WC Critique/Tweak Advice, 1st WC

[warriorpoet]

I was setting up my closed GPU loop when I realized it wouldn't work with the CPU air cooler I had installed (Scythe SCNJ-1000) . Long story short I have more parts for the rest of the system in the mail. Based upon years (literally) of research and the budget/ goals I have in mind, here's what I came up with:

Goals first (in order):
1. setup a system that remains enclosed within an Antec P180 upper compartment
2. create a system that rivals my air-cooled rig's sound levels, pre-x1900 (fairly quiet)
3. improve performance of stock x1900xt cooler and equal temps of SCNJ-1000 (CPU)

Cooled System Componentry:
CPU: Opteron 170 @ 2.4 GHz (29-48c)
GPU: ATi x1900xt (45-89c)
System Temp (on-board diode): 31-37c
Room Temp: 20-25c

Ordered WC Stuff:
CPU Block: D-Tek MP-05 SP LE (432 pin impingement) w/ additional restrictor plates for tweaking
GPU Block: Swiftech MCW60
Pump: Swiftech MCP350 (Laing DDC) + MCRES-1000 reservoir
Radiators: 2x Swiftech MCR120 Quiet Power

Can any of you out there give me some idea how this will perform, or recommend flow order tweaks? My current plan for flow is:
res./ pump->radiator (front)->GPU->radiator (rear)->cpu->res./ pump

Thanks!

[mx]

This thread would get much more replys over at XS or [H].

I'd buy a plexi top for the DDC to be able to mount the inlet barb on top for increased flow. If you dont want the bigger plexi top or don't want to pay for it you could just modify the original DDC top for increased flow.

[warriorpoet]

Originally Posted by mx

This thread would get much more replys over at XS or [H].

I'd buy a plexi top for the DDC to be able to mount the inlet barb on top for increased flow. If you dont want the bigger plexi top or don't want to pay for it you could just modify the original DDC top for increased flow.

Thanks, will do.

Where do I find that plexi top?

[mx]

I brought mine from from alphacool.com. You can find them in lots of stores here in Europe, I don't know where you can buy them in the states though.

[Long Haired Git]

I am assuming 1.5m of 3/8" tubing and all barbs are converted to 3/8".
I am using a MCW55 as I have not plugged in the numbers for the MCW60 yet.
Assuming 90 CPU Watts and 50 GPU watts, and all temperatures are based on RoboTech's die simulator.

Anyway, as an approximation, I get 2.5 LPM flow rate.
Fans @ 5v
Water temp at exit of 2nd rad: 8.87 dCaa (degrees Celsius above ambient)
CPU temp: 22.47 dCaa
GPU temp excluding TIM: 9.4 dCaa

Fans @ 7v
Water temp at exit of 2nd rad: 5.93 dCaa (degrees Celsius above ambient)
CPU temp: 19.52 dCaa
GPU temp excluding TIM: 6.44 dCaa

Fans @ 12v
Water temp at exit of 2nd rad: 3.92 dCaa (degrees Celsius above ambient)
CPU temp: 17.52 dCaa
GPU temp excluding TIM: 4.43 dCaa

Leaving the fans at 7v, and changing the pump to a MCP655 @ 5, you'd only get a half degree or less improvement.

When you assemble, please do the bucket fill flow test and let me know how close I am with my approximations

[warriorpoet]

Originally Posted by Long Haired Git

I am assuming 1.5m of 3/8" tubing and all barbs are converted to 3/8".
I am using a MCW55 as I have not plugged in the numbers for the MCW60 yet.
Assuming 90 CPU Watts and 50 GPU watts, and all temperatures are based on RoboTech's die simulator.

Anyway, as an approximation, I get 2.5 LPM flow rate.
Fans @ 5v
Water temp at exit of 2nd rad: 8.87 dCaa (degrees Celsius above ambient)
CPU temp: 22.47 dCaa
GPU temp excluding TIM: 9.4 dCaa

Fans @ 7v
Water temp at exit of 2nd rad: 5.93 dCaa (degrees Celsius above ambient)
CPU temp: 19.52 dCaa
GPU temp excluding TIM: 6.44 dCaa

Fans @ 12v
Water temp at exit of 2nd rad: 3.92 dCaa (degrees Celsius above ambient)
CPU temp: 17.52 dCaa
GPU temp excluding TIM: 4.43 dCaa

Leaving the fans at 7v, and changing the pump to a MCP655 @ 5, you'd only get a half degree or less improvement.

When you assemble, please do the bucket fill flow test and let me know how close I am with my approximations

Thanks!

Not to sound too n00bish (long-time hardware geek, 1st time WC geek), but what is/ how can I do the "bucket fill flow test", and what will it tell you?

