Fine tuning my CPU & NB watercooling setup

[megatron]

I have added a z-chip NB cooler to my Atlantis icicle and flow rates have dropped noticeably. It may be improved by replacing the kinked link between my NB and CPU cooler. I have been told that y-adapters can help. Would this be worth doing? I believe my CPU temp and system temp have risen after installing the new waterblock.

I do have an unusual setup (pics in sig). I use 10m of 1/2" silicon tubing so I can place my radiator outside and remove a noise source. I have a Eheim 1250 and the flowrate as seen in my resevoir has dropped alot. I will take some pictures of the watercooling system as it is currently and put them on my website soon.

The only options I have are:
1)shorten the length of 1/2"
2)hope the kinked tube being replaced will make all the difference
3)run the watercooling down from the cpu onto the NB instead of going past the slightest heat source first
4)Use Y-adapters and run a channel past the NB
5)Attribute higher CPU temps to heating the water up slightly due to passing the NB first.

Cooling the NB has not given any (may be down to bios I am using 3609 which sirens past 200Mhz) more overclocking potential on my rev 1.0 Epox 8RDA+ @ default Vdd = 1.6v. I will be replacing it with a Rev 2.0 or 2.1 (http://www.aoaforums.com/forum/showpost.php?p=191164&postcount=19) shortly.

[Ruantic]

I think you pretty much have all the right answers in your post, I don't think more fitting will be helpfull, although some folks seem to have good luck running NB cooler in paralell. I say clean it up as much as you can, kinks, short and direct as possible, then see how its performing...

[Player0]

Replace the kink, you have to do that.

Do not go to a Y-adapter parallel system. It will screw things up much worse than they already are.

Your pump is extremely underpowered for trying to do what you are doing with all that tubing. Shorten that tubing, keep the radiator, pump, and CPU at the same height (as much as possible) , and get a larger pump (like a 1060 or a MAG 800 or something).

[megatron]

http://www.liquidninjas.com/photopost/showphoto.php?photo=659

Shows the flowrate with a comparison shot before. In fact with a 3/8" barb attached inside the res the jet flies clean across and hits the res wall the other side with force (20cm), used to happen with 1/2" barb hence diffuser for bubbles. I have looked at a bigger pump but do not think it a prudent way to spend my money. I like the length of my tubing. There is only really one way I can setup my watercooling due to the size of the res, pump and rad. It gives excellent temperatures in winter (check out the 8'C water temp in the 2nd linked pic)! I will sort the kink which you can see in my gallery and have the tubing to do so now.

I believe a y-adapter would suit my requirements the best since I was happy will the flowrate before adding the z-chip. The 10m of tubing by itself does not present a problem for the eheim 1250.

I will see how it looks when I replace the kink, and then decide if making a parallel length past the z-chip would be worth it.

Do you think running down from the cpu to the z-chip will be less water restistant than going up the board from z-chip to CPU block?

[megatron]

You may like to see the range of Eheim pumps, the 1060 has been replaced with the 1260 and there are some higher flow rate pumps with 0.1m less head hight than the 1060 @ 3.7m:

http://www.aquatics-online.co.uk/acatalog/
Look under Power Heads & Universal Water Pumps .....

[Kabooka]

uummmm before you go buying another - or more pumps address that kink issue. Then do a test.

http://www.liquidninjas.com/photopost/data/500/368Kink_-_z-chip_-_Atlantis_Icicle-med.jpg?7284

[Player0]

Well you said you had 10m of tubing, that is a lot of friction for a pump to deal with. There is no doubt in my mind that you are running much less GPH than you should be. Small fountain pumps like the Eheim aren't meant for really long pipe runs. More GPH should improve your temps. The fact that temps rose when adding that NB block indicates to me that your flow is too restricted. The heat from the NB itself simply couldnt increase the temperatures.

Splitting the flow is a big newbie mistake. Everyone, and I mean everyone, beleives that having one waterblock (GPU, NB, RAM) before the CPU will increase the CPU temperatures. This is common misconception in watercooling. It is wrong. Experienced watercoolers have to explain this fact to newbies more than you can imagine. With PROPER flow rates, the temeprature measurement at any point in a water system, say befoer and after the cpu, or before and after the GPU or radiator will only be a fraction of a degree. That is because with a proper flow rate, the water in the system is acting as a single entity, it will have its own average temeprature, and the temperature will be fairly balanced throughout the whole system. This is good.

If you measure the water temps at two different points in the system, and you have a difference of more than a degree, or maybe two, then you dont have a high enough flow rate. Flow rate balancing is extremely important.

