dual mcp655 uber pump - Page 2

I use an RD-20 for 3 GPU's, 2 mobo, 1 ram and a CPU block. I also have about 2 meters of tubing and a huge air to water heat exchanger (ROG-BOX linky in sig, go to the 5th page as there are two variants). The pump is a beast, but it needs to be bleed properly as it does make a racket with air bubbles.

I'll not lie, the pump does get hot, but it adds about 60w of heat which isn't a very big deal (dunno about the RD-30 or RD-30x).

It isn't a very aesthetically pleasing pump, and no top mods or covers are available. So hidding it wouldn't be a bad idea. Though you get to show off a monster pump in the rig that most either don't know about or fall in awe.
Pics (Click to show)
Edited by Rognin - 8/25/12 at 11:37am

You mean because of higher flow rate? Yes, but not by much; after about 1.8gpm, you start seeing a whole lot less gain per gpm.

The D5 stays cool because it has a steel plate that transfers its heat into the water; the MCP35X does not transfer most of its heat to the water, thus why it has an optional heat sink and needs good air flow across it.

The RD-30 uses about 5 watts more than two MCP35Xs at max, but the RD-30 acts like the D5 and puts most of the heat in the water. I doubt you would want to run either at 100% though.

Be sure to read through this if you're considering the Iwaki.
http://martinsliquidlab.i4memory.com/Iwaki-RD-30-Pump-Review.html

Without knowing models, I can't know what their restriction is. Most EK and Swifttech blocks are middle-of-the-pack though.

The only reason you would want to split a single pump's flow is to increase the flow in the blocks by decreasing the pressure. Depending on the water block restriction, this is sometimes true with using D5s. It is definitely true with the Iwaki.

If you want to split the flow of a single pump, though, you want both of the sub-loops to be about the same restriction. If you want a split flow using the Iwaki, I would use the following:
Keep in mind that the Iwaki is a 24v pump and has 18mm (~5/8") in/out barbs.

Where your rads are at in the loop won't matter a whole lot, especially with such high flow rate. People generally try to put the CPU immediately after the radiators for that extra degree that the pump heat might add: it's not a big deal (~40-50 watts). Puting rads in the middle may help by a degree or two since in one sub-loop you have 3 video cards in series: so the incoming water for the third already has 500 or so watts in it; if your flow rate will be really high (from the Iwaki), then it's really not a concern though.


The problem with puting your rads in series with your loop in parallel is that if the block flow is say 1.8gpm, then your rad flow would be 3.6. Stren has said that radiators cap out faster than water blocks with respect to flow rate (I'm trying to confirm this currently): meaning that you may hit the flow rate where you're actually reducing the performance of your radiators by pumping faster. If you keep the rads in parallel, then each set of rads has the same flow rate as the water blocks. You'll need to try to collect data on your water blocks and radiators to see if you can determine what their maximum flow rate can be before they start to lose heat transfer performance. You could always combine and split the flow at multiple places too in order to keep the rads in parallel.


I wouldn't be concerned with where the pump dumps its heat so long as you can properly cool it. If airflow is a problem where the pump will be, then the MCP35X may not be the best choice.

You think someone with this build would be concerned with airflow? I'd assume a double wide is involved here.

Along with dual PMP-450S pumps of course.


Notice what he said in the thread, not just the OP.

Edited by superericla - 8/26/12 at 11:11pm

Sorry, not seeing the airflow issue here. If he used a 360 rad in the front, it would create the airflow itself.

You can use martinswatercooling flow estimator to get an idea of the flow, or look at pq curves of each.

The RD30 consumes 45W of power, 2 mcp35xs consume 36W, and 2 d5s consume 46W. Putting in equivalent of 5 full cover gpus, a high flow cpu block (which I would use a high flow with your components) and multiple fittings, 6 rads, the RD30 will have about 1.3- 1.4 gpm flow, the 2xmcp35x about 1.2-1.3gpm, and 2 d5s about 0.9- 1.0 gpm. 2d5 strongs at 24v each is similar head/flow to single RD30, which again is barely more than 2x mcp35x, but 2d5 strongs will be the least power efficient. The relative flows of each will be more accurate than absolute for your loop.

Since wattage is within 10W, and variable amounts in water, the difference in power consumption is irrelevant. What matters is do you want to get slightly more flow (.1 to .2 gpm max, which translates into ~0.2C better temps) of RD30 or stick with more convenient 12v of 2 mcp35x. Also may wish to listen to noise from each, RD30 is a little louder, though depends on what speed you run all your fans at how much noise you will hear.
Edited by opt33 - 8/27/12 at 5:45am