[build log] Water & Wood

[theseeker]

Mia,
Very ingenious!

[josk1]

love the idea !!
are you gonna get your desk a nice color finish, or will it stay "woody" ?

[miahallen]

I think it'll stay "woody"

[miahallen]

Water cooling components are now being flushed....here is a brief video showing the process:

[dengyong]

Anything coming from the rads will get stuck in the cpu block's pins. There may not be anything but I'd check to be sure.

[miahallen]

Oh, thanks dengyong....duh

[miahallen]

Yesterday I posted my video about flushing the loop....but I neglected to post the photos taken leading up to that point....I've also updated the OP with more parts pictures.

EK Full cover blocks:




3 way parallel bridge and bridge links:


Bridge links installed:


Bridge installed:








Well, progress continues...I worked on getting the rad wall setup last night.

I started by taping together some rad templates and placing them on the rad wall (the inner right wall of the PC opening in the desk).


One of my goals for this build (being a daily use rig) will be low noise, to mechanically decouple the fans from the desk, I'm using a 1mm layer of cork under the fans.


After getting the template placed, I drilled holes for all the fan screws, and at the center of each fan hole, then removed the template. Then I needed to cut the fans holes in the cork, I marked the holes with a compass (using the aforementioned holes as the center for my compass) and began to cut the cork by hand with a box cutter (razor blade).


Cork layer all cut.


Next, I removed the cork, and proceeded to cut the fan holes in the wood.


Next, I ran into a bit of a hurdle....the fan screws that come with the Swiftech rads are 30mm long, and I was using 25mm fans....which would work fine under normal circumstances. However, we know that ducted fans are more efficient, and I had a 16mm wood wall that I was mounting on, my plan was to kill two birds with one stone by mounting the fans on the outside of the wall and use the wood itself as a 16mm duct to the radiator. The problem was the 30mm screws were now 16mm too short. The Japanese HW stores carry plenty of standard M3 and M4 screws/bolts, but Swiftech uses the less common M3.5 size, which they do not stock. After wasting 3 hours going to various HW stores all over town, the longest screws I could find were 20mm....not 45mm like I needed. So, I brainstormed a bit, I thought about tapping the holes. But knowing me, I'd screw one up and get mad....on top of the fact that I'd have to by a tap and more expensive screws. I thought a simpler solution would be to open the corners a bit to allow the fans to be mounted from the inner flange, instead of the outer. The original fan corner:


Now, with the hole on the outer flange opened a bit, and the Swiftech provided screw for size reference:


This would give me a depth of 16mm for the wall, plus 5mm for the inner fan flange, plus 1mm for the cork...or 22mm total. 3mm less than the original fan depth. Since my plan was to decouple the fans mechanically anyhow, this left a little buffer room for a neoprene pad to dampen vibration. The final assembly looks like this:


Shaving from drilling out 48 fan flanges


Poking the screws through the drilled out flange, through a washer, punching through a 5mm thick peice of neoprene, and through the lower hole proved to be a PITA, so I sat down for some quality time with my wife and watched an episode of "The Event" and she helped me with the process


All finished:


Next, I just needed to mount everything on the wall:






Well, that's it for now...thanks for all the comments.

[defect9]

Damn, that's some kick-ass problem solving

[miahallen]

Thanks defect

[dengyong]

I like those rads with the U bend connector. Any plans for an air filter ?

[miahallen]

Yes, it's really dusty here 300 days of the year (too much orange dust from China ) so an air filter is a must. That will be a few more steps down the line

Here are a couple pics of the desk reassembled, the rig is on air for basic ops testing.

[miahallen]

Next, I wanted to setup the loop and continue flushing & leak testing. First I drilled the top of the desk for the fill port:


This is the underside showing the hole near the rear of the desk:


The mounting thread on the fill port was not deep enough for the depth of the desktop, so I had to increase the diameter of the hole on the bottom side to effectively secure the fill port. I used a sanding wheel on my dremel to accomplish this


After that the fill port went in without issues:


The fill port feeds straight into the top of the res....but I'll probably shorten the tube about 1/2" to get the res to sit up closer to vertical. The res/pump will hang here with support only from the tubes. Does anyone see a problem with that?


