Need a Recommendation on a H2O setup for Dual NV6800's (SLI) and an AMD FX-55

[Josh_Hayes]

Well, some parts arrived







Now I play the waiting game... Memory and Hard Drive should arrive next week. Then I gotta save for the other GPU and H2O Cooling (Ran out of funds too soon! )

-Josh

[Sc0rian]

oo sexy! lovely lovely lovely.. hey it's xmas... Cough;0.

[Cranky]

fx55 grrrrrrrr u lucky dog.....

[mtbdrew]

Originally Posted by pauldenton

the actual amount of liquid in the system will only affect the time it takes to reach equilibrium, not the equilibrium temp. the heat sources add heat at (virtually) the same rate irrespective of water temperature - but the rad(s) ability to shed heat is proportional to the temperature differential between the water and the air
http://thermal-management-testing.com/ThermoChill.htm

so the water temperature will rise until it reaches the point at which the rad sheds heat as fast as the pump/blocks add it.... the more liquid in the system the longer it will take, but we're talking minutes unless you have an enormous res.

once it reaches equilibrium it doesn't matter whether 1 litre circulates 4 times a minute or 4 litres once a minute - a smaller volume heated and cooled more often is just as good at moving heat from one place to another.

Thanks for the mental work out, you are making me dust the out corners I have not used in a while.

One point you are still missing though. You are making the assuption of the heat being applied at one time. However, it is being applied by five different sources which cause a 5:1 ration of heating to cooling. This of course is a simplified way to look at it. To do this accurately you would have to measure the time it take for a set amount of water to travel through the rad and each heat source. But for this example lets just say it takes 1 second for a millileter of water to travel through each device. In this case for every second of cooling you get 5 seconds of heating. The calculation is made moore complicated because each heat source is a different value but you get the general idea. To go back to my original state he will either need a monster rad or two seperate rads. Either methods should increase the amount of time the water gets cooled. Ideally you would want a 1:1 ratio to bring your calculation back into line.

[Tanaban]

Very nice stuff you have there so far. I live in RR and am actually planning to buy exactly what you have minus the case... (custom building some thing for my self...) Mind if I come over and borrow *cough*steal*cough* your stuff?

[pauldenton]

Originally Posted by mtbdrew

Thanks for the mental work out, you are making me dust the out corners I have not used in a while.

One point you are still missing though. You are making the assuption of the heat being applied at one time. However, it is being applied by five different sources which cause a 5:1 ration of heating to cooling.

err no - at equilibrium the cooling = the sum of all the heating (by definition - otherwise it's not at equilibrium.)

Originally Posted by mtbdrew

This of course is a simplified way to look at it. To do this accurately you would have to measure the time it take for a set amount of water to travel through the rad and each heat source.

that's the flow rate - in our example 4 litres/minute or 66.667 ml/second

Originally Posted by mtbdrew

But for this example lets just say it takes 1 second for a millileter of water to travel through each device. In this case for every second of cooling you get 5 seconds of heating.

ahh this is where you are misunderstanding i think... we are dealing with a continuous flow of water - in every minute we have 60 seconds of heating per heatsource and 60 seconds of cooling per rad - at no time is there ever less, or more water in a component.

Originally Posted by mtbdrew

The calculation is made moore complicated because each heat source is a different value but you get the general idea. To go back to my original state he will either need a monster rad or two seperate rads. Either methods should increase the amount of time the water gets cooled. Ideally you would want a 1:1 ratio to bring your calculation back into line.

what you have to take on board is that a rad's cooling ability is proportional to the temperature differential between water and air.
when you add additional heat sources then the equilibrium water temperature will rise until it's cooling ability matches the heat. to use the figures we used earlier:
CPU block = 70w
pump = 10w
total heat input 80w
say the rad/fans can cool this at a water temperature of 3C above the air flowing through it. if we add the gpu blocks (and imagine for a moment that the resulting flow drop has no effect on the rad's efficiency) then we have:
CPU block = 70w
pump = 10w
2xGPU blocks = 60W
total heat input 140W
now the equilibrium point would be +3Cx140/80 i.e 5.25C - so our water temperature has risen by 2.25C as a result of added heat input.
but at this (new) equilibrium the same calculations about temperature differentials apply-
the coolest water is directly after the rad, then we have the following rises after components:
pump 10/279= approx 0.035C
CPU block 70/279 = approx 0.25C
GPU bbocks 30/279 = approx 0.1075C each
total 0.5C...

