The northbridge block has been nixed; impossible to fit my block on a DFI LANParty UT nF3 250Gb.

No hunt for bigger louder pumps whatsoever; I'm using CSP-750s and sticking with them.

Yes, they're suspended--in midair; looking at the case from the side, the pump completely blocks your view of the CPU block inlet. If you want a pic, let me know, and I'll link you.

It's not just for overclocking. There's nothing wrong with wanting to overclock while at the same time achieving lower noise levels.

I'm not posting any pics here yet as they're not final; as a couple members pointed out, there's serious issues of serial air flow so I am very busy playing around with ducting arrangements so that both rads, believe it or not, pull cold air in from the front of the system. I think you guys can see the difficulty in that.

I have to manage to construct carefully fitted ducts that direct air from the wide open 3.5" bays into the rear radiator, while dealing with all the tubing and pumps that are in the way, and at the same time allowing the front rad's fan to maintain air flow across the chipset's passive sink, and all at once preventing the warm air from the front rad getting into the rear rad's intake area. My current problem is that the most recent duct arrangement is allowing the PWM temp report to increase 20-30C over any other duct arrangement, even though the chipset, CPU and GPU are reporting lowest temps with this setup and I can finally use my side panel again without adverse affects to temperature. Need to rectify the MOSFET temp issue.

Air tweaking has taken me well over a week alone

-Ed

EDIT: Btw I found that my desktop NewCastle with IHS is topping out around 2400-2450 while my DTR ClawHammer without IHS is topping out around 2500-2550, so once I get these air niggles dealt with, I'll be locking down the o/c and then cutting fan speed and dropping the voltage (I o/c at max voltage and fan speed until I hit the ceiling, then I drop the o/c a few MHz to an even number, and then drop voltage to the lowest 100% stable, and then finally drop fan speeds), so it really isn't just about o/cing.

Figured it out!!!

I've changed up the fan/shroud/rad assemblies so that now, both rads pull air from inside the case, through the radiators and blows it out of the system. The wide open 92mm opening in the side panel is plenty of bandwidth.

This completely eliminates my serial air flow problem once and for all; the only concern I have is that there's nothing blowing directly into the passive sink on the northbridge, but hopefully that just doesn't matter much...

I can't wait to see how it pans out now!

-Ed

I promised pics; here they are at last...

I was in a hurry when taking and processing these images, so color fidelity is not up to my usual standards, but I'm sure everyone will get the main idea, anyway. There were also some...interesting...artifacts related to the long exposure times I set so I could avoid setting up and running my excessively hot studio lighting. I did spend the time, as usual, to provide small pics for this post, and linked them to higher resolution versions of the same image.

Please enjoy the pics!

Starting off with three different angled shots of the inlets and outlets of the two individual reservoirs on each loop:



I staggered the inlets in such a way as to minimize conflicts in tube routing; too bad the same could not be done for the two outlets. I used TFE paste to seal the threading and it worked perfectly; TFE tape wasn't doing the job for me, but the paste got the job done.

Next are a couple shots of the pumps, obviously one for each loop:


I actually utilized a combination of four types of tubing for different portions of my loop, depending on how they needed to be routed and whether or not they needed to be weight/pressure bearing. The 1/2"ID lines coming out of the two reservoirs are stiff-walled, allowing me to actually suspend the pumps off the reservoirs without translating too much of the weight into the blocks. The line from the CPU loop's pump outlet runs directly into the CPU block inlet, as barely visible in the images above, and that line is also the same very stiff 1/2"ID tubing. The line between the GPU loop pump's outlet and the GPU block inlet is stiff-walled 3/8"ID tubing. All the rest of the tubing in the system is highly flexible Tygon R1000, with the line from the CPU block's outlet to its radiator being 1/2" R1000 and the line from the GPU block's outlet to its radiator being 3/8" R1000, while the return lines from both radiators to their respective reservoirs are 3/8". The R1000 is able to curve quite sharply without flattening out or kinking and I am quite happy with it.

Here now are the two radiator->shroud->pulling fan assemblies:


Both assemblies are arranged so that they pull cool air into and through the radiator and then expel it out of the case, with the CPU assembly exhausting out the front and and the GPU assembly exhausting out the back. The only real different is that the CPU rad assembly is inside the case, while the GPU rad assembly is out behind the case, with its fittings poking through. I replaced the two 120mm Globe fans at 12volts with two Nexus 120s without sacrificing any noticeable amount of performance while making huge gains in noise levels; I actually game with surround speakers now, and am not bothered too much by the noise of the system itself.