I know this is non-standard and a lot of people will question my decision, but I'm setting up the loop with all four blocks (3xGPU & 1xCPU) in parallel. I'm doing this for routing simplicity and ease of use...and I'll determine if performance is sufficient once I get up and running. THis is what I'm referring to:




Here are some shots of the loop in its test configuration:




And here is a quick video which gives you an idea of the flow:


I think this single pump will be fine. And to keep noise low, I'll be having the mobo CPU fan header control the speed of the pump automatically with PWM control based on CPU temp. The fans will all be controlled manually with the custom PWM controller xXSebaSXx and Bobnova built for me

[Vicey]

Having it just dangling there by the tubing alone makes me nervous. Having said that, we all know how hard it can be to pull tubing off a connector sometimes and almost impossible if the compression fitting is screwed on so I doubt it would fall off but still makes me nervous just looking at it dangling like that :P

[defect9]

I'd probably throw some eye hooks into the desk and run a bungee line or two under the pump, but honestly I doubt it's a necessity.

I will say, running all the blocks in parallel is visually rather striking.

[dengyong]

I don't think you'll have enough flow (if any) through the cpu block in that configuration. You could put one piece of clear tubing on the cpu block to see what the bubbles do.

[miahallen]

It was discussed in advance here:
http://www.xtremesystems.org/forums/...d.php?t=267745

My metric is cooling, not flow...if I can keep reasonable temps with my OC I'll be happy.

[theseeker]

Mia,
I am really enjoying your build and your problem solving skills. Everything looks great!

[dengyong]

It was discussed in advance here:
http://www.xtremesystems.org/forums/...d.php?t=267745

My metric is cooling, not flow...if I can keep reasonable temps with my OC I'll be happy.

In physics, the path of least resistance is always taken by objects moving through a system.
The three gpu blocks in parallel will effectively have no restriction to your volumetric flow rate so any restriction placed after them should get zero flow.
Martin said it would take massive pumping power because you would need to increase flow through the parallel gpu blocks until their resistance equals that of the cpu block to get flow through it.

[miahallen]

Mia,
I am really enjoying your build and your problem solving skills. Everything looks great!

Thanks theseeker...it's been fun so far

In physics, the path of least resistance is always taken by objects moving through a system.
The three gpu blocks in parallel will effectively have no restriction to your volumetric flow rate so any restriction placed after them should get zero flow.
Martin said it would take massive pumping power because you would need to increase flow through the parallel gpu blocks until their resistance equals that of the cpu block to get flow through it.

Thanks dengyoung...I'm not an expert, but I have a pretty good knack for this stuff. But in the end, we'll know for sure

The way I see it, the CPU block is not "after" the GPU blocks....they are all placed equally in the loop (parallel by definition). Therefore, all three GPU blocks will have equal flow, and the CPU block will have more or less flow (than any ONE of the GPU blocks) based on their relative restriction. My hope is that the CPU block is less restrictive than the GPU blocks, which will afford it a flow advantage over each of them.

I disagree with Martin....if my loop were setup all in a series, then his statement would make more sense to me. The fact that all the blocks are arranged in parallel would mean that the loop's overall resistance is far less than if they were in series.

You read Vapor's reply at the end, right?

[dengyong]

Thanks dengyong...I'm not an expert, but I have a pretty good knack for this stuff. But in the end, we'll know for sure

The way I see it, the CPU block is not "after" the GPU blocks....they are all placed equally in the loop (parallel by definition). Therefore, all three GPU blocks will have equal flow, and the CPU block will have more or less flow (than any ONE of the GPU blocks) based on their relative restriction. My hope is that the CPU block is less restrictive than the GPU blocks, which will afford it a flow advantage over each of them.

