Cooling Efficiency Tests
Diamond Radeon HD 4890 1 GB XOC
Diamond Radeon HD 4890 1 GB XOC graphics card works at higher frequencies of 925/4200 MHz, so we didn’t overclock it any more:
Let’s check out the obtained results:
Unfortunately, the cooling efficiency of Scythe Setsugen leaves much to be desired. In quiet mode at 800 RPM even with a Thermalright VRM2 heatsink, the new cooler can’t provide sufficient cooling for the PCB voltage regulator components, so it is hard to imagine what could have happened if we had used the bundled aluminum heatsinks instead of the Thermalright VRM2.It is interesting that the increase in the rotation speed of the Scythe fan doesn’t help improve the situation: the temperatures drop just a little bit. In the end Setsugen cooler with its fan working at the maximum speed turned out only as good as the reference cooler of the Radeon HD 4890 in its automatic speed mode. Not that impressive, I have to say.
The efficiency of our today’s second testing participant, Arctic Cooling Accelero Twin Turbo Pro, is on the contrary very high. This cooler remained ahead of the graphics card reference cooler in all speed modes of its two fans. And as we know, the reference cooler of this graphics card is pretty efficient. Even in quiet mode at 1080 RPM the new cooler can cope with an overclocked Radeon HD 4890 burned to the utmost by the monstrous FurMark. This result is truly impressive. And the advantage of the new Arctic Cooling product over the Scythe cooler sometimes reaches 30°C!
ATI Radeon HD 5850 1 GB
For the tests with Radeon HD 5850 we took a reference ATI graphics card with nominal frequencies. To make it harder on the tested coolers we overclocked the card from 725/4000 MHz to 900/4800 MHz without changing its voltage settings:
Before getting to the primary tests, we decided to check the efficiency of VRM heatsinks from Arctic Cooling bundle. These pseudo-heatsinks were installed using default sticky tape and compared against Thermalright VRM4 paired with Scythe Setsugen cooler at 800 RPM. The load was created using 3DMark 2006 test. The results turned out quite predictable:
Even under relatively moderate load created by the Firefly test from the 3DMark 2006 testing suite, the temperature of the PCB VRM components under the default heatsinks from the Twin Turbo Pro bundle got dangerously close to 110°C, so you can take a good guess what could have happened during FurMark. At the same time, Thermalright VRM4 heatsink provided more than 30°C better thermal conditions for the VRM components than the small Arctic Cooling heatsinks. Of course, during the primary benchmarking session we tested our coolers under two types of operational load using only Thermalright VRM4 for the voltage regulator components (I have to remind you that Twin Turbo Pro is incompatible with Radeon HD 5850). Here are the obtained results:
I have to point out right away a significant difference between the load created by two fireflies in 3DMark 2006 and the load in FurMark. We didn’t see such humongous temperature difference on Radeon HD 4890 even with the weak coolers. I also have to say that Scythe Setsugen could compete against the reference Radeon HD 5850 cooler. I really wish we could compare the two today’s leaders against one another.