Here's a picture of the system in profile completely open and then with the side panel on, which is how I actually run it:


This angle provides a clearer picture of how I was able to suspend the pumps off of the reservoirs utilizing the stiffness of the stronger tubing. The AF92CT is serving two purposes; one, to help prop up the video card (just an extra measure of safety; it works fine without anything propping it up, as the pump's weight is sustained by the line out of the reservoir, not the card itself, but I don't like how much it sags without help), and two, to get some air blowing towards the puny passive cooler on the motherboard's chipset; a larger passive sink isn't an option due to its proximity to the NV-68 block on my GPU.

Here I have close-up shots of the front bezel after a bunch of fudging:


I tried to seal off as much of the front as possible to prevent recycling of warm exhausted air. I did what I could utilizing bay covers that were included with the case, and then utilized leftover gasket foam from my shroud kits to seal up the rest (the black, 1/4" thick foam).

[url=http://www.ngtechnik.com/Duals/Large/Check-Res-and-Pump-Action.jpg]
The reservoirs are actually translucent, although my picture does not show this well, and the coolant can be seen through the plastic.[/url] I left them open for viewing so that I can keep track of the pumps' activity during operation. When the pumps are running, the coolant is quite turbulent in the reservoir, so it provides me an extra measure of security to be able to keep an eye on then. I used a bit of foam to seal off the openings to the sides of the reservoirs so warm air can't get back in, and so the opening in the bezel doesn't look quite so odd.

[url=http://www.ngtechnik.com/Duals/Large/Angled.jpg]
The whole front when sealed.[/url] The side panel opening, which is actually 92mm, can be seen more clearly this way.

[url=http://www.ngtechnik.com/Duals/Large/Gaming-Station.jpg]
Domo-kun says, "Hi!" [/url] That's RTHDRIBL running windowed 1024x768 with the planetary demo running with the nVIDIA and ITE temp. monitoring dialogues for the system open on the desktop at the same time, but my long exposure setting combined with Ultrabright seems to have completely washed out most of the image.

Yes, I know everything is dusty; a lot of dust accumulated during a LAN event I went to a couple weeks ago involving lots of carpeting, but it will be dealt with. I also know the wiring is an atrocity; it will be dealt with once I feel I am done tweaking (which actually means within the week, since I'm pretty much done toying around).

The system is able to maintain 435/1180@1.5volts on the GPU and its memory and 237*11.0=2607@1.75volts on the CPU with this setup; GPU goes up to 61C under heavy load over several hours and CPU reports no more than 59C after several hours of Prime95, rock steady, with ambient air temps of ~26C (I like it a little warmer in my room). The air noise of the system is no longer bad enough to intrude on my gaming, and this machine only runs during gaming sessions, so the only time the noise of the system is at all excessive is under heavy seeking on the hard drive, which sends a growling rumble through the case, but even that doesn't bother me excessively on this gaming-only machine, particularly since I have a better idea when it's accessing the drive this way.

That wraps it up from me for tonight.

-Ed

quite nice i'm inspired by your idea of deliberately hanging the pumps like that, because as long as they do not put too much stress on the surrounding components it'd be a quick and effective way of dealing with vibration issues (to an extent.)

given the same parts though, i would have set it up differently, all within the one loop to make sure the radiators are doing as much work as possible.. the way you have it set up at the moment means that if one loop is underutilised then the other loop can't do anything about it.

but i like your setup nonetheless good job!

Thanks, and true enough. Actually, I felt that this is the optimal way to take advantage of the design of CSP-750. Their light weight lends themselves to ease of suspension, their small size allows them to be placed in tight or unique places, and finally, I decided that by utilizing the natural 90-degree chance in direction of flow between the inlet and outlet of each pump, I can effectively cut down on bends within the cooling system to help reduce flow restriction.

The primary advantage to this setup is not inherent from first glance. I have tried three different DFI LAN Party UT nF3 250Gb boards, as well as four different CPUs, and having the two separate loops made it much easier to make component swaps. It may be a little difficult to tell this way, but with the way the loop is arranged, I can completely swing the CPU block/pump assembly out to the right and out of the way, and also swing the entire video card/GPU block/pump assembly upwards and out of the way, and it allows me to swap the entire motherboard without draining my loop! This came in very, very handy the past few weeks with all my tweaking and experimenting.

Another reason why I finally decided to switch up to two separate loops is the fact that the word on the 'net is that the CSP-750 pumps' seals can fail, allowing coolant into the motor assembly, if there is positive pressure at the inlet; in other words, running CSP-750s in series can result in failure of the pumps. There are, from what I can tell, three ways to avoid positive pressure: one, install the pumps in separate portions of the loop, so that components after the first pump will drop the pressure of the coolant before going into the second pump; two, install the pumps in parallel; or three, keep them in completely separate loops.

I looked at my situation and decided that the two completely separate loops would be the most convenient and make the most sense, for a few reasons...

