler and waited for 10-12 minutes for the temperature to stabilize between test cycles. The ambient temperature remained at ~25°C during the tests and is used as a starting point on our results charts. The tests were performed only in a closed system case. We didn’t check the coolers efficiency in an open testbed.
The noise level of each cooler was measured after 1AM in a closed room about 20sq.m big using CENTER-321 electronic noise meter. The measurements were taken at 3cm, 1m and 3m distance from the noise source. During the acoustics tests all three 120-mm case fans were slowed down to ~700 RPM and the 140-mm processor fan – to ~600 RPM. In this mode the background noise from the system case measured at 1m distance didn’t exceed ~32.8 dBA, and the loudest fan was the 130-mm fan of the system power supply. When the system was completely powered off, our noise meter detected 30.8 dBA (the lowest was 30 dBA), which may either be the internal noise of the device or the heartbeat of an excited tester :). The subjectively comfortable level is around 34~34.5 dBA.
We are going to test the cooling efficiency of our participants on a HIS Radeon HD 4870 graphics card:
This graphics card not only dissipated a lot of heat, but also allows monitoring the GPU temperature as well as temperatures of other components, such as voltage regulator or power elements. Since the latter get really warm during work (100°C+), we installed aluminum heatsinks on them and on the memory chips:
We removed two heatsinks from the two top memory chips above the GPU for Titan Twin Turbo. We removed a heatsink from the lowest memory chip for Arctic Cooling Accelero Twin Turbo. Besides a slight hardware modification of the graphics card, we also did some work on its BIOS. In fact, we didn’t perform any significant modifications. We lowered the GPU frequency to 250MHz and voltage to 1.1V in 2D mode (compared with the nominal frequency of 500MHz and voltage of 1.263V). As a result, we could reduce the current in 2D mode from 23 A to 11 A and lower the GPU temperature by 8-10°C. The graphics card worked at its nominal frequencies of 750MHz/3600MHz in 3D mode. You can download the modified BIOS here (7-zip file, 39.5 KB).
Besides the three coolers we have discussed in our today’s article and a reference ATI Radeon HD 4870 cooler (with the fans at 25, 35 and 100% rotation speed) we also added the results for Arctic Cooling Accelero S1 with a Turbo-module. We didn’t test the coolers during graphics card overclocking and later on you will see why.