so the position in the loop makes very little difference (unless you either have very low flow, or huge heat-eg TECs)

you could add a second identical rad, and (given the same flow airflow/air temp) then you'd halve the water:air differential (ie the +5.25C would drop to +2.625C) but the rises over the hot components wouldn't change (the ones over the rad would halve due to the temp. drop - each rad would now give an approx 0.25C drop not a 0.5C one) - the payoff from the extra/better rad(s) is lower water temps throughout: the rise over a component is unchanged(at teh same flow)

[Josh_Hayes]

So back to the topic...

How about this setup?

Reservoir    Rad1   GPU1   CPU    Pump
+--------+   +--+   +--+   +--+   +--+
|     -> =>>>=  =>>>=  =>>>=  =>>>=  =>>>+
|        |   +--+   +--+   +--+   +--+   |
|        |                               |
|     <- =<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<+
|        |                               |
|        |   +--+   +--+   +--+   +--+   |
|     -> =>>>=  =>>>=  =>>>=  =>>>=  =>>>+
+--------+   +--+   +--+   +--+   +--+
             Rad2   GPU2   Chps   Pump

2 Loops Sharing a Single Reservoir using Y-Connector

-or-

+<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<+
|                               |
|   Rad1   GPU1   CPU    Pump   |
|   +--+   +--+   +--+   +--+   |
+>>>=  =>>>=  =>>>=  =>>>=  =>>>+
    +--+   +--+   +--+   +--+

    +--+   +--+   +--+   +--+
+>>>=  =>>>=  =>>>=  =>>>=  =>>>+
|   +--+   +--+   +--+   +--+   |
|   Rad2   GPU2   Chps   Pump   |
|                               |
+<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<+

2 Individual Loops with no Reservoir

[Josh_Hayes]

Originally Posted by pauldenton

Is there any need for a chipset block on these boards??

Yes, the Chipset gets scalding hot! I'm not impressed with the Heatsink and Fan (8000RPM!) they chose to use on this motherboard, but it had to fit in a tight space... With GPU Waterblocks, I can use just about any Chipset block.

[Blackeagle]

If the NB is as hot as all that then I'd suggest routing the two loops a bit differant.

Loop #1

Rad=>CPU=>NB=>pump

Loop #2

Rad=>GPU1=>GPU2=>pump

My reasoning is that this would more nearly equalize heat load.

If CPU is @ 100w & NB @25w (guess) = 125w total

If GPU's are both at 50w = 100w total

I would prefer this as if you combine the CPU with a GPU you then get 150w vs 75w on the loop with GPU + NB.......should improve both the CPU temps and balance the GPU temps better.

Blocks

CPU => TDX with #4 jet installed & brass top to avoid corrosion risk, or.....Storm G4 or G5 both of which have non corrosive tops.

GPU => Silverprop Fusion with brass top X2 or Danger Den maze 4 if they now have brass topped versions.

Chipset=> Danger Den Maze 4 with brass top, or Fusion NB block with brass top.

RAD

As someone else posted get the Black Ice Pro which is 1" thick for best performance with those fans. Those are low noise fans you have, very nice, but also low pressure which means a 1" rad will perform better for you.

PUMPs

The DDC 12 pump is very good, low/no noise, and in dual loop can handle with ease the resistence to the above components. Very nice small pumps with high head to flow ratio.

50Z pump from Cooltechnia is also a very good pump with high head and should give a bit higher flows in such a dual loop system, but is also a bit larger to locate two of them and they are also a bit higher dba though they are not loud.

Both will do very well, the DDC is lower noise, the 50Z is higher flow/performance. No big advantage in favor to one or the other, just a judgement for you to make as to which is of more value, noise or flow rate.

