Acheron

[luxbel]

Alrighty, so now that I have basically everything here to start the build, I can actually post some pictures of what I've got to do. Every time I look at the disaster area that is now my living room, I can't help but grinning.

Pictures are not soo hot. I lent the Alpha DSLR to a friend for a photo shoot, and I haven't picked it up yet. So here's goes nothing with the pocket sized point-n-shoot that I have. There's some extra hardward on the table that probably won't go into this build. I'm waiting for my G.Skill RAM and Anfi's RAM cooler to finish this off, but at least this is the beginning.

Computer components (not in pictures at the moment):
Lian Li v600
Intel Q9550 (2.83GHz, stock speeds - may overclock in the future)
evga 790i Ultra
evga GTX280 (stock speeds - may overclock in the future)
4GB (4x1) G.Skill DDR3 RAM @ 1600MHz 7-7-7-18

Watercooling components:

Swiftec MCP355 with XSPC top and BP 40mm Z-Multi
4x BP 90* elbows
2x BP 180* with 2x BP temperature monitors
6x BP 45* rotaries
mCubed BigNG with a mCubed Sensorhub extender and a mCubed Flowmeter
Lian Li Fan Controller for 2x Silenx 72CFM/14dba case fans (definitely not the real specs, but oh well, they work fine for now)
XSPC R120v triple rad in a Koolance rad housing with 3x Zalman FM-3 fans (all nice and snug with anti-vibration gaskets) with AC Ryan mesh fan grills
Danger Den MC-TDX CPU cooler (will need to see if the ThermalRight backplate I modded will fit behind the motherboard with the stock DD mounts - will probably have to shave quite a bit on the poles that DD shipped with it so it doesn't interfere with the loop or the power supply)
EK 790i mosfet coolers
EK 790i MCP cooler
EK 790i Southbridge cooler (from a visual look at it, it will allow for SLI, but no way that tri-SLI is ever going to fit)
evga Hydro Copper 16 GPU Full Cover block

I ordered waaaaaaay too many of those BP/DD compression fittings, but I guess it's never a bad idea to have a few left over.

[luxbel]

Now, I have a question here: how difficult will it be to fill the loop based on how that res/pump is configured (as shown in the last picture of this sequence)? My idea was to have that located on the bottom portion of the 5.25 bays with a compression fitting towards the inside of the case for the return and, for now, a plug on the port that you see in the last picture. I could place it vertical, but I may run into a few problems, since the DVD drive has to go in the upper two slots, making it impossible to fill the res through the large top that the Z-Multi has. This is a 40mm version, so I'm hoping I don't have the issues I've seen with a few of those 80mm (guess that's why the sponge is there).

The exit tubing for the rad is that bottom slot that you see almost horizontal to the 5.25 bays. By imperfect measures it's almost flat to the outlet port on the XPSC top.

I don't have a whole lot of space to maneuver, but I guess I could put a fitting on the "fill port" and tube it up to a DD derlin fill port, if I make a hole in the case, somewhere along the top, if I can get the tubing there without kinking, but would like to avoid that if at all possible.

[MomijiTMO]

I wouldn't do that. I would have it standing up because you really need to open the top cap to fill your res. Oh yeah do not make the mistake of having your cap off when powering your rig or you will get a shower. I hate that res top and as far as I know it is hell to bleed. If you can, remove that tubing that goes from the inlet to the top of the res. It is next to useless.

Once everything is bleed [might take a few hours] then you can turn it sideways and mount it on the ground how ever you want.

Keep the worklog going and I want more piccies lol.

EDIT : I have 4 left over fittings and 4 hose clamps. You can't be sure one won't be dodgy lol.

[xtclocker]

I wouldn't do that. I would have it standing up because you really need to open the top cap to fill your res. Oh yeah do not make the mistake of having your cap off when powering your rig or you will get a shower. I hate that res top and as far as I know it is hell to bleed. If you can, remove that tubing that goes from the inlet to the top of the res. It is next to useless.

Once everything is bleed [might take a few hours] then you can turn it sideways and mount it on the ground how ever you want.

Keep the worklog going and I want more piccies lol.

EDIT : I have 4 left over fittings and 4 hose clamps. You can't be sure one won't be dodgy lol.