For one, having one loop for the CPU block and a separate one for the GPU block means that when it comes time to swap blocks, I only need to deal with draining half the loop and then when filling, priming a single pump.

Secondly, if you analyse the routing of my tubing, I actually worked very hard to tackle the issue of flowrate, not by increasing pump power (although I did, marginally, by upgrading from CSP-750 MkI pumps to MkIIs), but by tackling flow restriction everywhere else in the system, and making two separate loops ends up being the best way of doing this for my system. For example, the only 180-degree bend in the CPU loop is the return line from the radiator into the reservoir, and the GPU loop has only one 90-degree bend, being the line from the block outlet to the radiator inlet. I tried to figure out how to do a single loop with both pumps, all the blocks, all the radiators, and came to the conclusion that I would have to use more y-splitters or more curving in my lines to really manage such a setup. As a matter of fact, notice that the line from the GPU loop radiator outlet to the reservoir inlet is, in fact, a perfectly straight line! The lines into and out of the pumps in both loops bend no more than 45-degrees, and in all cases I was able to dramatically cut down on tubing in general.

Having the pumps in completely separate parts of a single giant loop would make it much more difficult to fill, prime and bleed the system, and having a parallel pump arrangement at the beginning of the loop would force me to use a more conventional layout, whereby the two pumps sit at the floor of the case and push upwards. In both cases, I'd end up working harder and/or running more tubing and dealing with the restriction that results from it. This setup also utilizes the absolute minimal amount of tubing necessary to get the job done, thus further reducing flow restriction. I very carefully planned out where the inlets would be and where the outlets would be in the radiators as well as on the GPU block so that filling and bleeding is as simple as tilting the system towards its back for a few second, and then tilting the whole system down on its open face for a few seconds! No crazy, convoluted shaking radiators, no fill, prime, fill, prime some more, fill some more, prime, rinse repeat over and over until a giant loop with random pumps all about finally gets properly primed (if even possible, as it really would be quite a bit harder when you have one pump at the start of the loop but then suddenly another pump halfway through the loop, and you're all the while trying to minimize positive pressure at the inlet of the second pump!).

Combine the new radiators when compared to the older single D-Tek Pro 120 heatercore, the direct-to-block pump arrangement, the large increase in 1/2"ID tubing in my system, the fact that I can now effectively utilize two, rather than one 120mm fan, and that the radiators themselves are far less restrictive than my original radiator, and you can see that I have obviously improved flowrate quite vastly over my original setup with the single rad, for both, coolant, and air, through the rads, allowing me to focus on both, performance, and noise, without resorting to louder, higher wattage pumps. I even cut down on the ratio of antifreeze in my coolant to reduce viscocity.

Finally, having this setup allows me to monitor the pumps' activities without relying on touching any part of the loop and feeling for vibration or flow; each pump's activity is visible just by looking at the turbulent action of the coolant in the reservoirs. Because my motherboard likes to vary the voltage of its fan power ports no matter how I set it, I needed to rely on straight molex to power them, but that nixes any chance of monitoring the speed of my pumps, as using a splitter that runs only the yellow line to my mainboard results in a shunt (I killed one of my MkIIs this way ; luckily, C-Systems was happy to replace it, since they don't warn of this issue anywhere--but they will be releasing a diagram to do so, soon). This meant that I had to rely on some other method to monitor those pumps, and having both in the same loop would make active monitoring difficult--two separate loops with two separate indicators are optimal in this respect.

The other thing that I want to mention is that I have learned a great deal about picking out and utilizing the optimal type of tubing in my loop. Using stiffer tubing where the run remains straight and or must be stiff to handle some weight, and using softer, more pliable tubing where it must be able to flex or bend without kinking or flattening out definitely made life much easier when working with the system as a whole. I don't need to use any CoolSleeves, which really causes huge amounts of turbulence at the tube walls, or even wrapping flattened parts out with larger diameter tubing and then snap clamping it down back to roundness, or run extra length to cut down the tightness of the curve, or come up with some sort of ugly rigid mounting setup to keep the pumps suspended in their optimal positions.

-Ed

EDIT: Found a typo/wrong word and fixed it.

Looking good, Ed. Incredibly better than your first loop. I like the case mods as well; they match your needs quite well.

How close is your rear fan to the wall? I'd guess that another inch or two might help airflow a bit.

I haven't tried my CSP on PWM but suspect that you might like a T-Balancer or similar. Something that allows you to monitor pump RPM and adjust the speed sounds like a great match for the Mark II.


Ed, you have a hardware problem.

Adam

Thanks Ed, (looking at fig.7) you've just given me all the excuse I need to treat myself to a nice 19" TFT for my 40th.