If you use any aluminum topped blocks in the loop be sure to add a anti corrosive to the water. I myself do not care for use of aluminum in a loop, but many do so without problems provided they use anti corrosives. Anti corrosives degrade your coolant's performance but they are a must with aluminum in a loop.

You'll have a really really kick butt rig, congrats to you.

[Josh_Hayes]

Well everything is here and she's running HOT!

I will take pics later...

My GPU's get up to 92*C each! I think its time to be introduced to Watercooling...

I've decided that the case I bought is not really going to work out for watercooling so I'm selling it and just purchased a Lian-Li PC-7077B case from Newegg.com. Its a lot like a stacker, but with the standard Lian-Li styling.

Do they make a 120mm Bore bit for drills to cut through aluminium?

[Josh_Hayes]

System Components:

Lian-Li PC7077B full-tower case
OCZ Powerstream 520watt power supply
Asus A8N-SLI Socket-939 motherboard
AMD Athlon FX-55 CPU
1GB OCZ PC3200 Platinum Rev2 memory
2 BFG Nvidia 6800 GT PCI-E graphics cards
SoundBlaster Audigy 2 sound card
Western Digital 74GB Raptor hard drive
Plextor Premium CD-RW drive

This case features one 120mm exhaust port on the back and one 120mm intake port on the front.

I am planning on using the back port for one of the radiators and I plan on drilling another port on the top of the case for the other.

My Plan:


Two independent loops, one for CPU and Chipset and one for both GPU's.

The pumps will be hung upside-down from the roof of the case (Is this okay for the pumps?). There will be a T-line up at the top between the pumps and radiators leading to a Fillport in the top of the case. Fans will be attached to the radiators using 1" shrouds and will be pulling air through the radiators and exhausting it outside. I will eventually add fans to the inner-side of the radiators once I can find another source for Globe fans. I will be using 1/2" ID tubing throughout.

Watercooling Components:

Danger Den TDX CPU block - brass top - #4 nozzle	$  61.94
Danger Den Maze4 GPU block - brass top (2)		$ 102.90
Danger Den Maze4 Chipset block - brass top		$  38.50
AquaXtreme 50Z-DC12-T pump (2)			$ 178.30
Black Ice Pro radiator and shroud (2)		$  99.00
Danger Den fillport for T-line (2)			$  23.90
Polypropylene T-connector (2)			$   2.00
HydrX coolant (2)					$   6.58
30 black plastic tubing clamps			$  15.00
10 feet of Tygon R3603 1/2" ID - 11/16" OD tubing	$  20.50
Total before shipping				$ 548.62

I plan on running two independent loops in the following configuration:

Loop 1

+<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<+
|                                      |
|   Res    Pump   Rad    CPU    Chps   |
|   +--+   +--+   +--+   +--+   +--+   |
+>>>=  =>>>=  =>>>=  =>>>=  =>>>=  =>>>+
    +--+   +--+   +--+   +--+   +--+


Loop 2

    +--+   +--+   +--+   +--+   +--+
+>>>=  =>>>=  =>>>=  =>>>=  =>>>=  =>>>+
|   +--+   +--+   +--+   +--+   +--+   |
|   Res    Pump   Rad    GPU1   GPU2   |
|                                      |
+<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<+

This will be my first attempt at watercooling. Please let me know if there are any changes I need to make as I want to make sure I do it right the first time. Also let me know if I've forgotten anything.

Thanks for the help!

-Josh

[Niatross]

With the kind of money you seem to have I'd look into phase change cooling

[Butcher]

92C? Woah, mine on run at 55-60C at load...

[MaxxxRacer]

You can get a hole saw bit from home depot that is slightly smaller than the 120mm hole.. its great for drilling the holes.. the only thing is that you need a larger drill, the bit and the bit mounter... the bit itself is 20 bucks and the mounting system is 35 bucks on top of that. Most power drills wont fit it though. You need one of hte big ones that are generaly powered on a cord. They are for concrete drilling and other stuff...

in other words, if you have a dremel its just cheaper lol.. me I have access to all the equipment so its great for doing the holes though...