+1 totally agree

[luxbel]

The 40mm has no tubing. I'm guessing it's too short. What it did come with, however, was a sponge, which I'm guessing will avoid a bit of the vortex. I'm playing with a couple of placement options. I'll take another picture of the case to illustrate what I mean, but there's a "non-standard" shelf between the 5.25 bays and the HDD rack that is a little bit higher than the pump. I can break out the dremel and accommodate the pump/res that way, so i'll have the full height, with enough space to fill and bleed between the res top and the dvd drive.

OR, but I haven't found this yet, is a top that for that res that has a G1/4 threading, and I'll just build out a fill port. At this point, both require a mod to the case, so which ever will work. I'm more comfortable modding the interior, since it won't be as obvious if I don't get it exactly right.

[MomijiTMO]

Ahh ok. Well the sponge will make bleeding super easy. You can remove it after there are no bubbles in your tubing and then top her up.

That mod sounds like a pretty good idea. =]

[pcaddict]

nice hardware.. but why zalman fans?

[SNiiPE_DoGG]

nice hardware.. but why zalman fans?

read vapors fan review, they are excellent fans, a bit more powerful than medium speed yates and they are quieter to boot

[luxbel]

K, so yesterday's experiment to make a perfect fit cover for the back end of that rad housing didn't quite work. 3 reinforced cutting discs later and still no dice. got sidetracked with a mATX build as well, so, yeah yesterday was cut short. the cuts were so bad, no pics of that. i'll tackle it again tomorrow. i really just need to install a small CAD programme (anyone can suggest one, that would be great) to print out the cut sequence, and try again tomorrow.

Waiting on: an derlin cover for the EK southbridge block, the copper top for the MC-TDX (that's in Miami right now, should have it by tuesday) and the anfi tec ram cooler, which i'm guessing will be here in two weeks.

now, if i can get the cover for the rad housing done, i might be tempted to build the box without the ram cooler, but i don't want to drain, fill and bleed all over again really. unless there's a way to keep most of the liquid in the system when i add in the block.

[luxbel]

So, after yesterday's epic fail on the rad housing backend cover, I decided not to wait around and actually start doing something. I went ahead and mounted the pump, res and bigng with the sensorhub while I wait for the last few pieces to come in.

http://www.mechanus.org/random/DSC01826.JPG

Can't really tell, but I cut the internal USB cables on the case to be a closer fit to the board. I have no idea why LL included like 40 inches of cable.

http://www.mechanus.org/random/DSC01830.JPG

test fit for the pump.

http://www.mechanus.org/random/DSC01827.JPG

one hole drilled in for the pump. the other one, I think it going to be a bit too far out - but we'll see. don't have a lot of space to maneuver there, but at least it was removable.

http://www.mechanus.org/random/DSC01829.JPG

http://www.mechanus.org/random/DSC01832.JPG

I could of gone for a round hole, would of taken a bit longer, but it's not visible from any angle, so why complicate myself with my inexperience with a dremel. As you can tell, it slipped on me at one point with the tungsten bit.

http://www.mechanus.org/random/DSC01834.JPG
http://www.mechanus.org/random/DSC01835.JPG

seems to fit ok.

http://www.mechanus.org/random/DSC01836.JPG

blurry shot of the bottom of the pump. i placed a temp sensor there that controls the front intake fan. at least this way, the pump is isolated, anti-vibration grommets on both sides on the screws (had to rig it, the screws i had were too short and the others too long - and that second hole, was a bit too far to the right), but with the impeller area uncovered. should keep cooler that way.

http://www.mechanus.org/random/DSC01838.JPG

there was a small space between the res and the bottom plate, so added extra foam to help with the spacing and just in case the res also vibrates.

http://www.mechanus.org/random/DSC01839.JPG

http://www.mechanus.org/random/DSC01841.JPG
http://www.mechanus.org/random/DSC01842.JPG

the bigng and the sensorhub installed and semi-cabled. it's a cable sprawl! wouldn't of been able to imagine how many cables are actually running from one place to another. it's going to be tight to get this all to fit and look good.

http://www.mechanus.org/random/DSC01840.JPG

see what i mean? it's going to stay that way until i get those last pieces in. man, it looks fugly.

[MomijiTMO]

Yeah I never used my classic mCubed thingie because of the mess.

[luxbel]

Finally got a hold of how to use the dremel. Now, all i need to do is paint the thing. At least I know the copper top for the TDX is here, now just need to wait patiently for the AF ram block and we're ready to rock.

http://www.mechanus.org/random/DSC01843.JPG
http://www.mechanus.org/random/DSC01844.JPG
http://www.mechanus.org/random/DSC01845.JPG

Test mount went OK, just need to put it under some even weight to level it again, although the torque pressure from the screws holds it pretty flat to the rad housing. the front acrylic piece to that is sitting in a negative mold so I can repour it in blue. Yes, you can buy acrylic premade sheets here, but it's not cheap at all. I only know of a single shop here that does them.

