How does this look for a flow plan?
Water goes into CPU, out to mosfet #1 then to chipset block, then mosfet #2, #3 and to rad
http://i232.photobucket.com/albums/e...2072010163.jpg
Printable View
How does this look for a flow plan?
Water goes into CPU, out to mosfet #1 then to chipset block, then mosfet #2, #3 and to rad
http://i232.photobucket.com/albums/e...2072010163.jpg
Very nice. No kinks at all.
Wes
Thanks Wes, final colour will be black, the silver is just for testing, but I'll decide a bit later whether to keep it.
What about the block order? I thought I'd want to get the hot CPU first, then over to the hottish dual N200's as soon as physically possible without messing the tubing up. Then to the other mosfets and back to the rad.
Order like that doesn't matter one bit. Not in the slightest. The water going over all of those components is the same temperature, called the equilibrium temperature.Quote:
Originally Posted by Biffa
At a given load (idle or stress), your loop temperature will be constant - equilibrium temperature.
Water does NOT enter your CPU block at 27C and leave at 35C, enter your MOSFET block at 35 and leave at 42... it just doesn't work that way.
The difference in water temperature from radiator in and radiator out is .1C if you're lucky. Route your tubing so it doesn't kink and route it so it keeps the tubing as short as possible. Keep your pump immediately after your reservoir to aid in bleeding the loop, and you're good to go. After that, order doesn't matter at all.
2-3C from coldest to warmest water if a loop with lots of hot parts and a rad at the end. Quite a bit more than .1C. You on the other hand will see a small amount, maybe .1C added by the mosfet then to the next mosfet then to the CPU that will add maybe 1C, then off to the rad I'm sure.
Still, routing to the CPU first really isn't a big deal, your just fine as it is.
I thought it matters only with cpu block as it has inlet/outlet? So in theory would this work for say a CPU+ChipsetQuote:
Originally Posted by Polizei
Cpu->Chipset->Cpu-> etc?
Excellent, thats sort of what I had experienced, and makes sense, but as I am rebuilding the loop anyway I just wanted to be sure as I don't want to do it again for a while :)Quote:
Originally Posted by Polizei
Interesting about the radiator in/out temps, I would have thought that as the job of the radiator is to cool the water that it would drop the temps more substantially as it flows through.
Also refering to gabes dual vs single loop thread I'm thinking of running my dual pump res in serial rather than parallel, would this work?
http://i232.photobucket.com/albums/e...P1040831-1.jpg
This should give redundancy if one pump fails, but because the inlet just goes into the joint res (old model) its not pushing water directly into the other pump. Any downsides?
So I have two 18W DDC/MCP355 pumps under the res. One would effectively be pushing the water around and the other would be pulling it back into the res. Would that setup be enough to handle CPU, mobo blocks, two radiators and 4 full cover gpu blocks? Or am I better assigning one pump to the cpu/mobo and the other to the gpus which is how I have it now?
payback time - hope this helps (at least partly) :)
http://www.xtremesystems.org/forums/...d.php?t=225249
Well all I can do is give it a try right? :)
Actual order of the loop makes minimal/no difference, not even the res before pump rule (although it does make filling/bleeding a lot easier if you follow that 'rule').
Just ensure that you are actually using the Inlet of the CPU block as an inlet.
With the jet-type blocks we're using now, getting the inlet and outlet ports backwards will have an effect on your temps.
Yeah its a XPCS Rasa, has the inlet and outle marked on the block :)Quote:
Originally Posted by Soulwind
Cheers.
So a few updates, I've got everything connected up and have done an overnight leak test and all is well. There have been a couple of modifications which I'll briefly cover but when I have a bit more time I'll throw together a worklog.
Again like most of my installs I would consider it neat and functional rather than super creatively awesome like the fantastic builds we often see :up:
So, kept the routing the same as in the original pic pretty much, just added the radiators and the dual pump res to the mix.
This is what I ended up doing to the res. Now its done I'm not 100% whether its an effective solution, it doesn't seem to be causing any problems right now but I'm not 100% sure if its any more or less effective than having one pump for the cpu/chipset and one for the graphics cards.
