I desperately need help from the system-building experts. I had wanted, for once in my life, to build a no-compromises gaming machine. Inspired largely by Tom's Hardware's March 2008 System Builder Marathon, I figured I'd try to set up a liquid-cooled computer. Who knows? Maybe I'd even play with overclocking a little.

But now that I have it all put together, I'm confused and dismayed. I seem to be getting temperatures *worse* than people are getting with stock CPU fans. I don't really know how to approach fixing the system. I don't know what things to investigate. Plus, I don't really understand the issues of core vs. CPU temperatures, bad sensors, and why different tools show different numbers. If anyone can offer guidance, I'd be grateful.

When I did a casual test yesterday, my computer was at 35C Idle (I think), and 65C after an hour at full load from Prime95. Those temperatures are from PC Probe II - an Asus-provided program. Ambient temperature was presumably 25C, but I didn't actually measure it then.

Here are my results from today:
Ambient: 25.3C
Idle
CPU Temp (PC Probe II): 46C
Cores (Core Temp): 49C 51C 55C 54C
Cores (Real Temp): 39C 41C 45C 44C
After 20 minutes of Prime95
CPU Temp (PC Probe II): 67C
Cores (Core Temp): 63C 65C 68C 64C
Cores (Real Temp): 53C 55C 57C 54C

The programs with which I tested were:
PC Probe II
Core Temp 0.99
Real Temp 2.60

So either I imagined yesterday's idle temp of 35C, or I messed something up between then and now that caused dramatic decrease in cooling. But even yesterday's numbers weren't good enough. The temps given by Real Temp don't actually look so bad, but I have no idea what they mean.

I wasn't smart enough to install a flow monitor into my system, but it feels like there's liquid flow when I touch the tubes. I see waves in the resevoir window. And the air coming from the radiator feels warmer than the air from the case fan.

The CPU block's connectors are 1/4" ID barbs, so that's a bit of a choke-point. And in a few places the tubing curves a bit more than it would like. I've got one elbow fitting, too, because there was no room for anything else between GPU-block2 and the power supply.

Pieces:
CPU: 3.0GHz Quad-core Intel Core 2 Extreme QX9650 (Yorkfield)
Motherboard: Asus Striker II Extreme nForce 790i
RAM: 4x 2GB Patriot Viper DDR3
Video: 2x BFG GeForce 9800GX2
OS: Vista 64-bit SP1
HDD: 2x Western Digital Caviar SE WD800AAJS in RAID-0
Case: Silverstone Temjin TJ10 Tower
Power: 1000W Silverstone Strider ST1000
Pump+Resevoir: Koolance RP-1000BK (5.25" drive bay mounted)
Radiator: Swiftech MCR220-QP (2x120mm)
CPU-block: Thermaltake W1
GPU-blocks: Danger Den DD9800GX2
Tubing: 3/8" inner-diameter clear (generic from hardware store)
Coolant: Feser One Non-conductive fluid

Loop: Resevoir -> Pump -> Radiator -> CPU -> "Fusion block" (mobo) -> GPU1 -> GPU2 -> Resevoir

All BIOS settings are factory-default except that the RAID controller is enabled.

The pump's temperature sensors - which I have taped to the side of the CPU block, 2nd GPU block, and radiator out port - show 36C, 34C, 32C when idle, respectively.

That's all the info I can think to provide. Like I said, I'd be grateful if anyone can help me figure out how to fix my setup.

Actually those temps are not bad considering the gpu and quad. to get better temps up the fan speeds across radiator, or get a bigger radiator. the other option is increase the capacity of the system. try to double the amount of water being held in the system, and it will help keep temps down.

Ok...where to begin here because this is going to be a long one...

First of all, you want to set up any cooling loop to maximize the flowrate. That being said, there are alot of factors that can go into influencing the effectiveness of your loop at various points of its path. It is pretty much a given that the CPU is going to be the most important component (that is, if you include it in your loop and decide not to cool it via convection). So, you want your CPU waterblock to recieve the best possible flowrate with the maximum "head" potential.

For purposes of definition, “Head” refers to the height of a vertical column of water. This is the maximum height that a pump can sustain any semblance of flow rate before it loses its capabilities. For purposes of an example we'll use a pump rated at 317gph with an imaginery "head" of 36 inches. At 0 inches of height you will have maximum flow rate and the pressure will be zero. Pressure is a measure of resistance to flow. Thus, at its initial discharge, at 0 height, the pump experiences its least resistance and generates its fullest flow. As the height in the cooling loop increases, the resistance to flow increases and the flow rate decreases. Earlier we said that our pump had a "head" of 36 inches. The closer the pump gets to its "36 inch" height, the less flow is generated.

So, at 0 height we have 0 pressure and 317gph. At 36 inches we have full pressure and no flow.

Alright, now the one component that is that hinders flowrate in any cooling loop the most is the radiator. Think about it, all the bends and turn in a tightly packed area. Now, I am sure that you are developing your own cooling philosophy but, at no point in the loop is there any better condition for cooling than when the coolant leaves the pump - best flowrate and head potential. So, ideally, you would want the most important component to be the reciever of that condition - the CPU. Placing the rad right after the pump is going to really hinder the flowrate and head initially. You should place the rad on the backend of the loop - that is what it is there for - toi deal with the heat.

