As with most VGA coolers, you lose the use of at least one, if not two PCI slots with the 9, despite its low profile design. Fortunately most motherboards come with all the bells and whistles built in, and needing those extra slots usually isn't an issue.
Testing
I'll be using the 6600's built-in thermal probe to monitor the temperature of the GPU core. I'm using Aquamark3 runs to generate load temperatures, and I'll be comparing the VF900-Cu to both the card's original stock cooler as well as the recently reviewed Sytrin VF1 set on high speed. Reason being? The VF1 captured the top position among the various VGA coolers I've tested, and I had to see how the Zalman would stack up against the best.
I've only included the lowest and the highest speed setting for the 9 in these comparisons. There really isn't a lot of difference in the amount of noise that the fan puts out at the different speeds, and even when on the highest setting I didn't notice it over the rest of the fans in my case.
The first graph shows temperatures at stock 6600GT speed of 500/1000.
As you can see, both the VF1 and the 900 are fairly comparable and exhibit a substantial decrease in operating temperatures over the stock cooler, dropping more than 10° at idle and over 15° at load. The VF900-Cu with fan on highest setting barely edges out the VF1 similarly set to high speed.
At 559/1170 the stock cooler throws in the towel, reaching nearly 70° at full load. In comparison both the VF1 and VF900-Cu are keeping things under control, although the VF1 has pulled even with the 9 at this point.
And in a surprising twist, the roles are reversed when I overclock the card to its maximum of 582/1245. This represents the absolute highest stable overclock I have been able to achieve on this card, and both coolers are capable of allowing the card to reach it, however the Zalman suffers a 1° defeat at the hands of the Sytrin VF1.
As I was not able to increase the clock speeds of either the core or memory any further than previously attained, I did not run the usual gambit of 3DMark benchmarks.
Final remarks and conclusion just ahead.