I think you'll find a slight increase in tube length is preferable to any acute/sharp/sudden change in direction in terms of minimising flow loss or adding to back pressure, even if the tubes don't flatten or kink


Having said that, it is a very nice job, well done. The fact that you can dissassemble the comp. hardware without upseting the WC side is sweet, nice planning.

Nicely planned and executed.

Keep up the good work and write ups.


Pete

very good points, very well explained your decisions have been more than justified!!

i can relate quite well to the disassembling problem, my first wc loop had everything crammed into the case and it was only barely possible to swing the cpu block out of the way to change the cpu. videi card or motherboard? forget about it

Well, the first DFI board I received from eMicroX, which was a Made In Taiwan piece, came from the factory defective; one of the inductor coils for the PWM circuitry was just bouncing around in the bag; my first desktop 3000+ NewCastle had already arrived days before, so I was in a bit of a time bind. I saw that the seating points were never punched through, which explains why the coil was never properly mounted. I had a friend punch through wit his iron and then seated and soldered the coil onto the board. It worked fine for a couple weeks but eventually died, so I sent it back to eMicroX for RMA, but I never dealt with them before and I wasn't about to wait an undeterminate amount of time for the replacement, particularly since their site showed no boards in stock, so I ordered a second board from NewEgg, now that they had them, along with a DTR 3200+ ClawHammer, and received a Made In China board at the same time. This board works fine, and I was using it to test the DTR chip in the system. While it and my first desktop 3000+ were both maxing out at 2.4G with the loud 12V Globe 120s, I decided to keep the DTR, since it was getting the same clock rate while having twice the L2 cache. My friend was interested in my NewCastle, so I loaned it to him to try it out, but he dropped it by mistake, so I went ahead and RMA'd it for him.

As it turns out, eMicroX was very quick with the RMA, and I got the replacement board, another Made In Taiwan piece, before the replacement NewCastle came in. When the RMA chip came in, I tried it out on the RMA board, and it turns out the replacement 3000+ NewCastle was able to hold 2.5, so I swapped the Made In China NewEgg board and DTR ClawHammer out for the new Made In Taiwan eMicroX board and 2.5G NewCastle in. I swapped out the fans for Nexus 120s to see what would happen and discovered, happily, that the thing held its 2.5G even with much less, "fannage." Finally, I was hoping for the luck trend to continue, and wanted to go back to 1MB of L2, so I decided to buy a desktop 3700+ ClawHammer. Luckily enough, it is doing 2.6, with its full 1MB of cache and all, with the quieter Nexus 120.

My friend who dropped the original NewCastle, btw, sprang for my DTR ClawHammer instead, since it's got more cache; so, this leaves me with two DFI LAN Party UT nF2 250Gb boards, one from each factory, one retail 3700+ CH and one retail 3000+ NewCastle.

That controller w/fan speed monitoring is a good idea; I should check that out some time.

The fan's only got an inch right now, but I agree with what you're saying, so I'll likely gap it out a couple inches; the only thing is that it will push my system into a more and more precarious perch about 4' away from a deadly drop to the floor, since my desktop, which is actually a dresser, isn't that deep.



Thanks so much!

I figure you may be right about the sharp angles, and there is one in particular that I'm sure you mean, being the one from the GPU block outlet to the radiator inlet; I am trying to come up with a solution to that one as I type. I'll post a pic of the solution I implement, but right now, I'm thinking to bolster that part where the tube goes onto the fitting with some 1/2"ID tubing clamped around it...

-Ed

This one's rather obvious, but that never stopped me from speaking up... The problem is the CPU loop's pump. I'd try moving that pump closer to its rad, rotated so the inlet faces straight up. You'd have a 90 bend from the res to the pump, a straight shot into the rad, and relatively gentle bends to the block and the res. Total tubing length would be about the same, but you'd lose the 180 into the rad. Add a little more tubing from GPU to rad, and that loop will look a bunch nicer.

Ah, I see; you mean to reverse the loop so the pump pushes into the radiator, and then it goes from the radiator into the block; I'll try that out next time I make revisions or changes to the loop. Thanks! The issue I feel more concerned about for now, though, is improving flow through the radiator for the CPU; I don't know why, but it when I leave the front bezel open and place my hand in front of the radiator, it doesn't feel like nearly as much air is flowing out of the CPU radiator fan as there is flowing out of the GPU radiator fan--doesn't make sense to me, since both assemblies are identical. I want to improve flow without increasing noise. Since the bezel is wide open when I do this feel test, I know this is a flow issue before the bezel and door even come into play.

-Ed

Do you have the side panel on? The 92 intake might flow better to the closer radiator, but I'm really grasping at straws here. I can't imagine the difference would be large.

Side panel is on, to prevent recycling of warm air that might creep around the system to the sides, once exhausted.

Opening the side panel didn't improve things, anyway. I'm confused.

-Ed

So I finally decided to try a different pump...