Reservoir - pump - CPU waterblock - GPU waterblock - GPU waterblock - rad - back to reservoir.

In this incarnation you have the rad dealing wiht the heat of the waterblocks in a position where it does not rob the flow and head potential as much as if it were at the beginning of the loop. Also, you have the reservoir coming immediately after it and it will also help in heat dissipation (however slightly).

Ok, rad placement is one thing, onto the next thing - the 1/4 ID fittings in a 3/8 ID loop. That doesn't help matters at all. In fact, that CPU waterblock is going to be a point of contention for the loop, much like the rad and how it affects the flowrate. Really now, if you can change those fittings to 3/8 ID than you will be doing yourself (and the cooling loop) huge favors.

That bay reservoir/pump kit includes fittings for 1/4 ID, 3/8 ID and 1/2 ID according to the Koolance website. In researching the kit firther I found that the pump in the bay housing – the PMP 400 – is the same thing as the Swiftech MCP350. It has a flowrate of 135 gal/hr and is ideally set up for 3/8 ID loops. That is not to say that it can’t be used in other ID systems, just that its natural ID is 3/8. Now, 135 GPH is ok, not as powerful when compared to the more popular pumps like the MCP 655/D5 Laing (which operates at 317GPH). So you need to think “maximize” the pumps potential as the whole loop revolves around the pump to begin with. So, 3/8 ID is where it operates best.

Alright, now even if you do move the rad and correct the fittings issue and have 3/8 ID maintained throughout the loop you need to alter your philosophy a bit about the heat potential you are dealing with. If you o’clock at all – and I mean even in the slightest, you are going to task a cooling loop that has a single rad (albeit, a dual 120mm). You have the heat from the CPU waterblock that is going to travel down the loop to the first GPU waterblock. Now that waterblock not only has to deal with the heat fromt eh CPU but the heat from the GPU as well. Further down the loop you have the second GPU that gets to catch all kinds of hell from the previous two blocks besides its own heat. That is asking a lot – and that is, initially, without any o’clocking. Now, think about eh heat buildup f you do o’clock ANY of those components.

If it is at all possible here, could you add a second dual 120mm rad? I hated even bringing that up because that would be asking the pump to do a lot but it would so improve the heat dissipation characteristics of your loop.

Reservoir – pump – CPU waterblock – rad – GPU waterblock – GPU waterblock – rad – back to reservoir.

In this loop the rads are at strategic positions to dissipate heat before it gets to another block.

Now, if a second rad is not feasible than get a triple 120mm rad at least – seeing as how you do plan on o’clocking – and position it like I had mentioned earlier.

Thanks for the feedback, guys.

Actually, the problem seems to have gone away after a BIOS update (to 706 from 406, I think). Go figure. So now I'm idling at 30C and running full-load at 52C. (Of course, we'll see what it's like when I make the graphics cards work.)

So I guess either: a) The old BIOS wasn't reporting temperatures correctly; b) the old BIOS was causing the voltage to be too high; or c) the old readings were right and the new BIOS is reporting temperatures incorrectly. I prefer to think it's a or b.

Thanks again for the help.

If you want to increase the cooling effectiveness of a dual 120 radiator, set up 4 fans in a push-pull config on both sides of the radiator. You can still run the fans at a fairly slow spin, but it will push-pull more air through the radiator than with just 2 fans running at a higher rate.

Also, be aware that your pump will dump heat into your cooling loop. You have active cooling from your radiator, passive cooling from your reservoir, but every other step is actively dumping heat into your loop. Your GPUs will want to run hotter than your CPU, so you will be dumping more heat into the back of your loop. Splitting up the radiator cooling would help a little, but so would increasing the effectiveness of you dual rad that you already have.

Don;t worry too much about your pump. Unimpinged flow rate has proven to be less of a factor in a cooling loop as previously thought. The more you can remove inpingement (anything that constricts flow rate) the better off you will be. THere is a certain amount of impingement that is unavoidable (such as water blocks and radiators), but if you can eliminate any sharp angles, or replace any blocks that reduce you loop inner diameter significanty ( I do believe Phreejak was mentioning this when he was talking about a water block that reduced from 3/8" ID to 1/4" ID) and you will see your flow rate increase. You really only need about 30-40 gph when it is all said and done.

As far as reporting temps from you bios update, then it sounds like that was an issue that your BIOS flash cleared up. Just monitor the temperature of your loop, and compare it to the reported temps from your on-board sensors. As long as you aren't suffering a boil-off and your reported temps stay under 70° C, then you should be sitting pretty.

Here is my Homemade PC
Click below link
http://www.tomshardware.com/forum/ [...] g#t1761518

thank you so much for the advice about water cooling. It was very helpful.

I believe Phreejak covered that quite well.

+2

Crazy system. Spending most of my college education understanding the flow of water, I will also agree with phreejak.

I'd at least follow his loop setup and fitting diameter advice, even if you're temps are fine.

This seems to be a problem. Is there any reason you are running 1/4" on this block and 3/8" on the rest of the system? As for the bends in your tubing, if you can't run good tubing like Tygon or something, you can at least run the coils around the tubes or just use zip ties and make those tubes a little more round.