Watercooling nighmare after dropping down to 10mm OD tubing!! (high temps)

Hi folks,

Wondering if you can shed any light on this situation, I've included some pictures of my new setup for reference.

Here's my current setup:

Hardware:

CPU: Core 2 Duo E6600 @ 3.4Ghz
Mobo: Asus P5N32-E Sli Plus
RAM: 2GB GeiL PC6400 4-4-4-12
Graphics: x2 8800 GTX in Sli
PSU: Zalman ZM1000-HP (1000w)

Cooling:

Pump: Laing D5 Vario (with G1/2" to G1/4" adaptor)
Radiator: XSPC R120-D (dual 120mm fan radiator)
Reservoir: XSPC 150mm passive aluminium res
Tubing: 10mm OD/8mm ID
Connectors: x9 10/8 compression (straight), x5 10/8 compression (90 degree elbow)
CPU block: Danger Den TDX Intel 775 (high flow)
Northbridge: EK Northbridge Asus 1 (high flow)
Mosfets: x2 EK Mosfet Asus 1 (high flow)
Liquid: Fesser 1 UV Blue


This is a upgrade from what I had before (a 2nd 8800 GTX, Northbridge block, mosfet blocks and a better PSU). In my previous setup I also used an EK-FC8800GTX full-cover waterblock (I've since returned that GFX card back to air-cooling for now) and 1/2" ID tubing without any mobo blocks.

These were the temps:

CPU: 43c idle, 47-50c load
GFX: 45c idle, 50c load

(ambient = 23c)

During my upgrade, I removed the watercooling GTX from it's slot (still part of the cooling loop) and installed the newly arrived (and air-cooling) 8800 GTX into the mobo's lower slot.

My CPU temp dropped to 30 idle, 34c under load!!!!! (ambient 23c again...)

Couldn't believe the difference for taking the 8800 GTX out of the loop so I decided to return to aircooling for the graphics cards and concentrate on CPU/chipset cooling for more stable clocks for these parts instead.

With the extra blocks required and the proximity to the CPU, I thought dropping down to a smaller tubing size (10mm OD/8mm ID) might make the logistics a lot easier and after reading serveral articles stating that the tubing size makes little difference to temperatures, I decided to convert to smaller tubing.

Big mistake, CPU temps have sored to 56c idle, 58c load!!! (ambient = 23c)

Something isn't quite right, I have put this down to:

1) The CPU block doesn't have enough tension for the mounting and therefore isn't seating properly as I had to adjust the tension during installation to not make the motherboard warp too much (especially the lower-right hand screw....see pics)

2) The smaller tubing/ x5 90 degree bends (especially coming out of the exit between the CPU toward the Northbridge) is killing the flow rate

3) The blocks I've purchased are designed for a high-flow rate and they aren't getting this currently with the bends/smaller tubing

Well they say pictures are worth a 1000 words so here's the rig:
















Many thanks in advance for any advice!!

I don't think the 8mm ID would impact the temps. It seems to be the amount of 90 degrees bends which is causing your issues. Try to remove them all and use the tubing only to connect all. Try to plan out so no 2 connections is too close as to cause kinking.

Hi Xilikon, thanks for the quick reply.

I thought the same; maybe the heat is building up by that first 90 degree bend coming out of the CPU?

TerminatorUK said:

Hi Xilikon, thanks for the quick reply.

I thought the same; maybe the heat is building up by that first 90 degree bend coming out of the CPU?

Click to expand...

That's very possible and by restricting the flow a lot, it will stagnate in the CPU block, causing temps issues.

It could also be a floating mount. Try remounting the CPU block once or twice, before you dismantle all of your tubing. Honestly, I think this was going on before as well- a 15* drop in temperature seems a bit much for just a GPU, even if it is an 8800GTX.

The elbows are playing a part, but not that big. I never considered the TDX as requiring high flow rates. It certainly benefits from them, but I don't think it's a necessity.

