[Build Log] - IRONBEAST - A fully water cooled, X99, Quad GPU, 900D build - Page 112 - Overclock.net

Forum Jump: 
Reply
 
Thread Tools
post #1111 of 1313 Old 09-09-2016, 02:44 PM
 
dilster97's Avatar
 
Join Date: Jun 2012
Location: Northern Ireland
Posts: 355
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 9
dilster97 is offline  
Sponsored Links
Advertisement
 
post #1112 of 1313 Old 09-09-2016, 03:08 PM
*cough* Stock *cough*
 
Angry-Hermit's Avatar
 
Join Date: Sep 2008
Location: New MExico
Posts: 290
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
O.O

devil.gifAngry-Hermitdevil.gif
Angry-Hermit is offline  
post #1113 of 1313 Old 09-09-2016, 03:52 PM
Overclocker
 
d0mini's Avatar
 
Join Date: Aug 2013
Location: England
Posts: 1,067
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 68
d0mini is offline  
Sponsored Links
Advertisement
 
post #1114 of 1313 Old 09-11-2016, 04:21 PM - Thread Starter
 
OCDesign's Avatar
 
Join Date: Nov 2014
Location: England
Posts: 578
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 44
Progress update:

After an anxious 48 hours things are finally clam and I have some time to report on the last couple of day’s developments.

Test loop, Idle

As you hopefully worked out from my previous - and uncharacteristically short - post, I began the initial water testing on Friday. (Sorry that one pic was all you got, suffice to say I was more than a little tired out by that point. You would not believe how many attempts it took just to type that one vaguely intelligible line). Though the process was not exactly plain sailing to begin with, I am delighted to say (at the risk of jinxing things) that the test loop is now stable, the crosslinked pipework configuration works, and that the system has been filled and running completely leak free for a little under 36 hours so far. The Aquaero is hooked up to my test bench’s power supply (so the pumps are following my normal PC usage), going through several cycles of 4-6 hour continuous running, 1-2 hour breaks, and sitting idle overnight without any problems.

Test loop, Running

To take this back a bit, last week I was doing the air leak test of the individual components and subassemblies – which all went very well. Unfortunately when I tried to test the whole system, it was useless. I’d forgotten, because this little project has been going on a while now, that those reservoirs had been shortened (by about 15mm) to fit the space more comfortably. Consequently the tubes lost their top screw thread and the caps have been reworked as push-fit. This doesn’t matter, because the water level doesn’t reach them, but it does mean that they are no longer air tight. Therefore, at that point, I just had to take the plunge and try it with water.

The vast majority of the loop was absolutely fine (thus far not one of those rotary fittings appears defective!), however those acrylic pipes at the bottom have been problematic. So I took them back out, checked and re-sealed the caps with even more flexible adhesive to be absolutely sure they were ok, and reinstalled them as carefully as I could. This appeared to work, and the system ran all afternoon and evening (continuously, from just after 5PM through to midnight) without so much as a drop (whilst I paced around nervously, constantly checking, and taking some photos for you lot). Sadly, overnight, it leaked a distinctly unambiguous amount of water over a similar time period from several of the same places . However since late yesterday morning, after a lot of painstaking adjustments to the tightness of the compressions fittings, it has remained absolutely bone dry. Which is a massive relief.

Reservoir, Idle

I believe the reason why the leak showed up overnight is due to a small issue with backflow. From about nine tenths full, the reservoirs will drop down to half level as the pumps fire up and water starts to get circulated through the system (the return to the reservoirs is at the top and a lot of the CPU-side components are higher up in the case, so the loop will not fill completely under gravity alone). When the pumps turn off, the flow reverses under the head pressure, the water is forced back up through the bottom of the reservoirs, and they slowly rise to their previous level. This means that 1. The volume and weight of water in the reservoirs is a lot more when the system is idle - which I think is why the leaks in the lower acrylic pipes showed up during that time, rather than over an equivalent period when the system was running. 2. Every time the pumps turn off, air gets re-inducted into the top radiator - which more than a little unhelpful. 3. (and most importantly of all) the reservoirs aren’t full to the top of the sleeve window during operation - and thus, as you can see from the pictures, only the bottom half looks the way it’s supposed to (utterly unacceptable!).

I really should have foreseen this one – but I didn’t. Fortunately it is a relatively simple fix, it just requires the addition of a check valve on the output from each reservoir to prevent any backflow of water.

