|
Review Arctic-Cooling Silencer NV5 English |
|
martes, 20 de marzo de 2007 |
|
Página 7 de 10
As we can clearly see in the heat sink pics, the fins don't cover the copper base completely but the center of the heat sink, this means that the heat coming from the GPU finds the way out straight upward, but the heat generated by the memories have to cover the distance that separates them from the fines, hence the efficiency is not good enough in this part of the video card. It is evident that the behavior cannot be the same for the memory chips, since the chips and the fins are not located at the same distance.
This is one of the things that we want to find out: how effective is the NV Silencer 5 to each of the components. To do this, we have no choice but to use independent temperature sensors on each element, as the video card cannot provide us with this information on its own.
For that purpose, we use a set of nine thermal probes that are installed in the back part of the video card, bellow each of the elements to be measured -in this case, eight memories and the GPU.
We already know that the reading will draw an error, since the heat to get to those probes will have to pass through the entire PCB of the video card. Nevertheless, we decided to carry out the measurement with the available means. But before we continue, we want to inform the reader that this is just for an informational purpose and should not be considered as something strict.
The set of thermistors
Bellow are the probes placed in position and kept by OC Squadron's backplate which covers all the area and gives a certain degree of insulation to the sensors regarding the room temperature.
To conduct these measurements, we used the well-known 3Dmark 05 Free Version, as it is a test that can be perfectly run by anyone without the temporary circumstances that may arise during the normal course of a game.
This time, unlike the previous BF2 test, we waited one minute from the first measurement until the benchmark began, and we kept a record every thirty seconds.
With all the options activated, we let the 3Dmark 05 Free Version finish, and waited until the temperature dropped to normal levels to observe the evolution of the entire process, which took fourteen minutes. We first compared the temperature obtained from the video card drivers to the one obtained from the probe located under the GPU on the other side of the PCB, just to get an approach to the difference given by the two readings.
In this chart, it is possible to see in green colors the readings performed with the video card running on default mode. The highest reading corresponds to the driver and the lowest one to that of the probe.
In blue shades are the same readings, but with the video card at 400/1100. A similar behavior has been noticed between the driver and the probe.
By way of comparison, the default and overclock values along with the stock cooler were added in red colors.
The difference, with a 10°C average, turned out to be greater than expected, making the measurement impossible to carry out as we planned, since it invalidates any estimate we could make. In any case, a set of measurements was taken.
What it isn't clearly shown in the charts is the thermal inertia that was noticeable in the temperature measured by the probes. It was evident that the heat took sometime to reach and leave the PCB's opposite side, this was the reason why we always had a delay in this matter.
|