I'm a fan of big tubes, but there's no way a diameter change should impact temps that much. The difference should be marginal at best (as long as you've got a decent pump - which you do).

I'm with mavalpha: if you suspect a mounting problem already, that big temp difference probably confirms it. Remounting will probably take care of the issue.

On a side note, I really don't like the way everything is in series there. The MOSFET blocks really don't need the same amount of flow that the CPU gets, and they add a heck of a lot of restriction to the loop. I'd put the MOSFETs and the NB all together in one lobe that's parallel to the CPU to get more flow to the CPU. Basically:

pump - Y - (CPU on one side of Y), (NB and MOSFETs one other side of Y) - Y - GPU - ...

Then again I can't blame you if you don't want to redo the whole loop; a simple remount will probably take care of 80-95% of the problem.

It states above that by removing the GTX (unplugging) temps were 30c, that sounds like a fine temp.

My question would stay around the flow and the radiator/fans. Do you feel hot air comign out the radiator? I'm thinking radiator saturation makes the most sense, that coupled with poor flow = heat issues.

Sunin said:

It states above that by removing the GTX (unplugging) temps were 30c, that sounds like a fine temp.

My question would stay around the flow and the radiator/fans. Do you feel hot air comign out the radiator? I'm thinking radiator saturation makes the most sense, that coupled with poor flow = heat issues.

Click to expand...

Doh! I wasn't even paying attention to the rad size. I was thinking that anyone running an SLI rig would be running at least a triple. That's a big load to put on a double. I'd start digging in some of the calculators available in the WC forums on xtremesystems.org to see exactly what to expect from that setup.

Thanks for all the advice guys...really helpful!

Here's an update:

CPU block

I've re-mounted the CPU block and made sure there is a decent tension on all the springs but this doesn't make any difference.

I think the problem is going to be the circuit (I love the Y-piece parallel circuit by the way nonlnear)

I don't think this is going to be down to the CPU mount because the temps start at a nice 28c and steadily increase to to a peak of about 58c.

My theroy is the exit from the CPU is so tight, it isn't allowing the heat to escape through this route.

Radiator

The air through this is being pulled (not pushed) with a built in metal shroud/rubber mounting. This is mounted at the front of the case internally and the air through this does not feel too hot.

Loop

Just to clarify the loop is currently:

Pump - CPU - NB - Mosfet - Mosfet - Radiator - Reservoir - Pump

(graphics is all air cooled)

Voltages

Forgot to mention, the CPU and NB voltages are increased to the following:

CPU = 1.475v
NB = 1.45v

Heat in the loop

Would the ammount of heat generated by the NB and two Mosfets be greater than that generated by a single 8800 GTX?

TerminatorUK, sorry to go off topic but I love your style with the compression fittings. I'm a watercooling noob and want nothing but compression fittings for piece of mind and security.
Please tell me how you attached compression fittings to that pump.
It looks like the same Swiftech/Laing pump I am going to buy but I had no idea you could attach compression fittings to that particular pump.

Hi Xerran...thanks for the nice comments.....I've seen neat setups using compression fittings before which inspired me to this build.

I was using 1/2" ID tubing before (great flow, nice and secure with jubilee clips but a bit bulky) and have used 10/8 push-fit connectors (never again!!! - leak city....killed an 6800 series nVidia card!!) but these connections are easy to use and very secure (I only used PTFE tape on the pump connectors).

It took be a A LOT of searching to find the fitting that allowed me to convert the Laing D5 into 10/8 compression but I finally found it from the following place (if you're in the UK)...strangely as a fitting for a water chiller but fits absoutely perfectly!:

http://specialtech.co.uk/spshop/customer/product.php?productid=2462&cat=575&page=1

(looks like I got the final two connectors as this is out of stock now!)

Well to give you my results from last night.

42c idle cpu
43c idle gpu
32c system temp reading

I have a Thermochill PA 120.3, D5 pump, EK NB, EK SB, EK 8800GTX full, and I forget the cpu cooler but it is a high flow item.

