That's the way you can use it, given Zalman's commitment to that goal, but after their first model -- the tops of the Alum heatsinks for the B and D-HP models are slotted and for certain graphics cards, Zalman either recommends or insists that you should use their OP-1 80mm fan. The fan, like their 92mm jobs, runs about 2,950 rpm, or using their fan splitter that changes the fan from 12V to 7V, it also runs in "Silent" mode. But "Normal" mode is nothing that I can hear.
You could, if you wish, use any kind of fan or jerry-rig a bracket fan above the thing, but if the graphics card requires an OP-1, I still think it can bring in temps a degree or two below the Silencer. But in the end, it WOULD depend on how hot the GPU runs . . . so your point is well taken.
For purposes of comparison, I'm using a dual-heatpipe ZM80-D-HP:
ZM80-D-HP
with an ASUS V9980 FX5950 Ultra 256 AGP card. Although this is last year's flagship model for the nVidia line, it still requires the additional power of a four-pin molex plug. Temperatures I've measured with a thermal sensor right at the GPU core registered as high as 128F with a GeForce4 Ti4600 card, and I can't imagine the FX5950 -- with GPU @ 475Mhz and memory @ 950Mhz -- running cooler than the GF4 card. For the Ti4600, the cooler without a fan dropped the GPU temp at room ambient of 75F to about 105F, measured with the same taped-on sensor at the core. With the FX5950, the GPU has its own thermal monitor built in, and the temperature value is available through Display Properties, Advanced, FX5950 tab. At about the same room ambient, it registers something between 38C and 40C, or 104F.
This is a competitive business, and I see ThermalRight and others producing products under similar concepts. I have a Silencer now on the Ti4600 -- a problem of space and conflict between CPU and GPU heatsinks. I have the ZM80D-HP on my FX5950.
I haven't determined a way to measure this, but I have it from my electronics-tech friend back in Virginia that my suspicion that the mobo circuit traces conduct heat as well as signals -- is correct. I DO believe that cooling high-temp components without needless electrical usage or additional fans and fan noise will allow for more control of other component temperatures. I haven't been able to TOTALLY confirm it, but measuring the temps before and after VGA and other cooling seems to substantiate it. I'd like to find out what other people show as chipset temps before and after their VGA cards were cooled with solutions more effective than the stock HSF.