The revised CSP-750 Mark II pumps, like those sold by D-Tek Customs, have two slits cut into the motor housings to prevent overheating of the motor. Unfortunately, this also became a major noise leak. This was never a problem with the CSP-750 Mark I pumps (like the ones that failed on HammerSandwich and his friend) or the initial CSP-750 Mark II pumps without cutouts, such as those sold by Bigfoot early on. I was using Bigfoot, uncut pumps when Dave from C-Systems decided to send me revised pumps, telling me that it would be best to do a review based on the revised model. Not until I received the revised pumps did I realize the mistake they made in cutting out those slits!

In the end, I had to deny them a review because it would simply end up very negative, due to out-of-box acoustic performance. Although, as you see in my previous images, I was able to do some crazy looking mods on the pumps to block out most of the noise they emitted, they did, nevertheless, still make more noise than the previous CSP-750s, and at this point I deemed them inadequate for silent PC building. Around this same time, HammerSandwich's Mark I (which I sold him, in perfectly fine working order, I might add... ) and his friend's failed them. Hammer was wise enough to do some price shopping and came across a holiday special from Danger Den whereby there's a nice discount on all orders over $200; so, I chipped in and we made an order for three pieces, bringing the total above $200 and netting us a decent discount (sufficient for offsetting the additional cost of having two of the pumps redistributed).

Obviously, integrating the Laing DDC (aka Swiftech MCP-350) into my system would take major changes. I had to switch from two pumps with 1/2" fittings to a single pump with 3/8" fittings. On the same token, I felt that if I'm going to go through the trouble of (yet again... *sigh*) draining and dismantling my system's liquid cooling loop(s), I might as well take the opportunity to make other changes to the system. My focus was two-fold: decrease temps and decrease noise. So, what did it take to do this? Well, I had to do a little more than a pump swap, of course...

First thing to deal with, is that, as some might recall, I had an imbalance in air flow. My front radiator was flowing much less air than my rear radiator, even though the fan, radiator and shroud were identical, and both were blowing out through grille-free openings. I realized that it wasn't so much as the front exhaust being excessively restrictive as it was that my rear exhaust was flowing very, very freely; I decided to take advantage of this, by upgrading the rear radiator. I swapped out the rear Black Ice Pro for a Black Ice Xtreme.

I had experimented with using a second fan, blowing through the front exhaust, to help bolster flow through it. What I discovered was that in the case of my system, having a second fan blowing through helped flow of the front exhaust quite a bit (two fans in serial). The reason for this remains unknown to me, but the fact of the matter is that a fan, spinning around ~900rpm, to support the Nexus 120 that's pulling, improved flow by some ~75%, judging by the highly (inaccurate ) anemometer I have with me--my hand. So, long story short; after I had assembled the new loop, I did lots of air flow and temperature testing and achieved a very good balance, whereby I added aluminum Evercools to the blowing in side of both radiators (the front one, in particular, fully shrouded) and have them software regulated to remain relatively slow, and spin only fast enough to improve air flow dramatically without adding much perceptible noise.

My next objective was to deal with the new pump. As I said, the switch is from dual cooling loops with dual 1/2" fitted pumps into a single loop with a 3/8" fitted pump. Moreover, I can no longer rely on the direct-to-block pump setup because the new pump's inlet and outlet are parallel. In the end, I took a queue from HammerSandwich and suspended the pump off of the reservoir (having only one loop now, I was able to dump the second reservoir, too), pumping back up a bit into the front radiator, then from the front radiator into the GPU block, then into the CPU block, then into the rear radiator and finally back to the reservoir. The best part of all? I was able to retain the major ease of use factor--I can still do as much as a whole mainboard swap without even draining my cooling system!

Next, I took the opportunity to remove the mainboard, making room to take out the PSU and mod it. I removed the two stock fans from the OCZ PowerStream 420, which, while relatively smooth, spun much too fast. In place of the rear (exhaust) fan, I installed a Nexus 80mm fan and ran its wire out the front of the PSU, where the rest if its cables come out. The front (intake) fan was removed, then the motor and blade assembly were cut off the frame, and the frame was reused to mount the wire grille to the PSU (not very restrictive, but important for safety; I don't like the idea of possibly letting stray wires or other things get into a gaping maw in the front of the PSU. Plus, this looks more well finished. Doing this mod alone dropped system noise a significant amount.

Finally, I did an analysis of the voltage I was feeding to my CPU in order to achieve the overclock that I had, and found it peculiar that I was using so much more voltage than others (everyone was using no more than maybe 1.75V while I was at 1.9!). I did some playing around with it and realized that by backing down on my memory timings a tad, I could shave over .15V off the Vcore and still have 100.00000% stability. It must've been something about the memory controller on the CPU wanting a whole lot more voltage for a negligible gain in memory performance (in my case-the only difference lie in very obscure timings controls that aren't even available on other mainboards).