I have no fittings in the system it is straight tubing from piece to piece.

I think my temps are a bit high at idle I expected them to be around. But then again 32c is awefully hot for the system. I mean that is 90F... Hmm wonder what that is about. I should check ambient in my condo before doing the next set of tests.

I never got to load the system as I am running down a few issues atm!

Personally I think your killing your flow with the mosfet coolers and that is causing your issue. Try pulling them out of the loop and test it.

I really think it is a flow issue. How much pressure is coming out of the return into your resevoir? it shoudl be forcefully hitting the back wall of your res if you have good flow!

I'll second and third the flow issues. I think that's what is killing your temps is low flow.

With that little variance in idle/load, sounds a lot more like a really bad mount.Those 90* elbows aren't helping anything, but they're definitely not enough to explain that big of a difference, especially with a D5 (have you tried changing it to 5?).

Awesome advice guys...many thanks.

I think I might swap out the mosfets with some passively cooled ones, make a loop does not use 90 bends and just connects the CPU/chipset blocks up and see what that does to the temps.

I'll let you know what happens after I make the changes and might post some new pictures

Your pump is more than enough to push through the loop. I guess you can try removing one block at a time until everything stablizes.

Hi guys!

Great news...temps have dropped a lot; a big thank you to everyone that offered advise on this.

I took the mosfets out of the loop completely (put some passive heatsinks on for now that aren't getting too warm at the moment), removed the 4x 90o bends around the CPU area away and just have a simple system that goes to the cpu, a generous u-bend between the CPU and chipset and a long section from the chipset.

I also re-mounted the CPU again but I don't think this was the problem as it had a nice, even square of AS5 on the base and seemed secure enough already; however, I did place a small blob of AS5 right in the centre of the CPU heatspeader just in case it was a little thin (I do normally spread the AS5 out myself when I place it on the heatspreader).

I didn't get a lot of time to test my results but I ran 3DMark 2006 three times and these were my results:

CPU = E6600 @ 3.4Ghz (1.475v)
Ambient = 23oC
Idle temp = 34oC
Load temp = 38oC

I need to do more extensive testing but a drop of 20oC is amazing so thank you all for the great advise!

btw - when I run Orthos, the CPU temp (occording to nVidia monitor) jumps upto around 50oC (immediately)....and when you stop it immediately jumps back to 38oC....that can't be correct surely?

Actually what are you measuring your cpu temps with? Get coretemp or something like that.

Next that might not be that surprising depending on the voltage your running.

My C2D E6600 at 3.6Ghz at 1.4v hits a max of 62c, Idles at 42c. At stock voltage and speeds of 2.4Ghz and 1.235v Idles at 34c, fully loads at 45c, if memory serves me. My ambient was 22c yesterday and system was at 35c.

That's good temps considering the cpu block is pretty old and not optimized for CPU with a IHS. I'm glad the changes helped a lot

For the mosfets, it's worth watercooling only if you are pushing for extreme overclocking.

@ Surin - nice overclock....I had my E6600 hit 3.6ghz and get into windows for the first time; evidence that the chipset cooling is doing it's job However - I cannot get it to run Orthos for more than 5 mins without crashing (and that was at 1.55v vCore...don't really want to go any higher).

What is your northbridge/FSB voltage set to? I currently have mine at 1.45v

I was only using the nVidia monitoring tool that works with my 680i (well hybred 650i with dual 16x pci-e lanes chipset) to monitor the temperature....I'm guessing coretemp is more accurate.

@ Xilikon - I think you are right about the mosets...they are only slightly warm even under load without much airflow using passive cooling....think they'll be going on eBay soon then hehe

I wasn't aware the TDX wans't very good with an IHS cpu such as a C2D....do you have any recommendations in this respect?

I'm gonna do a good session of gaming tonight (which I haven't done for over 2 weeks with all this messing around!!!).

I'll report back some more realistic load figures - probably gonna be an oven in the case after 3 hours of Crysis with 2 8800 GTX's on stock cooling!!!