If you doing two rads, i suggest one at hte top (the backplate whereh the psu is flips over so that the psu isnt at the top and the fan is) attached to the ceiling, and have the other rad on the floor of the case.. if you leave the feet on the case it will provdie enough airflow so the rad will not be starved for air.

[Josh_Hayes]

Originally Posted by Niatross

With the kind of money you seem to have I'd look into phase change cooling

:shrug: AFAIK Phase-change cooling is only for single devices and is WAY over my budget of $500...

[Josh_Hayes]

Here's a picture of the two BFG Nvidia 6800 GT cards and the CPU and Chipset coolers...

[The GrRingo]

Nice! Currently I have an AMD 64 FX-55 sitting here awaiting use, along with a MSI K8N Neo4 Platinum SLI mobo . I'm going to purchase 2 XFX 6800 GT cards and a WD 74GB Raptor.

I actually have a Storm/G4 waterblock i'm going to be using for my FX-55 and also a Silverprop Fussion HL i'm going to TRY to mount to the 6800 GT (one for now as I need anotehr block.) Anyone mount one? Work good?

I'm actually going to test the setup with a dual 120mm Black Ice Xtreme II radiator and two Criticool Waterplant res's (I used two just for looks, I know it adds to the demise of my system. ) For now I'll just have teh Storm/G4 and Silverprop blocks on the same loop with a mcp650 pump.

I wonder how 2 Silverprop blocks and the Storm/G4 will react with only one dual 120 rad and the mcp650. Has ANYONE hooked up the Silverprop to the 6800's yet? Just curious as I cant find anyone who can give their opinion.

------------------------------------
George
Ex Airborne Infantryman (U.S. Army)

(Yeah, would have had this done earlier if my engine didnt crap out. Droping a Jap engine in my '92 Prelude though.. there goes my money for my compuer )[IMG]My Comp[/IMG]

[MaxxxRacer]

grRingo: the HL's are actually lower in height than the stock 6800ultra cooler so they will work just fine. This is coming from the owner of silverprop btw. I have seen them used and they work fine, so no worries.

btw with two of those blocks a storm and two rads i suggest you get another mcp650 or a Aquaextreme50Z. the 50Z would be the better choice of the two, but bother are good.

[The GrRingo]

Hey MaxxRacer, thanks man! I apreciate it. Well see, unfortunatly I do not have the money to get the second 6800 GT so I'm going to put both the Storm AND Silverprop on the same loop with only ONE (dual 120mm) radiator and ONE pump. Once I get the money to splurge on another 6800 GT, I'll get a second Silverprop and hook it up to the loop. I'm going to test to see if I can overclock AND run (safely):

-1x Storm/G4
-2x Silverprop HL
-1x Black Ice Xtreme II (Dual 120mm)
-1x mcp650 (actually the DangerDen 12v, same thing)

Im actually worried about the chipset as I dont know how good the stock fan/heatsink is on it. On the MSI Neo4 SLI, when I plug the second 6800 GT in, it will go OVER the stock chipset fan/heatsink leaving me no room to replace it with a better one.

[MaxxxRacer]

Dont worry about the NB. It gets really hot but it doesnt get unstable even when oced.

[Josh_Hayes]

New case just arrived and came bearing surprises!



A 120mm blowhole in the top which saves me $40 worth of modding hardware!



Here's a pic of my current setup before the new case and watercooling.

[Josh_Hayes]

It was suggested in another forum that I should split up the GPU loop where each GPU gets their own line as follows:



Are there any benefits to this? I see logic in their suggestion as each GPU would get equally cool water, but wouldn't this negatively affect my flowrate? And would it even matter as the water going into the radiator would be the same. The only difference would be that the second GPU would get cooler water than it would when in a series.