On with the pics!!!

[url=http://www.ngtechnik.com/DDC/High/Overall.jpg]
An overall view of the machine with the side panel removed.[/url]
The rear radiator has grown considerably in depth, and this image provides a good angle to compare it against the front radiator, which is identical to what used to be in the back. As you can see, the new loop still allows me to just bend everything out of the way to make room for relatively effortless components swaps. The AF92CT doesn't ever turn on; it's just there to help support the weight of the video card, which sags under the weight of the Danger Den NV-68 block if not supported. The pump is suspended, and gently set on top of some foam, primarily to keep it from leaning into the side panel and making a massive humming racket. The RAM looks funny because I removed the heatspreaders from them, but they're two 512MB sticks of PQI Turbo PC3200.

[url=http://www.ngtechnik.com/DDC/High/FrontBottom.jpg]
Here's a closer look at the pump and the front exhaust.[/url]
I would consider suspending the hard drive were it not for the fact that doing so would take away greatly from the system's ability to travel without having to open it up and make preparations. This is, still, meant as a portable, liquid cooled overclocked gaming rig. Now, it happens to be very quiet to boot.

[url=http://www.ngtechnik.com/DDC/High/FrontTop.jpg]
Here's the upper section behind the front.[/url]
The change to single reservoir is visible here, as are the lines into and out of the pump, into and out of the front radiator, into and out of the CPU block and into and out of the GPU block. As HammerSandwich suggested, I was able to completely eliminate 180-degree bends. I apologize for the poor cabling; there was just too much of it to deal with.

[url=http://www.ngtechnik.com/DDC/High/RearBottom.jpg]
The pump and GPU block from a different angle.[/url]

[url=http://www.ngtechnik.com/DDC/High/RearTop.jpg]
Supplemental rear exhaust fan that only activates at higher temperatures,[/url]
CPU block from another angle and the lines into and out of the rear radiator.
I had a bit of difficulty trying to get that rear fan to stay due to the way the rear radiator is mounted, but I finally figured it out and used foam to wedge the fan in from both sides (can't see it, the way it's done). The computer doesn't normally sit so close to the wall during operation, but when shut off, I push it back towards the wall so it's not as easy to trip on, like during my quick photoshoot this evening (it's all on the floor; I got these angles by using a minitripod right on the floor). The depth of the machine had gotten too much to deal with on top of my shallow desk, so I placed it, for safety reasons, on the ground.

That sums up yet another entry in the log. Any bets as to what I (might) do next?

Btw, temps are down across the board thanks to the combination of less voltage, more flow and more radiation. Noise has been cut by a huge margin, and I can easily sleep in the room with this computer on, under load. In fact, I did all my stability & stress testing overnight for several nights with this machine and it didn't affect my sleep one bit.

Per usual, I'm looking for comments.

-Ed

EDIT: Minor fixes for spelling, vocabulary and grammar. Took out some extraneous content.

Look kids! Here's an example of what you can accomplish with a lot of work when you start with the wrong case!

Um, sorry, I think.

Yet again, you've managed a significant improvement, Ed. Can't argue with lower noise and better temps (even if that's partly from the voltage). Are the Evercools run by Speedfan or what?

The tube running into the GPU looks acceptable but a bit tight. I'd have probably gone from front rad > CPU > GPU > back rad, but I wouldn't bother changing it just for that.

Also, would you please detail how the front rad is mounted? That's a lot of weight cantilevered there. If the Nexus is joined to the front panel with regular fan screws, I'd worry about it. Long bolts thru the entire sandwich are probably fine.

Can't think of a better case for the job; a steel case would hurt portability quite badly and the Super LAN Boy is not as adequate as this case for the job.

Evercools are controlled using ITE SmartGuardian. It's an excellent utility that's included with the DFI LANParty UT nF3 250Gb (of which I proudly own two). I can control the temp the fan shuts off, the temp it turns on, the temp at which it should hit max speed and what speed the fan should be at when first activated. Two separate controls, one based on CPU temp and one based on PWM temp.

I did it from front rad->GPU->CPU->rear rad->reservoir just to ensure absolute maximum ease of filling & bleeding (I'm sure you can see how it differs from the arrangement you're suggesting--focus on the CPU block). Of course, with the DDC's flowrate and head, it's probably worrying over nothing.

The shrouds come with gaskets and long bolts; the bolts for both fans (which means the whole assembly) bolt clear into the rad.

-Ed

looks good, but why did you remove teh heatspeaders from your ram?

Also did you notice a temp difference without the voltage drop?