Here's hoping the temps stay nice and chilly

I'd say the following is causing the problem:



Even though you've solved it now, this may still be an issue with your latest setup.

By having that bit of tubing so short and close you're probably making the waterblock sit unevenly on the CPU. You should always use as long as tubes as possible when connecting to a waterblock so as to ensure that they don't create any unnecessary pulling and tugging on the block.

I'd say re-arrange your tubing accordingly to fix it. Just ensure every piece of tubing going into whatever block has enough slack to move fairly freely and easily and this'll ensure that it won't be tugging on the block unduly.

Spine, did you bother to read the latest reply from the OP ???

TerminatorUK said:

@ Surin - nice overclock....I had my E6600 hit 3.6ghz and get into windows for the first time; evidence that the chipset cooling is doing it's job However - I cannot get it to run Orthos for more than 5 mins without crashing (and that was at 1.55v vCore...don't really want to go any higher).

What is your northbridge/FSB voltage set to? I currently have mine at 1.45v

I was only using the nVidia monitoring tool that works with my 680i (well hybred 650i with dual 16x pci-e lanes chipset) to monitor the temperature....I'm guessing coretemp is more accurate.

Click to expand...

1. Get Coretemp program, it shows each individual core. Under the overclocking area of this forum there is a guide to OCn the C2Q and C2D. It is a decent guide with good utilities! NVmonitor runs about 5c hotter then the actual cores!

2. My NB is still at stock 1.4v

3. Yeah I've been told that although the cores can take above 1.6v, that is the limit for a daily overclock! I would guess, in the middle or researching this, that your max core temp should be 70ish celsius. The cores should never hit over 80c imho. Even though they are set to hit a max of 85c. Really no reason to push it like that for a daily OC.

4. I have not puhed my vcore that high yet. 1.45v produced 67c core temps for me, and if 1.4 produced 62c, I would think 1.5v would be 72c, and 1.55v would be 77c, which I would never want to have as my daily.

Whatever you do have fun with it and good luck!

Xilikon said:

Spine, did you bother to read the latest reply from the OP ???

Click to expand...

Yes, I read the whole thread, but I thought I'd highlight the original issue since I see it alot in people's watercooling systems. People seem to focus on waterblock design and spends loads on an amazing block only to then do a poor mount making it all redundant. You should never underestimate the stress the tubing is going to apply to the waterblock mount. It's vital the tubing is free to move.

Just thought I'd make that point, that's all.

@ Spine - you are definitely right....that part that you highlighted (really well I might add!! ) I think was definitely the issue; I have a long bend in it's place now that is fairly slack and seems to have really made a difference; i'll post some pics on the weekend so you can see. I had to scrap the idea of having two adjacent 90o bends between the mosfet coolers in the upper left hand corner as it was litterally pushing the blocks off their mounts!

@ Surin - I think you're right....I had to push the vCore to over 1.55v to just get the chip @ 3.6ghz to get into windows and do a 5-minute Orthos test before crashing....I'm quite happy with the performance at 3.4ghz overall. However, even at that voltage and ghz, my temps were around the low to mid 40's mark.

Unfortunately I didn't get to extensively test the setup last night but again it seemed to sit at 38oC with nvmonitor under load so I'm really happy overall

Cheers chaps

Congrats man! I know I'm enjoying my system. I think I am going to have to lap the cpu IHS and the CPU Waterblock. The temps are just not what I expected. I really had anticpated a solid 5-8c lower than what I am getting and I used MX-2 in the correct fashion. I may try tightening it some and then if that doesn't work I'm going for the lapping.

Enjoy your new cooler stable system!

Thanks for all your help guys....I can't get the CPU to go over 38oC whilst playing games even for long sessions so I'm very happy overall

I did test using Coretemp and Orthos for 2 hours and it reached 54oC but then dropped straight away to 38oC again once finished so I'm pleased with that.

Nice and stable