But I think it looks cool! :D :up:
http://i232.photobucket.com/albums/e...P1080573-1.jpg
Heres a detailed shot of the cpu/mosfet area:
http://i232.photobucket.com/albums/e...e/P1080570.jpg
And the whole shebang:
http://i232.photobucket.com/albums/e...e/P1080563.jpg
Its still bleeding at the moment, and there is much more turbulence in the reservoir than before, with a constant noise of bubbling water.
The water doesn't seem full of bubbles unless tip the case around, but I'm not sure if thats the res sucking air into the loop or the loop disgorging some trapped air into the res. :shrug:
So I'm going to give it a bit of a run and see if I can get as much air out as I can, then I just recieved a part to make a drain line. Not sure where tho, I was thinking just to the right of the motherboard tray, right at the top of the lop where it goes horizontally from the res over to the inlet on the PA120.3 unless someone has better idea. :confused:
It would appear that with your current gfx piping, you have created a short cycle at the top card which would restrict the flow to the lower cards. Water will take the path of least resistance don't forget :)
Flow in the top card and return from the bottom imho.
His flow setup is good (parallel flow works for GPU). The lowest resistance path in this case is through the 4 blocks simultaneously.
For your drain just think that you'll have to tip the case so it becomes the lowest point in your loop, makes it a lot easier. The PA120.3 is on top, am i right ?
And whoa, impressive setup, is this going to be a crunching powerhouse ?
It would appear so but the temps are quite even across all cards, with them all folding at the same time and an ambient of 30°C the cards are all within a few degrees of each other, which isn't any worse than they were in serial :confused:
To flow in at the top and out at the bottom I'd have to jury rig it like the serial setup I had with a different on the bottom block with the fittings offset from the other blocks and custom loops of tubing to link it all up (like my avatar). Problem is there is no room between the bottom card and the PSU to put any connectors in :(
I've got a new AquaGrafx TwinConnect coming (hopefully soon) which does serial and parallel and will fit in the space (similar to the EK connectors)
At the moment though its the CPU thats getting hot
The way this is set up should provide all cards with about the same amount of cooling.Quote:
Originally Posted by P5ym0n
It does seem to be cooling them ok at around 60°C give or take 1°Quote:
Originally Posted by Alexandr0s
But the CPU is getting hot, 84/81/81/83°C :shocked:
Its stabilized at that temp, and its folding as well as the gpus so its not slacking :D
But I wonder if I hang a TFC 120 off the back fan will that have a positive effect on the entire loop to make a difference.
Also still not decided on the dual-pump-res-flow-mod as to how effective it is.
Thanks for all your comments guys. :up:
60c :eek:
That's ok given the amount of components that are being cooled there...
And the ambients, if it was winter or I had aircon I reckon GPU's would be circa 45° and CPU around 65°
biffa how are gpu temps in paralell compare to before??
Not too much difference.
Redoing the loop at the moment, different order and added an extra 120 radiator off the back fan.
All 1 Loop yeah?
Dual pump res, but I removed my mod on the res. I'll explain more later when I have a bit more time. Suffice to say the turbulence and "gravelling" sound is gone
parallel would work well with the dual ddc's, and did it not work because the IN part of the pump is not directly connected?
These cpu temps are on what speed and what vcore?
4.2Ghz and 1.375, similar to yours I think.
You are correct its not directly connected but the pump pulls pretty hard on the downflowing in-pipe.Quote:
Originally Posted by TRICHEESE
I've taken the connectors off the pump, cool as it seemed in my head it just was causing alot of probs without any gain.
Now I've got it setup like this:
http://i232.photobucket.com/albums/e.../pumpflow2.jpg
I've left it running on this fairly hot day and see how it gets on :up:
Id move the vrm/nf200 block and the 120mm rad to the gfx loop dude.
360 is WAY more than enough, infact you could comfortably use the 240 with gt1850s.