Without the voltage drop I was already able to get same or better temperatures with less noise, but that's because of the massive drop in PSU noise. Actually, I do get less noise in regards to 120mm fans as well, since the new rear radiator is shrugging off more heat than the previous one. Plus, because we have two rads at once cooling either chip, that also reduces heat (the dual loops wasn't too great an idea, more of a way to avoid issues of positive pressure at the inlet of CSP-750s).

The heatspreaders were removed in an experiment to test if they really did achieve tighter timings or higher speed rates with the same timings (several members of another forum reported that removing the heatspreaders on PQI Turbo PC3200 had this effect), but my testing showed negligible improvement or detriment to removing the spreaders. That's with only one stick installed, too. The reason why I mention this is because with both 512MB sticks installed, the single-channel controller of S754 K8 must resort to relaxed timings to maintain stability. This is the primary drawback to S754 vs. S939, but any sane person will point out that this is purely an issue for overclockers and noone else. On the same token, it should be noted that in 99% of realworld usage, S939 does not yield noticible performance improvements over S754; K8 in and of itself does not benefit from the raised bandwidth in the manner of NetBurst P4.

-Ed

Ed,

To you and all of the other contributors to this thread...

This thread (and the one that preceeded it with your initial planning and pump purchases and eventual trials with bacteria) were fantastic pieces of "reporting" with tons of valuable information. Brilliant work!

It has been great to stumble upon it 4 months later and be able to go through the whole time line in one shot. I laughed, I cried, I sat on the edge of my seat, and I shout for more!

I'd love if you could post a quick follow-up now that it's 4 months on? I did a search on your profile but couldn't find any threads along those lines or any other mention of this rig.

1. How is that new single pump holding out? It seems a little puny.
2. How is the system in general holding up?
3. Are you still using it? / Are you happy with it?
4. Any problems with leaks?
5. Any "valuable lessons in watercooling" a wisened fellow such as yourself could impart to us lowly neophites?

Thanks to you, and all the other posters here for all the work and great posts! It has also been a great one-stop-shop for everything someone intersted in doing his first WC system (me) needs to know.

Cheers,
CList

Hey, CList: WELCOME TO SPCR!!!

... nice to see yet another NYer (represent!).

Anyhow, moving on to your questions:
1) The DDC is doing great; I cannot A/B it now against when it was new, but it feels as though after a little break-in, it has smoothed out a little. It is still quite audible when the blue LED fans are shut off, because I have it at a full 12V and completely unisolated, but flawless operation is flawless operation.
2) No biogrowth, no corrosion, nothing. It's working extremely well, although I know now that the NV-68 block is a big mistake.
3) I am definitely still using it, as it is my gaming rig, and I do get time on the weekends to hit it up, plus it's the machine I let other people use when I get visitors (which is quite often, believe it or not). As for satisfied, I am mostly satisfied, but I believe I can do better; I think this loop can be further refined, tightened up, cleaned up and more efficient.
4) No leaks whatsoever once I started using teflon paste on the reservoir and plier-shut nylon snap clamps at all fittings.
5) Unfortunately, there's too much info for me to impart even a good fraction of it. To put it in another way, I'm one of those people that learns best from $%^&ing up, and from when I started until now, I uh...did a whole lot of learning.

If I did have any really good tips, it's to first do as much research as possible; cram your brain full, then cram it even more, past capacity. Mistakes made when you actually begin making purchases will cost money. Knowledge is free, all it costs is time. Another thing is like what we use in the culinary world, mise-en-place. Plan and be prepared for every single little detail. Some of the more critical things include how the loop is layed out (more importantly, minimal pump intake restriction and absoulte maximum ease of bleeding air out), the metals in the loop (avoid galvanic corrosion--I haven't had it, but it sounds absolutely awful), what's in the coolant (do it wrong and mother nature will do its best to ruin your experience; nothing frustrated me more during my learning period than biogrowth) and overall compatibility of components. Also, considering you're building for silence, a good amount of focus needs to be placed on the pump. The DDC is a good choice, but I personally am not confident I could build a completely inaudible rig using it without a lot of work to isolate/muffle it.

Please feel free to post more specific questions, as I'd love to help.

-Ed

Ed,


Word up!


Thanks much for the quick reply!
I haven't even bought *any* of the parts for my new rig yet - no case no nothing. I will be going with a 6800GT and those things are all loud as hell, so I'm just assuming I'll want to WC it. I've also read so much about it now, that I'm itching to try it, just from a "having a new project" perspective. For now though, as you suggested, I'm trying to do as much planning as possible up front.

I'm pretty handy and have a good selection of tools, and have done things like stereo installs and what-not that came out pretty well, so I'm not averse to modding. Plus living in NY with easy access to Canal street and it's many foam, plastic, and metal stores leaves me with a lot of things to play with!