Which brings me to point two. Parallel plumbing. It's very difficult to do (just look at any diesel engine fuel rail). Each lenght of tubing on each leg of the split must be the exact same lenght, the exact same height, and the exact same diameter, or you will get unequal flow through each branch of the split.

http://www.liquidninjas.com/photopost/data/6026/1p0_rearview.jpg?7107

Beleive me, I've tried this. Ive done it very carefully, and it still didnt work. See those two pelt waterblocks? One would run at 30c, the other at 80c because I could never get the flow to balance. This is bad for peltiers, and worse for your CPU.

My first thought is ditch the NB block. They don't work either. A good HSF is all you need for this, even when seriosuly overclockign. Ive chilled them, Ive put pelts on them, ive had every heatsink i can think of on those NB, and nothing even seems to make a difference. If you ever get pelts for the CPU and the GPU, then yuo cant use a NB waterblock anyway. It will just be in the way. If you dont plan on using that...at least dont go to a Y split. You cant make that work well enough.

Now. If you do Y split, make sure to downsize the splits! Use pi * r ^2 to figgure out the inner volume of your current tubes, split that by two, and find a tube with THAT inner radious. No, it wont match exactly. Maybe split down to 3/8" ID for the CPU, and 1/4" ID for the NB and keep another 1/4" ID loop in there going nowhere, or to the GPU, to keep the pressure balanced. You cant split the 1/2" ID line in to two 1/2" ID lines because then youll have a terrible time with getting air out of that system (been there too).

Anyway, I hope you understand that this is the voice of experience. You cant do anything that I havent tried yet. Parallel splits will give you a serious migrane. It can be done, but its beyond difficult to get it right.

[megatron]

I have attempted to run a pelt twice. First it failed from not protecting the exposed socket 462 which made frost and then water when I engaged the TEC and 20sec later the CPU. Second attempt I believe the pelt PSU gave out, since I was assured the pelt was not at fault even though I got a refund. The PSU has been in use for 1-2a load for the rad fans for a number of months now as opposed to 21a load for the TEC.

Concerning y-split. I believe you are describing a situation where I run the CPU and north block in parallel as someone else suggested (and I believe you have tried y-spliting for dual CPU pelts?). This is not my intention. Jess who owned OCPC which has now gone under - has advised people in the past to put the main flow through the CPU block and then to keep flow rate high run the north block (and optional GPU block together in series, I think) in parallel with a free running tube. This is what I said all along and I am not talking about running CPU and NB in parallel since the CPU is far greater a heat load than the NB.

Regarding temperatures. The water temperature I believe is the same temperature as before ~ 30'c. The CPU temperature has risen about 5'c judging by overnight temperatures assuming the ambient temp overnight has been constant. So this means for sure that flowrate has been harmed. My cpu block as said in a review likes high flowrates.

There is no reason to expect to get equal flowrates when running a y-split as described above, and no reason to want it to be so. I have no problem with bubbles in my rig since the res outlet is the highest point in my system and having a res with a water drop automatically purges any air. I have seen y-split 1/2" all round and y-split 1/2" to two 3/8". I am not worried about bubbles in the y-split but I am concerned with getting the best flowrate for the z-chip.

I just thought of an advantage for running up the board from NB to CPU - more chance that air will self purge.

[Player0]

Not very good luck with TECs there, but they are still a very good system, you just need to be sure to build the system properly, condensation proofing and of course having the proper spec'ed PSU to run it with. Dont give up just yet :)

Ah, yeah, I thought you were spliting the flow between the NB and CPU, sorry for confusing that with what you are doing. And I have no idea what you are doing. I can't get a clear picture in my head of what you will be splitting with that Y splitter from your description. Maybe a diagram would help? It sounds like you want to maintain flow rate by letting some of the water bypass the NB all together. Oh, maybe thats what your doing.

Okay. I see what your getting at. Yeah, that can be smart. Your NB might only get 20% flow, but it should increase the flow over the entire system. However, Y adapters reduce flow themselves, and you will be adding more tubing in to the mix. You may not end up with much more flow than you already have. Its smart though, worth a shot I guess. But I dont think its a cure-all.

See how temps go with the Y and getting rid of that kink (that might be more of a problem than the NB block itself :)) and go from there :)

Maybe buy a second Eheim instead of replacing the big one. Put one pump outside or something. Just a thought...Koolance does it :)

[megatron]

---pump---rad---cpu block-----y-split----z-chip---y-join---res---pump
.................................................. ....\---------/...............................

Dots are nothing, just for positioning of lines.

My psu is rated at 30a + 10a + 6a for three different outputs respectively.