Yeah, I used to live in Houston. Who knows, I might of even been neighbours with R3.

[luxbel]

A note to the wise, if you are looking for a block that will use compression fittings out of the box, do not get the MC-TDX with the copper top. The acrylic hold-down that comes with it does not fit two compression fittings directly without a 15 minute date with the dremel. a little sanding is all that is required. in my case, a 45* and a compression fitting was all i needed. Will take pics in a few.

Updated after the fact:

Soooo, I probably did something I shouldn't of. This morning, before I left to DHL to pick up the copper top, i measured with the original acrylic top and cut the mounting poles. so now that I modded the hold down so it would fit the compression fittings, so results that the hold down is higher up on the block than the original mounting scheme.

i'm a little worried about the mounting tension, I have to compress the springs about 1/4 of the way for the thumbscrew to thread though the pole. if i try this just against the thermalright backplate, it's on there pretty tight. i'm wondering:

1.- should i worry about this with a spring mount at all? I remember reading, I think it was a mounting discussion where martin chimed in, that spring mounts do not put through 75lb of pressure to crack the IHS, however, that being said, this is on there pretty tight - i can still thread the thumbscrew further without much of an effort though.

2.- if i were to forgo the springs and just "hard mount it" at what point do I stop turning the thumbscrew? when it's hard to turn with bare hands?

[luxbel]

Hmmmmm, what is this? Looks like big yellow bags from overseas...

http://www.mechanus.org/random/DSC01846.JPG

Why, yes. They do bear gifts from overseas:

http://www.mechanus.org/random/DSC01847.JPG
http://www.mechanus.org/random/DSC01848.JPG

Man, I really need to get my camera back - this point n shoot is killing me.

[omaryunus]

AWESOME! stuff congrats on all the SWEET stuff :p

[luxbel]

Thanks, omaryunus. I'm cleaning the parts as I type this. Each is getting a nice 5 minute bath in near-boiling distilled water, basically, and i have a weak fountain pump that is pushing through another half litre of near-boiling distilled water. the rad got a treatment the other day with distilled and 98% alcohol - since vinegar looks like it's not the best idea, unless it's for a few minutes and highly diluted?

[luxbel]

It's up and running! Yeah, living on the wild side here. leak testing was a total of 30 minutes, HA!

I'll post pictures of the build shortly, but here's a few tidbits:

1.- I definitely think I have faulty DTS sensors on the chip. While today is not exactly cold (ambient around the comp is at around 22c +/- 1c), idle temperatures are too high for this to be working. It's reporting the EXACT same temperatures as it was on air at idle, cores are at 44, 30, 53, 53. I know with real temp that the last two cores have sensor movement of no more than 2, which probably means they're stuck. Cores 1 and 2 have a sensor movement 5 and 4, respectively. However, GPU did fall about 10c, from 48c on air to 38c on air. Everest reports MCP/SPP at 31 and 38, respectively. I think maybe the SPP does not have enough mounting pressure, but the same pattern repeats with the MCP, it was running around 40 on air.

2.- Now that being said, I can say I was quite disappointed, until I started running a few benches. We all know that the DTS sensors on the 45nm are not accurate at all at idle, so I ran Intel Burn Test after the liquid had normalized and the comp was up and running for about an hour. I remember I was really nervous about running IBT when I was on air, because cores 3 and 4 (the stuck ones at 53) would creep up to 70/71 during an IBT run. However, on liquid, the cores were running at 50, 47, 59, 59. Hmmmm, that's a pretty big drop.


3.- Lesson totally learned. I wanted to put a constraint on the project because I love my v600, it's pretty simple to transport. However, loaded with like 6 kilos of copper and a little over half of litre of water, it's no longer as transportable. yes, it's doable, but no, not as easy as before. Next time, I'll just get a bigger case.

4.- The Koolance QDC are not being used in this build, mind you. LL apparently does not make the PCI slots wide enough for the QDCs to fit through. I don't know if this is a general problem, or just one with the v600, but beware of that if you're going to buy those QDCs.