I may very well just decide I don't want to bother with the WC since I don't plan on OC'ing it - might just go with a Zalman on the CPU and one of those artic-cooling blower things on the GPU, we'll see.

If I do go with WC, I'll be sure to take some pics and post them here with a little bit a of a different perspective as noise will be a primary goal of mine.

Again, just wanted to say thanks for all the great info.

- CL

I've ordered a pair of CSP-MAG pumps. While my guess is they won't create as much flow in my system as the DDC does, I'm hoping I can develop a lower noise floor with them. I'd easily sacrifice some flow performance for lower noise compared to the DDC, and as I found in the past, the super light weight, small size and 90-degree angle between inlet and outlet allow me to be very creative with the design of the loop.

Likely changes will also include a switch to Socket 939 and one of the new 90nm, 1MB-cached cores on a fanless SLI board (but just one video card for now). I'll be changing the video card block to a Maze 4, as I've pretty much settled on the fact that the NV-68 is a waste of flow and doesn't perform as well from a GPU-focused point of view compared to the standard Maze 4.

Additionally, I will likely drop down to a single radiator because the new CPU will be significantly cooler, and there will be no GPU memory heat included in the loop. Judging by the design of the P180, it is quite possible that I will be able to finally swap out the system from the Tsunami Dream--I no longer need the portability.

-Ed

Hey, Ed. Long time, no talk!


Initial reports at Procooling imply that Mags are incredibly quiet. CSP-750 owners, remember that C-Systems has an attractive upgrade offer.

BTW, I must have a particularly good DDC. It's simply not a noise problem in my setup.

The new Swiftech MCP655 is reported to be a "silenced" MCP650.

Hey, Adam; how's everything? Indeed, long time no talk/type--my new job eats up so many hours.

I do not like the idea of undervolted pumps, particularly DDCs, which were known to have starting issues (although admittedly, they were due to a programming defect in the controller chip on only the first 200 pieces off the assembly line, I'm still not taking chances). It is also likely that my DDC is quite the opposite; not that quiet. I know for a fact that the noise floor is currently controlled by the pump, as even all four fans at once do not contribute sufficient noise to exceed that produced by the pump (although yes, all four fans are undervolted sleeve-bearing Yate Loons). Dave has e-mailed me personally and informed me of several nice tidbits:

According to his ears, not only is the CSP-MAG quieter than CSP-750 Mk II (which were louder than Laing DDC), but they are in fact even quieter than CSP-750 Mk I!

The motor compartment is filled with silicon, so it should be fine underwater, but it is not tested for this or purposefully built for this and its warranty is void if this is found to be the cause of failure. He said if I wanted to try it, I should do a couple drives on it just to be sure it's 100% full (of silicon).

CSP-MAG has both, higher head and flow than any of the CSP-750 models.

Rotor rides on fluid bearing surface once started; bearings are jeweled ceramic (sapphire to be exact).

Almost no metal-to-fluid contact at all in the new design, except for the bearing support, which is now aluminum with a stronger than ever anodization performed on it. Delrin cover reduces metal content of the pump even more.

No shaft or shaft seal whatsoever to worry about thanks to the mag drive design; no way for coolant to reach the motor assembly (except maybe if the whole pump were submerged). This also means the pump works safely in serial as well as parallel, as there is no shaft seal that can fail under positive inlet pressure.

Tolerance for running even in zero flow without failure.

They are also releasing a cooling block model; it's literally a CSP-MAG with a copper microchannel block at the bottom, turning it basically into a pump-cum-CPU/GPU block. I've sent a query to see if it will come with any platform-specific mounting hardware (Socket 754/939/940, etc.). They told me that the pump had this usage in mind from the very beginning of its design period, and much focus was placed on virtually eliminating any possible emittance of interference, so it can sit right on a chip without affecting operation. This item is also known as the Head-Unit.

Choice of covers; one with rear inlet and top outlet (90-degee in-out direction change, like CSP-750s) or one with top inlet and outlets (180-degree in-out direction change, like DDC). I made a special request with Dave to purchase spare covers, so I will have one of each direction for both of the CSP-MAGs I ordered.

The pump draws only eight watts--yes, eight watts. There is little to no chance of overheating the pump, which is why it is a sealed case design once again, just like the far quieter CSP-750 Mk I (compared to Mk II), and will dump an utterly negligible amount of heat into the cooling loop itself (1-2 watts, typically, according to engineering tests).

Overall, I am quite excited about this pump and hope it performs as quietly as people are saying (Dave actually handles support for C-Systems and the majority of his CSP-MAG issues are with people unable to figure out of the pump is running or not!). Dave actually said the pump might be only half the noise level of a CSP-750 Mark I! If that's the case, this pump handily trounces anything else I've ever seen (or rather, heard) for the price range.

-Ed