I disagree with Martin....if my loop were setup all in a series, then his statement would make more sense to me. The fact that all the blocks are arranged in parallel would mean that the loop's overall resistance is far less than if they were in series.

You read Vapor's reply at the end, right?

loop.jpg

The resistance being so low on the gpu blocks is the problem. They need resistance equal to the cpu block. You'll need around 1gpm through the cpu block to have good results.

You can do a quick test by using clear tubing and introduce some bubbles into your loop by putting a straw into your res near the intake and observe the flow.

[defect9]

as all things, blocks, tubing, etc, introduce some measure of reduction in flow through friction (pressure drop, basically), the more flow you get through the blocks, the more backpressure you effectively get. obviously backpressure doesnt mean the water starts flowing in the opposite direction, but it introduces an entrance pressure to that block. it's why 3 equal blocks in parallel equal out. it's the same reason that the CPU block will get flow. maybe not enough flow, but it will get flow.

however, aesthetically, that part of the loop is in.

[miahallen]

loop.jpg
The resistance being so low on the gpu blocks is the problem. They need resistance equal to the cpu block. You'll need around 1gpm through the cpu block to have good results.

You can do a quick test by using clear tubing and introduce some bubbles into your loop by putting a straw into your res near the intake and observe the flow.

Maybe drawing the picture a little differently will help you see what I'm talking about....



The CPU is not competing with all three blocks....but all four blocks are competing with each other for flow. Naturally, the block with the least resistance will get the most flow....since all the GPUs will have equal resistance and equal flow, the question is simply: "Is the Apogee XT Rev2 more or less restrictive than an EK-FC580?" If the answer is "less restrictive"...then the CPU will get more flow than each GPU block, and the temps will be best for the CPU (the goal)...if the answer is "more restrictive", then my CPU temps will suffer.

If the CPU block has the same resistance as each GPU block, than every block will get the same amount of flow, and in that case my loop will need to perform at 4GPM in order to achieve 1GPM over the CPU block. I'm not sure I'll reach that goal, but like I said, my goal is based on temps/performance, not GPM

[dengyong]

The CPU is not competing with all three blocks....but all four blocks are competing with each other for flow. Naturally, the block with the least resistance will get the most flow....since all the GPUs will have equal resistance and equal flow, the question is simply: "Is the Apogee XT Rev2 more or less restrictive than an EK-FC580?" If the answer is "less restrictive"...then the CPU will get more flow than each GPU block, and the temps will be best for the CPU (the goal)...if the answer is "more restrictive", then my CPU temps will suffer.

If the CPU block has the same resistance as each GPU block, than every block will get the same amount of flow, and in that case my loop will need to perform at 4GPM in order to achieve 1GPM over the CPU block. I'm not sure I'll reach that goal, but like I said, my goal is based on temps/performance, not GPM

I don't want you to think I'm being an ass, but it doesn't work like that...
The water inside the tubing needs some pressure to move it through the entire loop.
If you put a short cut in the middle of the loop with no resistance to the flow (in this case the 3 gpu blocks) then there is no pressure left to push water the rest of the way round... it simply completes it's circuit via the path of least resistance. If the first two gpu blocks were to flow freely enough you wouldn't get any flow through the third or the cpu.
Think of it as short circuiting an electrical circuit.
I'll leave you to your testing now.

[meanmoe]

You're going to starve your cpu block. It's a relatively restrictive block and will cause a bad flow imbalance. Like dengyong, I'd recommend the cpu block in series and then gpu blocks however you want them. At very least, the way you have it is not optimal.

Otherwise, very nice build. I really like how you mounted the fans. I'm out of room and I've been debating a desk or wall mount. There have been a few recently, but your approach is more my taste. Looking forward to the finished product.

[miahallen]

I don't want you to think I'm being an ass, but it doesn't work like that...

Not at all
We just disagree

If I had the equipment to measure the resistance of a block, then I'd have no doubt. As it is, I'm not sure if my idea will be good or not, but if I'm not happy (or maybe even if I am) I can switch the loop around and run the CPU in series to see the difference