CPU alone on the 240 getting cold air and everything else on the 360+120 getting case exhaust air would be the way i'd set it up for best temps. Youre pumping the heat from the gfx cards straight through your cpu loop rads at the moment. They generate far more heat than the cpu, so youll be better off with the cpu heat going into the case and the gfx/nf200/vrm going straight out.
Looking at your kit id go like this:
360>gpu>120>mb>vrm rear>vrm top>vrm front>360 - pump/res either infront or behind the 360, however fits best for tubing route. My only concern is gpu flow rate. Its a restrictive loop and the parallel setup will reduce flow to each block again. Test temps carefully. If they're high go back to serial to get gpu flow back up. DDC should be good in that restrictive a loop though. Ive run my gtx260 parallel sli with a mb block and an EK DCP2.2 pump - flow was pathetic but temps were still fine. I'm confident but be cautious :).
CPU loop would then be fast flow/cold air and simply pump/res>240>cpu. Push/Pull the fans on the rad if you have room. GT1450s for silence GT1850s for low temps.
Hmm I thought best practice was to have the rads cooling the water before it enters the res? Thats why I have the 360 after the cpu going into the Res and the 240 after the gpu's going into the res.
See what you mean about swapping the 360+120 for the gfx cards and the 240 for the CPU, only problem I might run into is the HDD's are right behind the 240 but worth a try. :D
Thanks :hammer: just when I didn't feel a yearning to drain my loop again this week :D
At least I put a T line in this time round. :up:
Actually I was a bit concerned at how soft the tubing was getting :eek:
Current temps are 80 on the CPU cores and 60 on the GPU cores, mind you it is pumping out nearly 62K PPD on its own,
Yeah tubing gets soft fast in hot setups, make sure everything is tightly clamped around fittings :)
Best practice is having the reservoir just before the pump inlet (negative pressure in the res then), and then everything else in the order that produces the shorter tubing length, and the least elbows.
PiLsY: parallel setup for GPUs work very well and have actually a positive impact on total loop temperature. See the magnificent thread that Gabe posted around here not long ago on that topic.
gmat - I've been running parallel sli for nearly 2 years now. Its taken a long while for the general readership to catch up with me (and several others here who i got the idea off) and I agree its better in most situations. Gabe's thread has nothing in it that I haven't been trying to tell people for years. People seem to have forgotten the negative aspect though - flow reduction. By running multiple blocks in parallel you divide flow. In tri and quad setups you need to be careful to have enough flow in the first place so that the split does not leave you below flow capacity for the blocks.
Biffa's quad setup on the loop I suggested would be VERY restrictive. Flow would be down below 1gpm even without the gfx blocks. Add those in and split the flow by 4 and you end up with each gpu block receiving somewhere between 0.15 and 0.25gpm. I know my EK FC260 gt200b blocks work "ok" at approx 0.25gpm - temps are about 6c higher than at approx 0.5gpm and 8c higher than at approx 1gpm. However temp increase is not linear so id expect it to fall over quite quick below 0.25gpm.
Just needs some caution testing as nobody actually knows the lower limit for flow through the various full cover blocks that are out. Biffa's going to be pushing the envelope. I was told catagorically by many that below 0.5gpm per block would end in an instant cloud of smoke. That's obviously since been prooved utter rubbish when I went ahead and tried it.
I'd urge Biffa to try - everyone here seeks the perfect setup for their rig, just be careful doing it :). As I said, I reckon you'll be ok, but I wouldnt want you risking your expensive hardware on my supposition. Just cycle the pump a good few times to make sure each gpu block has good flow and no trapped air. Do the finger test while its booting - if a gpu block has too low flow/air pocket blocking flow itll get hot to the touch FAST. Finally load up evga precision and watch the temps carefully on first boot and during testing.
gmat: Its a dual pump bayres, can't get much closer than that :)
PiLsY: Temps on the cards have been good throughout, much better than air with around 60°C max when spanking the bj's out of them :D Its the CPU thats suffering, its hitting nearly 90°C worst case and averaging around 82-85°C
Its not crashing or anything and I'm pushing it all pretty hard (100% on all cores and all gpus)
I'm going to try rearranging rads again when I get at chance and see if that makes a difference. But I think the 1850 fans might be the kicker :D
. . .and if your not running with comps or clamps, this is when your tubing will blow off your fittings in a right hurry. :yepp:Quote:
Originally Posted by gmat
Really interested to see how this turns out. I suspect gfx temps will be higher again, but those fermis run 95c+ stock so not a problem. CPU temp im hoping you get a good 5c lower, maybe more.