[Player0]

Yeah, thats what I thought. Again, splitting that flow in to a Y and recombining it looks like it will increase flow, but...the resistance added by the extra Ys and tubing themselves may offset that benefit. By how much, I dont know. If your flow is 100GPH, and you pick up 5GPH from that mod, will that be enough, or do you really need 150GPH to really make this work well. I've had the Eheim 1250 pumps.. they are a strong little pump with about 3ft of head. You're pushing quite a few pounds of water with that pump. Probably 8-12 pounds with all that tubing.

If adding the NB block is causing all these problems...maybe its not the NB blocks fault, maybe your flow just wasnt adequate to begin with, thats all.

[megatron]

Ok I fixed the kink, ;D. I got these readings from Epox USDM:

01/07/2003 22:17:43 - Power FAN:0, Chassis FAN:0, CPU FAN:1638, CPU Temp:28, System Temp:27, Vdd:1600, VCore:1696, Vdimm:2752, +5V:5053, +12V:12403, -12V:12116, -5V:5350, Vbat:3216, 5VSB:5042, 1057094263
02/07/2003 00:39:26 - Power FAN:0, Chassis FAN:0, CPU FAN:1577, CPU Temp:21, System Temp:20, Vdd:1600, VCore:1696, Vdimm:2752, +5V:5026, +12V:12403, -12V:12198, -5V:5401, Vbat:3184, 5VSB:5042, 1057102766
02/07/2003 00:44:26 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:25, System Temp:21, Vdd:1600, VCore:1696, Vdimm:2752, +5V:5053, +12V:12342, -12V:12116, -5V:5350, Vbat:3200, 5VSB:5042, 1057103066
02/07/2003 00:49:26 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:27, System Temp:22, Vdd:1600, VCore:1696, Vdimm:2752, +5V:5053, +12V:12403, -12V:12116, -5V:5350, Vbat:3120, 5VSB:5042, 1057103366
02/07/2003 00:54:26 - Power FAN:0, Chassis FAN:0, CPU FAN:1622, CPU Temp:28, System Temp:22, Vdd:1600, VCore:1696, Vdimm:2752, +5V:5053, +12V:12403, -12V:12034, -5V:5350, Vbat:3120, 5VSB:5018, 1057103666
02/07/2003 00:59:26 - Power FAN:0, Chassis FAN:0, CPU FAN:1622, CPU Temp:28, System Temp:22, Vdd:1600, VCore:1696, Vdimm:2752, +5V:5053, +12V:12342, -12V:12034, -5V:5350, Vbat:3200, 5VSB:5042, 1057103966

Looking at the CPU temps. The first reading is before sorting the kink. After that is showing the water is heating up from no load to equilibrium temperature for the cpu idle. I still think these temperatures are reading higher than before the z-chip; for whatever reason.

I have some pics to illustrate the change in flow rate going from no z-chip > z-chip & kink > z-chip no kink. Also just a general idea of the flowrate when at the end of 10m the water drops into my res through a standard 1/2" barb:

http://www.liquidninjas.com/photopost/showgallery.php?ppuser=368
Enjoy the pics! :D

It may be a possibility that I could sell my 1250 - they are always in demand, and get a 1260 or higher! :D They cost double of the 1250's pricetag. What other high power pumps are available? Damn I have such a cheap ghetto rad. (http://www.doomiii.pwp.blueyonder.co.uk/FrontalRad.jpg)

[Player0]

The pump is going to add ALOT of heat in to the water by itself because you are putting a lot of pressure. Turn off the CPU and bypass the radiator, and you will see the parasitic pump heat rising the water temperature, much quicker than usual because of all that plumbing. the pump heats up the water cause its being strained, and the water heats itself up because of the extra friction inside the tube (yes, water will have friction with itself).

Im sure you can sell the 1250. Its a high demand item these days. You can look in to DannerMag drives. The 1260 will work too, Ive used a couple of them. Can you move to 5/8" plumbing? tubing is cheap enough. So are barbs.

[megatron]

5/8" plumbing when all blocks only ever go to 1/2" sounds...pointless. I am water cooled already, I don't fancy alot of cash outlay for something I already got. My flow rate is up now to 90-95% of what I had before. I could do 2500Mhz on 22'c water, that is not really a lame cooling system. You gotta remember I have the advantage of outside air cooled radiator and can have up to 14.8v through my fans (which is bloody loud for people outside).