5.- That being said as well, I'm not overly enthusiastic about flowrates. The flowmeter that came with the mcubed module is reading at 2.1L/m, stable, which translates to about .55 GPM. I definitely need to improve that. I'm open to suggestions, but none which require another loop. I can't deal with that in this case, and unless someone foots the bill for the case, not going to go there "yet" (read: in other words, not this month. Maybe the next.). I do have an additional MCP355 here with me, so I could add it in the loop, but will need to make major modifications. I will be changing out the mcubed flowmeter (it's one of the 180* turns) for the Koolance one. Just need to know the impulses per second to get the mcubed to read it correctly.

6.- If anyone goes that external rad housing box from Koolance, note that the fan power adapter works great BUT it seems that an mcubed cannot control those fans. It'll seem them, but it won't be able to adjust their speeds - it seems as if the fan adapter does not pass-thru the commands. The Navigator software will tell you it's sending the power modifications to the fans, it will assume voltage has changed, but at least I wasn't able to notice a difference. The one thing about that rad housing that bothered me was that the fan adapter only has 2 plugs for 3 fans... hmmmm, strangeness indeed. However, the Zalman fans, even in a horizontal orientation to the floor are not loud at all at 100%. Yes, they're noticeable, but I think just about any fan at 100% will make enough noise to let everyone know it's running.

7.- Pump is totally silent, except, obviously, as it finishes bleeding the system. Don't hear many bubbles anymore, but I bet there are still some around in there.

Pics are being transformed as we speak and will be uploaded shortly.

[luxbel]

As promised, the pics of this insane build:

Here's the board as it was before, there's no turning back now.

http://www.mechanus.org/random/DSC01851.JPG

Nekkid ram:

http://www.mechanus.org/random/DSC01853.JPG

Step 1, forget everything you know about putting together heatsinks for RAM modules, and start here:

http://www.mechanus.org/random/DSC01857.JPG

Step 2, remember to buy thermal tape. mental note, remember to buy thermal tape. Repeat after me, remember to buy thermal tape. Forgot that one and had to run out and find something suitable. At the end of the day, worked great, chips make contact and since this G.Skill modules only have chips on one side, made it a lot easier.

http://www.mechanus.org/random/DSC01859.JPG

Step 3, repeat Step 2 about four times and you get:

http://www.mechanus.org/random/DSC01861.JPG

Note to the wise on the above process, make sure those square nuts are aligned well and DO NOT MOVE from their little grooves. Saves a whole bunch of aggravation later on.

Nekkid board... sure a lot of TIM on them here chips:

http://www.mechanus.org/random/DSC01862.JPG

Halfway through the build, a couple of friends came by to "help." Felt like I was a public works guy, 3 guys watching, one actually doing the work. Great amount of MX-2 went into this build. BTW, I hate that ceramique stuff, much more difficult to work with than MX-2.

http://www.mechanus.org/random/DSC01866.JPG

Board with all the blocks on. Had to fanagle a bit with that DD hold-down and the EK mosfet block to get it to bolt down correctly. I still think this might not be that good of a setup. See above on temps for why.

http://www.mechanus.org/random/DSC01868.JPG

Even more nekkidness. I play way too much WoW at this point, I think.

http://www.mechanus.org/random/DSC01871.JPG

Decency has returned!

http://www.mechanus.org/random/DSC01872.JPG

Bad, blurry shot of how the board looks like with everything installed. Not a lot of direct lighting at my place, so it's tough to take good stable pictures. Need to get a spot lamp.

http://www.mechanus.org/random/DSC01874.JPG

Loop built, me doing who knows what at this point. Yes that is the primochill Pro LRT tubing. Looks purple with a flash, but it's really blue.

http://www.mechanus.org/random/DSC01881.JPG

Liberal use of coils! That's the final routing currently being used. I think it can be worked on and tweaked (once again, see above for why). But I can tell you, working on this size of a case, with all the junk that's in there was quite a challenge.

http://www.mechanus.org/random/DSC01883.JPG

Cable management hell! Yeah, any effort I was going to do was shot by that PSU placement. Next time, I'll just buy a bigger case.

http://www.mechanus.org/random/DSC01884.JPG


Well, there you have it. It can use a lot of work, but for a first time, I guess it's not that bad. Need to definitely think about a bigger case next time, better routing and more fittings. Spacing between the CPU and the mosfets, for example, don't allow for easy tubing (can't use 2x 45*, the turn is too aggressive; can't use 2x 90*, the fittings are too long), and that DD hold-down won't fit 2 compression fittings without modifications. I might spring for a LL TYR x2000 before too long, however, so it might get improved, looks wise, before too long.