Been wrong plenty of times before though!
Seems as if you have your loop the wrong way, If anything your cpu should be running way cooler then your gpu's, try splitting half the rig..
Haven't had a chance to move anything around as yet as I've been away for 12 days.Quote:
Originally Posted by Johnny87au
By "splitting half the rig" do you mean run two separate loops?
I had two loops before but the way the bayres is setup you have a shared res of heated water whatever way you do it.
I've got the Aquagrafx twinConnect multiSLi connector now so will try PiLSy's suggestions with the new connector and see if that helps :)
Thanks again for all your ideas folks.
Just been playing with the twinConnect, and one of the things it offers with this motherboard (with 7 pcie2.0 slots) is the ability to move the bottom card up a slot :D meaning I could do inlet at the top and outlet at the bottom while staying parallel.
Now all I have to do is drain it all, rearrange everything and find out it didn't make any difference :)
Will check back here later dude, GL :).
Dunno what im more curious about, cpu temp on smaller rad and colder air or GPU temps at low flow. Bit nervous too, I could end up looking a right plank here :lol:.
LOL don't worry, all advice is considered, Its not like its going to burst into flames, my only misgiving is the concept that the entire loop stabilises at a certain temperature after a while no matter what order you have your components in, well thats the theory. And I'm loathe to question it, however what I don't know is how effective the placement of radiators and fans to cool the liquid at certain points in the loop will have an effect. It shouldn't have any effect theoretically (practically I don't have enough time to experiment every which way) but my thick skull keeps saying that radiator placement and flow strength should somehow have an effect.
I know I'm going mad, I've been up 36 hours straight and am probably on caffeine overdose :) Time to rip the loop apart! or maybe not... :hitself:
The water goes through your loop way too fast to build up any significant temperature differences. Go with the easiest mount options, and easiest tubing.
Yes but the further you go below 1gpm the more noticeable that difference is.
As gmat inferred though the difference is small (likely 2c at most in a low flow large loop) so you can get close to that by taking the shortest route with tubing.
If I move my rads to after the cpu instead of infront I see a 1-2c increase across all cores but a 1c load temp drop on both GPUs. 1c is nothing for gfx cards so went with cpu before rads :).
I'm thinking of going back to a dual loop now. With the 240 on the CPU and the 360 and 120 on the GPU's.
As I'm currently using the XSPC bayres with two DDC 18V pumps, but its the old version without the separator in the res for dual loops. So I thought I'd just swap it out for the separated dual loop version.
But what I'm concerned about is that with just the little bit of perspex separating the two parts of the res, there must be some significant heat transfer through the material, or would it be so much less transfer of heat compared to the single res I have now that it would make enough of a difference.
Problem is I don't really have room for the two DDC's and two separate res's in this setup with the configuration I have now, and I *really* need to get this finished before it takes over my entire life :)
Biffa, how much room do you have in the drive bays?
5 bays.
From the top:
1. empty (but radiator comes pretty close)
2-3 XPSC dual pump res
4 DVD
5 Fan controller
Hmm. . .OK, if you get (or make) something like this and then you could mount 2 setups like this side by side on it.Quote:
Originally Posted by Biffa
Height is approx 5⅛" tall (50mm res)
http://a.imageshack.us/img248/8553/koolance50mm.jpg
or, if you can swap (or lose?) the DVD with the fan controller.
Height is approx 6-5/16" tall (80mm res)
http://a.imageshack.us/img638/4306/koolance80mm.jpg
Another option aside from Koolance would be a pair of EK DDC X-RES 100 which would basically do the same thing.