Reservoir, Running

Although I have (for the moment) got those acrylic pipes sorted out, I must admit that I remain a little nervous about them. From what I can tell, the leak issue is to do with the seal at the O-ring rather than the seal between the tube and the cap. The rubber is quite thin and not that squishy, so it takes very little in the way of a misalignment for it not to compress evenly enough to seal properly. This may well just be down to shoddy preparation of the tubes on my part (more than likely, I am a total novice at this after all) but it’s also not helped by the fact that three of those pipes are connected to an assembly of rotaries at both ends, so the whole thing is not particularly rigid. I’m a bit terrified about knocking one of the tubes, even slightly, during the hardware installation (probably whilst doing the awkward job of fitting the cables) and then getting a nasty surprise when the loop is refilled. But then again some of that pipework does now have to be modified anyway (to fit the check valves mentioned above) so perhaps I’ll simply make a better job of it 2nd time around.

Beyond this a few other little issues were identified during the test so far. Firstly the CPU-side flow sensor (the Bitspower one) isn’t running. It may be that the jet pressure isn’t high enough and it needs the smaller reducer, but from the look of it I think it’s just stuck. Secondly the resistance of the GPU side isn’t nearly as high as I feared it might be, and as such the transfer tube isn’t fully able to compensate for the disparity when all three pumps are running at full speed. So No.1 and No.3 just need to be turned down a bit. Finally I discovered how very helpful it would have been if I’d actually remembered to label which set of cables goes to which pump.

All this notwithstanding - the concept works, the construction is holding, and - when full - the reservoir lighting effect looks the way I wanted it to. The water is currently going round and round on the inside of the tubes and not leaking all over the floor. So on balance, all things considered, I think we can chalk this one up as a success.

One more step forward.


Quote:
Originally Posted by Kojaqe View Post

Wonderfully , technically , creative build - subbed...

Thanks very much, its kind of you to say. I’m hoping it won’t be that much longer before you get to see the actual hardware fitted.


Obsessive Compulsive Design...a chronic condition for which there is no cure.


The IRONBEAST Project
“Obsession is the Only Path to Perfection”

post-flame-small.gifASUS X99-E WS Owners Clubpost-flame-small.gif
NVIDIA GTX 980 Ti Owner's Club
OCDesign is offline  
post #1115 of 1313 Old 09-11-2016, 06:21 PM
WaterCooler
 
Bloodbath's Avatar
 
Join Date: Jul 2011
Location: Australia
Posts: 1,481
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 62
OP how tiny are you're fingers? Keep up the good work.
Bloodbath is offline  
post #1116 of 1313 Old 09-12-2016, 06:57 AM
 
dilster97's Avatar
 
Join Date: Jun 2012
Location: Northern Ireland
Posts: 355
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 9
Quote:
Originally Posted by Bloodbath View Post

OP how tiny are you're fingers? Keep up the good work.
OP got noodles for fingers.
dilster97 is offline  
post #1117 of 1313 Old 09-12-2016, 06:59 AM
WaterCooler
 
Bloodbath's Avatar
 
Join Date: Jul 2011
Location: Australia
Posts: 1,481
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 62
Quote:
Originally Posted by dilster97 View Post

OP got noodles for fingers.
Noodles, more like tweezers.biggrin.gif
Bloodbath is offline  
post #1118 of 1313 Old 09-12-2016, 03:06 PM - Thread Starter
 
OCDesign's Avatar
 
Join Date: Nov 2014
Location: England
Posts: 578
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 44
Quote:
Quote:
Originally Posted by Bloodbath View Post

OP how tiny are you're fingers? Keep up the good work.
Quote:
Originally Posted by dilster97 View Post

OP got noodles for fingers.
Quote:
Originally Posted by Bloodbath View Post

Noodles, more like tweezers.biggrin.gif

Well I hate to disappoint, but they’re just the usual standard-issue set. Basic functionality, entirely unremarkable. They’re not even robotic…


Obsessive Compulsive Design...a chronic condition for which there is no cure.


The IRONBEAST Project
“Obsession is the Only Path to Perfection”

post-flame-small.gifASUS X99-E WS Owners Clubpost-flame-small.gif
NVIDIA GTX 980 Ti Owner's Club
OCDesign is offline  
post #1119 of 1313 Old 09-19-2016, 03:44 PM - Thread Starter
 
OCDesign's Avatar
 
Join Date: Nov 2014
Location: England
Posts: 578
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 44
A little progress update.