Also it would cost pennies to upgrade to a parallel loop of the z-chip. For me getting the most bang for buck is what its all about. :alienbird :nuke:

[Player0]

Hi Megatron,

Actually, I have charts here from KMS and also references from BillA that support 5/8" plumbing as really making an improvement. It reduces tube wall friction, and it will actually force the accelleration of water down in to the waterblock (like putting your finger over a hose). I never would have thought a year ago that having different size plumbing in a system would be worth anything. It turns out that different size plumbing can make a BIG performance difference because of the accelleration and decelleration effects it has on waters. A good rule of thumb is: wide tubing, small waterblock channels/barbs. Yep, serious :)

Yep, best bang for buck is always a good thing, and the parallel will work. And the system will continue to work fine irregardless of what else you upgrade. But, there are definite upgrade paths open to you, and thats all im saying :)

[megatron]

Update: I swivelled the cpu block round so that the inlet was lower than the outlet. Coincidence maybe but I got better temperatures and also I am running 205Mhz FSB and vCore 1.7v instead of 1.65 with mutliplier the same at X10. I raised the vCore simply to negate the possibility of my cpu falling over. Maybe there is better layer of artic ceramic paste.

02/07/2003 15:58:17 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:28, System Temp:23, Vdd:1616, VCore:1680, Vdimm:2752, +5V:5026, +12V:12403, -12V:12198, -5V:5401, Vbat:3200, 5VSB:5018, 1057157897
02/07/2003 16:05:07 - Power FAN:0, Chassis FAN:0, CPU FAN:1622, CPU Temp:28, System Temp:23, Vdd:1600, VCore:1744, Vdimm:2768, +5V:5026, +12V:12403, -12V:12198, -5V:5401, Vbat:3200, 5VSB:5018, 1057158307
02/07/2003 17:05:26 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:34, System Temp:24, Vdd:1600, VCore:1728, Vdimm:2752, +5V:4999, +12V:12464, -12V:12198, -5V:5350, Vbat:3200, 5VSB:5018, 1057161926
02/07/2003 17:19:34 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:35, System Temp:25, Vdd:1600, VCore:1744, Vdimm:2752, +5V:5026, +12V:12403, -12V:12281, -5V:5401, Vbat:3184, 5VSB:5018, 1057162774
02/07/2003 18:19:34 - Power FAN:0, Chassis FAN:0, CPU FAN:1622, CPU Temp:32, System Temp:25, Vdd:1600, VCore:1744, Vdimm:2752, +5V:5053, +12V:12403, -12V:12116, -5V:5350, Vbat:3200, 5VSB:5018, 1057166374
02/07/2003 19:19:35 - Power FAN:0, Chassis FAN:0, CPU FAN:1591, CPU Temp:28, System Temp:23, Vdd:1616, VCore:1744, Vdimm:2752, +5V:5053, +12V:12403, -12V:12116, -5V:5350, Vbat:3200, 5VSB:5042, 1057169975
02/07/2003 20:13:30 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:26, System Temp:21, Vdd:1600, VCore:1744, Vdimm:2752, +5V:5053, +12V:12342, -12V:12116, -5V:5350, Vbat:3200, 5VSB:5018, 1057173210
02/07/2003 20:17:03 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:26, System Temp:21, Vdd:1600, VCore:1744, Vdimm:2752, +5V:5053, +12V:12403, -12V:12116, -5V:5350, Vbat:3200, 5VSB:5042, 1057173423
02/07/2003 20:18:03 - Power FAN:0, Chassis FAN:0, CPU FAN:1607, CPU Temp:26, System Temp:21, Vdd:1600, VCore:1744, Vdimm:2752, +5V:5053, +12V:12403, -12V:12116, -5V:5350, Vbat:3200, 5VSB:5042, 1057173483

I will see if I can get past 205 FSB with my stock Vdd of 1.6v.

[megatron]

Originally posted by megatron
---pump---rad---cpu block-----y-split----z-chip---y-join---res---pump
.................................................. ....\---------/...............................

Dots are nothing, just for positioning of lines.

I did this and have increased the external case tubing size to 3/4" and used a bulkhead to go to 1/2" inside the case. I will be making tube length optimisations and cutting my case so the weight of the connectors/tub/water can be held by the side panel. I am having no trouble with my Y-split and my flowrate is better than it even was without changing the pump even! Eheim 1250. I bought some tygon and could replace the pump outlet tubing with a step up adapter. The tygon will work well in the case with some relatively shart turns. It is as hot today as its going to get and temps are looking good considering I have had to remove two fans from the rad to see if I can repair the one not working. Otherwise have two full depth fan shrouds for my remaining 60mm fans. The are outside in all weather so it hardly surprising they break sometimes. I will rotate the existing ones so the wires are at the bottom for drips and seal the PCB to ttry to prevent losing more. I will get one big one if any more die.

No pics since I won't have a camera for a couple of weeks.

[megatron]

BTW y-splitters rock! They facilitate higher flowrates without needlessly upgrading your pump.