Hello bob4933,
Thanks for your input. Never heared about cavitation O_O
Could you please tell me how this occures and why having a pump under the reservoir can prevent from it ?
Cheerz
EDIT : nvm I g****d it
gah, haha how deep do you want?
I will give you the sesame street version. Basically, when you mount the pump above the reservoir, you are limiting whats called "pump head". The output of a centrifugal pump is based entirely on pump head. If you have low pump head, you aren't putting out much pressure or volume. On top of this, with a low suction pressure, the coolant will actually drop pressure and create bubbles on the impeller of the pump (the part that spins to move the coolant), and over time this will DESTROY your pump. Not "it might happen", it WILL happen. The question is simply "how long?".
Not to mention, if you put it lower, your pump will operate more efficiently, which means higher flow, which means lower temps (theoretically). Its truly in your best interest to get that system layed out a little better. It doesn't have to be completely below the reservoir, but the lower the better.
Bob is right about the cavitation and you're seeing foam or bubbles in the loop because the pump is sucking in air. This is because you don't have enough fluid in the loop. The pump being in the second reservoir which is at a high point, when running doesn't have enough liquid in it is sucking the air. Make sure you have enough fluid in the system and problem solved.
I want to clear this up...because there's some stuff in here that's pretty wrong.
Firstly, when your pump is cavitating, that just means it's sucking air, no more. Water is incompressible, so you won't be able to make bubbles out if it without there being air in the loop already...shy of supercavitation at least. This is bad because the pumps that we use for WC are largely based on ceramic bearings which rely on the fluid they are pumping to cool and lubricate the bearing. Run dry, the pump can fail quite quickly.
Secondly, the location of the pump in a closed loop generally won't matter...certainly not in the case of the pumps that we use in any common computer case. You're talking about the pump's head needing to overcome restriction because it's at a high point, but this is incorrect. Any fluid that it fights gravity to move up eventually gets a gravity assist back down, this is the nature of the closed loop.
If you have your res immediately before your pump (which it should be, and in this case is an inherent feature of the res/pump), it does not matter where in the loop your pump/res is located. Air naturally tries to work its way to high points (flow in the loop can move it from these points, it won't just sit at high points), so if you have your pump near a high point and the res is not immediately before it, then you can run into issues of sucking air through the pump...which, as I mentioned earlier, is bad.
Hopefully that clears some stuff up. Feel free to have reasoned discussion on this.
Moving to your actual issue, I'm relatively certain that you've got it completely misdiagnosed. I believe your issue is two-fold. Primarily your issue is that your loop has a lot of air trapped in it. When your pump kicks on, your res levels probably change pretty significantly as the air is compressed. You don't see the air when it's running because it is pushed into your rad. When you reconfigured your loop, you did a poor job of bleeding it, and with your rad setup with barbs down, it's a big air trap (you can get around this, barbs down is fine, you just have to know how to work around it). The second issue, which I think is worsening the first, is that by bracketing the rad with reses (which is to say running res->rad->res), I suspect there is a strongly averse effect on flow. On the WC board (why isn't this posted there?) we had a long discussion about this the other day, and arrived at no real conclusion, but I'm of the mindset that reses should feed into each other...essentially creating one giant res.
Read up on how to bleed your loop, our sticky in the WC section is great, and try to get that air out of your loop and see if anything changes. Also, it may be worth reconfiguring your loop again, and removing the new res, or changing the loop order if you like the aesthetic of it.