[Metzen]

I use the SwiftTech 240x120mm rad + MCP350 + MCW50 (x2) + MCW6000 + a Zalman (GWB1) GPU cooler for the chipset all on a single loop. The hose totals 8' in length amoung all the blocks. The setup goes > MCRES-525 > Pump > CPU > GPU1 > GPU2 > Chipset > Rad > Res. Cooling the Rad is 2 - 14dBa SlienX fans. I believe performance is pretty good (going for quiet). Hottest measured CPU temp is 53-55C (measured in software) while running CPUBurn. Hotest recorded GPU temp is ~60C measured by the Nvidia control panel (running FarCry in a Window @ 1280 x 1024 with the "desktop" resolution @ 1600 x 1200). I suspect these temps are not accurate to any real degree as the GPU's get to their "peak" of around 60C the CPU still reads 53C. Unless that's to be expected? :shrug:

Idling both the CPU and GPU's read the same @ ~40-45C. I'll have to double check that though.

I admit I'm a pretty big noob at this and got a lot of help building this PC but I'm pretty satisifed.

System configuration is as follows:

6800U x 2
A8N-SLI Deluxe
FX-55 @ 2.7GHz w/default voltages.

I'll measure out the LPH tomorrow for interests sake.

As you've read above, there are big temp advantages for the GPU's in watercooling that I've observed. On air-cooling my 6800U's were hitting an ambient temp (measured directly opposite of the die IIRC) of ~100-105C with a GPU temp of 70-80C. With water-cooling ambient temp and die temp stay the same. GPU2 always seems to be hotter for some reason, around 5-10C air-cooled and 1-5C water-cooled (I don't know how Nvidia numbers their cards though, so I don't know which is top and which is bottom).

I'm looking to change the fans to something a little more powerful. I have heard SilenX has a bad reputation for giving out "optimistic" specs and trying to find a comparable fan confirms this. I don't know which fan would be best though to replace it. I've been reading good things about Globe or the Nexus (too ugly for me) but they're specs are loud (starting at 24dBa IIRC) and, unfortunately, SPCR has not reviewed any of the SilenX fans in their sound chamber nor recorded their CFM flow so it seems impossible to get a fair comparison going.

Edit
I chose the Zalman GWB1 waterblock because it's the only water-cooling block that fits with 6800 Ultra's. The MCW50 is too big, maybe even too big for the GT. The A8NSLI has PITA holes that your block has to line up against to mount it. Though the GWB1 waterblock works, I had to invert it's mounting gear to get it to fit correctly because the holes are so close together. IIRC, I had trouble with the MCW50 block because of this.

[Josh_Hayes]

**Hopefully Final Selections**



Watercooling Components:

Swiftech MCW6002 CPU Waterblock
2 x Danger Den Maze4 GPU Waterblocks with Brass tops
Danger Den Maze4 Chipset Waterblock with Brass top
2 x D-Tek Pro-120 HeaterCore with Shroud
2 x Fan Shrouds for read of Heatercore
2 x Y-Connectors for GPU loop
2 x 5 1/4" Bay Reservoir 
2 x AquaXtreme 50Z-DC12-T Pumps 
30 black plastic tubing Clamps 
15 feet of Tygon R3603 1/2" ID - 11/16" OD Tubing
2 x 2oz Bottles of HydrX Additive

Questions:

1. Are brass tops recommend over Acetal ones? Why?

2. Will 15 feet of tubing be sufficient for my setup?

3. Will 2 x 2oz bottles of HydrX be enough for two loops?

4. Am I missing anything?

5. Can you see any problems with my final selections?

[Josh_Hayes]

I just completed the first mod on my new case...

Here's something for those of you that might think my OCZ PowerStream 520 power supply wont be able to supply enough power for my rig...



I just added my PC Power and Cooling Silencer 400 ATX power supply as an internal secondary PSU. I gutted an old DVD-ROM and reinforced the interior of the housing. I then drilled holes in the side of the PSU and the top of the DVD-ROM. I have three threaded bolts with rubber grommits holding it in place. The mounting is solid as a rock, its not going anywhere!

I'm planning on running the power cable out one of the PCI bracket slots or out the bottom or back of the case.

-Josh