With the watercooling the highest GPU load temperature I've seen is 53c.
Compared to a good watercooling setup which would be 3-5c above ambient temperatures that seems quite high.
It's nothing compared to the stock cooler though.
With the stock cooling my 8800GTS would reach 80c under load very easily.
The fan settings for the 8800GTS per Rivatuner are:
Duty Cycle Min 25%
Duty Cycle Max 100%
T Min 72c
T Range 32C
T Operating 82c
T Low Limit 72c
T High Limit 92c
What that seems to mean in practice is that the graphics card fan stays at 60% speed until the card reaches 82c (You can see that in a few of the rivatuner temperature pics on the first page of the thread). After that it starts to ramp up.
3D Mark 2006 Stress Testing
You've probably seen the SPCR Aerocase Condor review where they discuss loading graphics cards using synthetic benchmarks? In particular where looping 3D Mark 2005 (as seen in the Madshrimps review) isn't sufficent to get the card fully loaded.
Room temperature today is 24c approx.
First of all I left the card looping the full 3D Mark 2006 test for an hour. (all default settings 1280x1024 no AA)
At the end of the hour the GPU temperature had risen to 51c.
This was only for about 50% of the time however. Whilst doing the CPU benchmarks the GPU temperature fell to 47c.
After leaving it to cool back down for a few hours I then left it looping the first 3D Mark 2006 Shader Model 3 benchmark (the airship and giant sea dragon one) for an hour. This is at default 1280x1024 no AA.
After roughly 45 minutes of looping the SM 3.0 benchmark the GPU temperature had reached 53c. It then stayed at that level for the next 15 minutes (11 run throughs) until I stopped the test.
Room temperature had increased from 24c to 24.4c (using a digital thermometer). As I'm writing this the room temperature is gradually increasing as the reservoir cools (room temp is 24.6c 30min after stopping the benchmark, the GPU temperature has dropped to 44c)
The temperature of the silver aluminium reservoir felt quite warm to the touch at the end of the test. It wasn't hot enough to burn you but you wouldn't want to keep your hand next to it for a long time either. The black vanes weren't as hot.
I didn't have anything to test the water temperatures with unfortunately. I suspect it was quite hot though.
You really wouldn't want to put too many hot components in one water loop using this.
Have a look at these two reviews where they use the reserator 2 with multiple components:
http://www.xbitlabs.com/articles/cooler ... html#sect0
http://www.pcper.com/article.php?aid=34 ... pert&pid=9
For cooling a single graphics card it seems to be good enough though.
There was also a noticeable "ticking" noise coming from the reserator however. Now that it's cooled down a bit this has stopped again.
It's difficult to describe the noise: think hot metal expanding perhaps.
If you want me to upload the rivatuner temperature monitoring logs of the two tests let me know as they're only 640kb each in size.