Power, Temperature, And Noise Benchmarks
Now that we know the relative performance of these cards, let’s look at their other statistics:

We'll start with power draw. Note that we begin the graph here at 200 W, which is close to what this system draws with a power-sipping GeForce 8400 GS running at idle. These results paint an interesting picture.
Starting with Gigabyte's low-end GeForce 9600 GT solutions, we see a uniform result from these cards with the lower-clocked passive version using a little less power. This makes sense as the passive GV-N96TSL-1GI has a slightly lower clock speed than the actively cooled GV-N96TZL-1GI.
The GeForce GTS 250 solutions show us a marked difference in idle power use. In fact, the Asus EBGTS250 Dark Knight uses less power at idle than even the GeForce 9600 GT cards do. This is because the Dark Knight has a very aggressive power-saving mode and only runs the card at 300 MHz core/100 MHz memory in 2D mode, according to the GPU-z utility. Once loaded in a 3D game or overclocked, both of these GeForce GTS 250 cards demonstrate virtually identical power draws.
The Asus and MSI GeForce GTX 260 cards show similar idle power draw, but the Asus Matrix shows a much higher load power draw. This is likely due to a number of factors: the MSI’s 10-phase power management is likely more efficient than the Matrix’s and the Matrix card allowed us to over-volt to a higher level than with the Lightning mode. We should note that it would have been possible to under-clock and under-volt the 2D profile of the Matrix card to achieve low-power results similar to the Dark Knight's power draw at idle, while the MSI Lightning has no provisions for under-clocking and under-volting as far as we could tell.

The GPU temperature graph doesn’t show us much except that all of these cards run their GPUs in a similar temperature range. All of the coolers here are doing good jobs, with loaded overclocked temperatures remaining under 80 degrees Celsius. The Asus ENGTS250 Dark Knight and Gigabyte GV-N96TSL-1GI get props for keeping loaded temperatures notably less than 70 degrees Celsius. Especially impressive is the Gigabyte GV-N96TSL-1GI, since it is a passively-cooled card.

We started the noise graph at 45 decibels. This is because the Gigabyte passively-cooled GV-N96TSL-1GI produces no noise at all with its passive cooler and we measured 47.5 decibels of noise caused by the system itself.
While the temperatures are uniform, the amount of noise produced is not. The Zotac is hard to miss with its loud reference cooler at 56 decibels. Every other card managed to keep things reasonably quiet below 53 decibels.
Would you run the first passive card with negative air pressure?
@mi1ez. Not too sure how yours is set up. But for most living rooms I'm sure the average Ikea or Argos T.V. cabinet wouldn't have adequate airflow for that. Getting waffted with heat in the summer wouldn't be my choice.
I don't have a passive card. I was just curious really given what they said about the card drawing cool air from outside the case.
The MSI overclocking panel appears to claim the fan is at 600,000rpm!
I wonder if there's an option on the Asus ENGTS250 Dark Knight card that allows you to run the cooler passive when idle? Seems like the heatsink could pull that off easily, doesn't it? It would be nice to have silent card when in 2D
so the question is can you control RPM of the fan and if so, can you put it to 0% when idle?
Well I just bought the Asus ENGTS250 card based on this review, and not only is the memory clocked slower at 2000Mhz DDR, the card does not slow it's clocks in 2D mode leading to substantially higher power consumption at idle.
I'm not impressed.