The initial water test has been running for a little over a week now, following my normal PC usage pattern, with no sign of any further leaks. I’ve tweaked the speeds on No.1 and No.3 pumps down a bit, with a slight bias (about 200 rpm) to the former. The setup is running smooth and quiet, at a fraction over 200l/h indicated (GPU-side), with no disparity in the fluid level between the two halves of the circuit. That’s really rather better than I’d been expecting frankly, given the complex layout.

Unfortunately the backflow and air re-induction issue (that I mentioned before) has been proving a persistent problem every time the system starts up, and I spent a lot of this morning trying to deal with some stubbornly trapped air bubbles that were causing a few difficulties for the pumps. However the new check-valves have now arrived, and I think they can be installed relatively easily with only minimal modifications to the pipework. So hopefully that should fix that.

As such I’m intending to drain the loop down again and make the alterations at some point during the week. Also amongst my latest order of parts are the two remaining Aqua Computer mps flow sensors, so I may try to put those in too so we can see how even the flow split is between the GPU blocks.


Obsessive Compulsive Design...a chronic condition for which there is no cure.


The IRONBEAST Project
“Obsession is the Only Path to Perfection”

post-flame-small.gifASUS X99-E WS Owners Clubpost-flame-small.gif
NVIDIA GTX 980 Ti Owner's Club
OCDesign is offline  
post #1120 of 1313 Old 09-25-2016, 02:30 PM - Thread Starter
 
OCDesign's Avatar
 
Join Date: Nov 2014
Location: England
Posts: 578
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 44
Another little progress update:

Over the last couple of days the test loop has been drained down, two new Alphacool check-valves have been fitted, the pipework readjusted appropriately, the flow indicator fixed, and the loop refilled again… I am very pleased (not to mention relieved) to report that the modification has been successful. The system now works as intended, and there have been no leak issues in the altered pipework thus far.


Check Valve, GPU-side, on reservoir feed to pump No.1

The two valves look very nice and match with the Bitspower ones really very neatly (if anything that photo exaggerates the slight disparity in the silver colour, to the eye they blend in perfectly well). Each has been placed as close as possible in the flow path to their respective reservoirs, being the first active components after the supply on each side of the circuit. I have used a couple of extenders to fit the one above, rather than trying to do another very short section of acrylic. The pipe that was there previously has been reused and is the one you can see in the picture below - with another little 5mm extender to make up the difference. You may also note in the picture above that an additional drain port has been added (the male QD3 just on the edge of shot), which helps enormously with emptying the water from this very complex layout.


Check Valve, CPU-side, on reservoir feed to pump No.2

I did have a little panic moment with this valve, when I filled up the loop and found that it seemed not to be letting any water through. First suspicion was that it was installed the wrong way round, but no. So the system was drained down a second time (at least partially - thanks to the new drain point mentioned above the water in the subfloor section could be ignored) and the valve taken out to see if was broken or jammed. At first that did seem to be the problem, but then when it went back in it did the same thing again. Eventually I worked out that the problem was actually a pocket of trapped air in No.2 pump, an issue which proved a bit of a faff to fix. Indeed purging the air from the system was a major aggravation in general (enough to push me into total despair at one point), but it is done and the new valves are successfully preventing the backflow problem. So that is a monumental worry (i.e. that my loop design was fundamentally flawed) safely off my mind.

The system is now running beautifully smooth and quiet. With all the air finally extracted from the circuit the pumps are all but silent - there is just the faint (and very soothing) trickling sound of the water flowing back into the top of the reservoirs. If anything the output flow from the four split-return jets is actually too laminar, and as a result the illuminated bubble-swirl effect isn’t quite as pronounced as I’d have liked. It looks best, and most like what I wanted, immediately after being powered up. The static pictures below don’t really do it justice, but I hope you enjoy them anyway.







Oh and if you’re wondering why the centre bit has gone missing again, the Motherboard Mounting Plate has been taken out to have the standoffs fitted - in preperation for the deconstruction of my faithful old Phase 2 test bench. Removing the plate would have been required anyway, in order for me to attach the 3rd aluminium bracket (the one that supports the GPUs) which shares one of the same screw holes. Or rather it will do, once I’ve finished making it.


Obsessive Compulsive Design...a chronic condition for which there is no cure.


The IRONBEAST Project
“Obsession is the Only Path to Perfection”

post-flame-small.gifASUS X99-E WS Owners Clubpost-flame-small.gif
NVIDIA GTX 980 Ti Owner's Club
OCDesign is offline  
Reply

Quick Reply

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off