Testing Configuration
We used the following hardware configuration to test the GFXChilla:
- Intel Core 2 Quad Q6600 @ Default Frequency
- Asus P5K-E Motherboard (P35 Chipset)
- 2x1024MB Buffalo Firestix PC2-6400 Memory
- Sapphire ATI Radeon X850XT, 256MB
- Maxtor Diamondmax 10 SATA Hard Drive
- Pioneer DVR-212D SATA Optical Drive
- Generic Mid-Tower case with one Noctua NF-P12 120mm exhaust fan.
- Corsair TX750 Power Supply
Ambient temperature was maintained at 19°C for all tests and was not permitted to deviate more than +/- 0.5°.
For testing, I decided to use only a single 120mm exhaust fan and kept our reference mid-tower case closed. I wanted to see how the case temperature changes when using a down-draft versus exhaust style GPU fan.
I placed an external sensor in the case, close to the lowest PCI slot to measure the case temperature during all tests. GPU load was simulated using the ATITool application and it’s “Fuzzy Dice” rendering. In my experience, this puts a very heavy load on GPUs. GPU and PCB temperature was monitored using ATITool as well. ATITool was also utilized to control the fan speed of the OEM ATI cooler.
To test GPU stability in the short amount of time I had, I relied on ATITool’s fuzzy dice rendering and looked for hangs and artifacting to judge stability. The card had to be able to render at any given frequencies for at least twenty minutes without crashes or artifacting. This is definitely not a 100% indication of GPU stability, but is good enough for our comparative purposes.
Test Results
I had no clearance issues while installing the X850 with the GFXChilla mounted. I utilized one of the motherboard fan headers. I was immediately amazed at how quiet the GFXChilla is compared to the OEM cooler. It is not silent, but can barely be heard above the CPU and case exhaust fans.
As can be seen, the GFXChilla runs circles around the OEM cooler. Not just by a small margin either, we’re talking an almost 40°C delta. Impressive considering that it managed this at lower noise levels too. Even with the leaf-blower blaring at 100%, the GFXChilla still manages a 21°C lead. As you can see, the case temperature increased slightly when comparing the OEM ATI fan at full speed and the GFXChilla. The OEM fan flows quite a bit of air out of the case. Even with a single case fan, the GFXChilla maintained a case temperature of only 1°C above the ATI fan. As long as there is healthy ventilation in the case, there should be no concerns about giving up the exhaust properties of the OEM fan.
Overclocking ability is almost always improved with lower temperatures. I had always assumed that the X850 GPU had very little headroom as it was ATI’s last 130nm GPU and really pushed the manufacturing technology to its limits. Although not a huge gain from the 520MHz default frequency, the low temperatures allowed 594MHz. Only 561MHz was possible using the OEM cooler and default fan speed. I had always thought the memory had reached its limits at 595MHz regardless of cooling—I was wrong. With the OEM heatsink contacting the ICs, its no wonder the memory hit a brick wall at those high GPU temperatures. A 30MHz increase on the memory is quite impressive.
The above graph depicts each cooling configuration at its maximum overclock. You can see that the already toasty OEM cooler gave up another 5°C. Amazingly, the fan remained at less than 10% speed with the GPU at 85°C. The GFXChilla only gave up 2°C at its maximum overclock.
Articles RSS