What did this guy do wrong?
 

http://forums.silentpcreview.com/vie...=208493#208493

it seems somewhat impossible for 2 x GPU, 1 x CPU to be that hot. You think he done something wrong due to the way he loop his pipe, or you think zalman just can't cut it?

With all those blocks on a 1046 he will have a very low flow. Don't quite know the zalman blocks internal design is, but it is quite possible that they are a tad restrictive too.
Couple that with fluid xp and a small passively cooled surface area, y-splitters et. al. I'm not surprised.

Its probably not the flowrate. even with a super-low flowrate the results can be quite ok. you won't get the best performance with an outdated pump, but if it is still moving the water it won't let your pc get so hot.

Because the idletemp is quite ok it can't be the pump because it has a constant flowrate. The idletemps would be worse if it was only the pump.
I think its obvious that it is the Radiator and maybe the cpublock too. The Reserator simply has not enough power to cool the cpu, northbridge and two graphics cards. A single black ice pro with an pabst supersilentfan would be more powerful than this passive Radiator. Second thing is the waterblock. Zalman uses a really old blockdesign similar to an old Innovatek Rev.3 but a little smaller. There is at least 5mm of copper between the water and the DIE, which isolates and is'nt good for the heat transfer.
The Radiator and the Waterblock are the most important components in a rig like this. The pump can theoretically give some extradegrees, but when the radiator is so hot like here a larger and better pump would even perform worse because it produces more heat that the Radiator can't handle. A block like the zalman is a laminar flow block which doesn't scale with added flow. A watercooling rig is alwas as good as the worst part of it...


EDIT: I see that this guy used a parallel way of flow to otimise flow. In most cases a parallel waterflow is wors than a serial one because of different pressures. Maybe his cpu doesn't get enough waterflow because most water is going trough the graphics card cooler. it can be that he gets this low idletemps because its an A64 with cool n quiet.

Parallel loops without flow tuning between the loops. If you've got pump head and flow to spare you can add a variable restriction to each parallel loop to tune the flow and make it work. As it is I think his CPU/NB loop is getting a trickle as it is likely the more restrictive parallel loop and he's already splitting the flow of a 1046 in a restrictive system. He's using long small tubes, a 1046, and FluidXP with what look like restrictive blocks.

His planned solution of adding another radiator will likely make things at best no better. If he plumbed it in series and upgraded the pump to something that could push FluidXp with some more head and flow then we'd then be able to see the performance of the Zalman Reserator. As the loop stands now there are more likely candidates for the weakest link.

Just my opinion, ymmv.

Saw your note about the plumbing, so I take it when you wrote this you were under the impression that the loop was serial. Anyway, I doubt your logic. Think about a variable speed CPU fan. At its lowest speed at idle CPU the temperature is fine. Why then at load as temperatures climb do we increase the fan speed? By your logic a constant flowrate is ok, so we should never have to increase the fan speed. With water cooling we generally don't vary the flow, so we've got to have enough to deal with the load.

The key telling point was his CPU block was too hot to touch and his FluidXP was warm. These are very innacurate measurements, but they tell a story. Based on his untuned parallel loops and obvious flow problems as a result, I am not willing to pan the Zalman yet.

Reading the updates in the thread linked from the first post apparently some bleeding problems were causing some of the high temps. Even with the plumbing issues the temps have gotten a lot better in the most recent updates.

If the coolant is hot then its the rad.
If the coolant is ok but the block is hot, its the block.

Trickle of cooling flow with G4 and BA radiator cost Cathar a mere 4 degrees (by memory).

Parallel paths indeed rarely help. You want every drop going through the CPU block and radiator.

And if the block is cool, it's the mount.