[MomijiTMO]

Woohooo that is a lot of updates in a short period of time. Great job on just about everything. The only thing I would ask you to reconsider is do you really want that flow meter. Restriction hell and flow meters aren't really necessary when you can monitor the rpm of your pump.

[luxbel]

Oh yeah, 12 hours yesterday just on this, crashed around 3am, but was worth it.

Well, that flowmeter was part of the mcubed purchase. The deal was, and I had the reason initial reasoning as you, that with just the rpms, I could guess at the flowrate, but if I look at the RPM, it's running at ~4200, however, flowrate is terrible. I was shooting for about 1.1 to 1.2, based on the estimates that I did empirically and with Martin's spreadsheet. However, I'm getting about half that. I partly blame the flowmeter, probably losing about .25 GPM just right there. I'm a monitoring freak, in that sense, so i do want to see some numbers on that. Based on my experience with these DTS sensors, I really can't base myself on those, at least not at idle. i'm still wondering if it's a bad mount issue as well

[Cpt_Zero]

omg thats too much for the pump i guess

i would recommend to make it dual loop...

just separate the video card coz they make tons of heat. :)

you can make reference of this build of docrj if you want coz this is also a full watercooling setup like yours.


http://i96.photobucket.com/albums/l165/nellrollon/teaserpic03.jpg

http://i96.photobucket.com/albums/l165/nellrollon/teaserpic05.jpg

[Webster]

Good call Cpt_Zero.

Since you have coils there really is no reason to have angle fittings between the Mosfet and CPU blocks. It just adds some restriction to an already strained pump.

Edit: I completely forgot to mention how nice this build is. I love the RAM block. Don't see that everyday. :D

[luxbel]

Capt'n, she won't give me any more power!


True, true. Dual loops at this point make more sense and I very well may move to that in a few months. I believe the DDC3.2 would of handled this without the flowmeter or the RAM cooler, however, measuring flow to know if I was in the 1+ GPM range would of been difficult.

I'm sure the GPU has a significant impact in the loop - no doubt about that. However, interestingly enough, the delta between air/water temps is about 10c. Ambient was about 20, water temps maintained stable around 30 to 31. Definitely, if I want to push the performance envelope anymore, I will need to go to a dual loop setup.

If I do go down the route of dual-loops, definitely would have to look at another case. The problem with something as complex as this in a v600 is that the PSU slides on top of the CPU area, making any kind of routing difficult because you have about a 4 inch clearance - which is why I tried to use angle fittings to minimize how high the tubing went - not 100% successful on that and I am quite certain that the PSU is exerting pressure over at least the tubing going from mosfet to mosfet. I can make a larger PSU bracket so it's further back, and that way minimize that impact, but it still would have the issue of doing a second loop. Too bad the Cora rads are alu - I could manage an interesting setup with a few of those, and the "passive" rads from XSPC. They would, however simplify the problem a 100-fold. I did, at one point, consider a full alu loop option, but couldn't get all the blocks in the same material, so i ditched it and went down this route.


As far as a new case, I was looking at the x500 (http://www.performance-pcs.com/catalog/index.php?main_page=product_info&products_id=24342) as a viable option. I currently run my rig with a single internal drive - have an external which images my internal drive once a week, so I could fit it in the 3.5 external bay. That would allow me to mount the mcubed, with the 3.5 bay adapter in a HDD tray. I could probably get away with sticking in a BIP Compact 240 (http://www.performance-pcs.com/catalog/index.php?main_page=product_info&products_id=3985) with a pair of the thin Yate Loons (http://www.performance-pcs.com/catalog/index.php?main_page=product_info&products_id=22998) as a loop for just the mosfets/chipsets/ram and re-use the external 360 that I currently have for the CPU/GPU loop. I can't imagine sticking the CPU in a loop with that BIP, I don't think it would perform as expected. I'd switch out the res/top combo I have now for the XSPC ones, and mount them more or less how you have them in your build, definitely like that cleanliness. That looks like a UN Bracket, no?

I'd just have to see how I mount the second MCP355 I have, since it has no "ears" for the screws, but I'm sure I'll find something.

Also, would probably have to make a hold-down for the MC-TDX. It's on there "ok" but I know for a fact that it's rubbing up against the mosfet. I need to thin out the "legs" going to the mounting poles to avoid this problem. It'll probably be easier to simply make one out of acrylic or sketch it up and make one out of copper/brass.