Thank you especially for the MCP655 comparison, I was considering it, but from what I gather it'll be too loud for my purposes.

Two questions:
1. The x1900xt card consumes 125w+ at load, are you sure the GPU gives off only 50w?
2. My CPU, according to an admittedly inaccurate estimation program, consumes ~163w at load (dual-core, 1.5v)

Thanks again, 'git, that's exactly what I was looking for. I'll be sure to help your database any way I can.

[mx]

The bucket fill flow test is done by timing how long it takes to fill a bucket with all components in the loop. This way you can tell what flow rate you have.

[warriorpoet]

Originally Posted by mx

The bucket fill flow test is done by timing how long it takes to fill a bucket with all components in the loop. This way you can tell what flow rate you have.

What size bucket is best? (unwashed American, I'll have to convert Gallons->L)

edit: Seriously though, how do I ensure a consistant water supply for the ~1min. I'll be running the "bucket test"? Can I just add tubing from my res to another bucket full of water? Also, is this an effective way of bleeing bubbles from a new system?

[BGP Spook]

Bubble bleeding in my experience can best be done with time and some aerobics.

Best method I can think of for bucket test:

1. Fill bathtub with a fair amount of clean water.
2. Hook up everything and set it in the water with the return in a bucket. *
3. Get out of bathtub. ( :buzzet: )
4. Turn on pump/start timer.
5. When bucket is full stop timer, turn off pump, drain everything.
6. Divide liters filled by number of seconds it took to fill. (L/sec)
7. Multiply by sixty for liters per minute. ([L/sec]*60 = L/min)

* Bucket has been pre-measured to a convenient level. Say 10 liters, which is about half a 5-gallon bucket, or whatever is convenient and preferably more than 4 liters for accuracy.

[Long Haired Git]

Originally Posted by warriorpoet

Not to sound too n00bish (long-time hardware geek, 1st time WC geek), but what is/ how can I do the "bucket fill flow test", and what will it tell you?

Check out the www.overclockers.com.au WIKI for how to.

Originally Posted by warriorpoet

Two questions:
1. The x1900xt card consumes 125w+ at load, are you sure the GPU gives off only 50w?
2. My CPU, according to an admittedly inaccurate estimation program, consumes ~163w at load (dual-core, 1.5v)

Well, marketing departments like to sound cool. The data sheets I refered to for the Opterons were posted as a link from OCAU's news page a while back. Basically most of their CPUs are 89W items. Now that's standard Mhz and Voltage, so no doubt tweaking will impact.
I struggle to accept that a video card will thus put out more watts as heat from the GPU block (as distinct from energy consumed) as comes out of a dual core Opteron.
Just let me know what wattages to use and I will re-approximate.

[warriorpoet]

Originally Posted by Long Haired Git

Check out the www.overclockers.com.au WIKI for how to.

Thanks, knew I could count on you for good info, 'Git.

Originally Posted by Long Haired Git

Well, marketing departments like to sound cool. The data sheets I refered to for the Opterons were posted as a link from OCAU's news page a while back. Basically most of their CPUs are 89W items. Now that's standard Mhz and Voltage, so no doubt tweaking will impact.
I struggle to accept that a video card will thus put out more watts as heat from the GPU block (as distinct from energy consumed) as comes out of a dual core Opteron.
Just let me know what wattages to use and I will re-approximate.

I trust your figures, but am curious how much OC'n affects temp output and the validity of the high figures being tossed about the 'net.

I realize that VGA power consumption is much different than GPU heat output, but am interested about the process used to isolate heat dump from total wattage used. Considering the card uses 512Mb 1.1ns RAM, the differences are not hard to account for, just difficult to calculate in relation to other current architectures.

[warriorpoet]

Originally Posted by Long Haired Git

I am assuming 1.5m of 3/8" tubing and all barbs are converted to 3/8".
I am using a MCW55 as I have not plugged in the numbers for the MCW60 yet.
Assuming 90 CPU Watts and 50 GPU watts, and all temperatures are based on RoboTech's die simulator.

Anyway, as an approximation, I get 2.5 LPM flow rate.
Fans @ 5v
Water temp at exit of 2nd rad: 8.87 dCaa (degrees Celsius above ambient)
CPU temp: 22.47 dCaa
GPU temp excluding TIM: 9.4 dCaa

Fans @ 7v
Water temp at exit of 2nd rad: 5.93 dCaa (degrees Celsius above ambient)
CPU temp: 19.52 dCaa
GPU temp excluding TIM: 6.44 dCaa

Fans @ 12v
Water temp at exit of 2nd rad: 3.92 dCaa (degrees Celsius above ambient)
CPU temp: 17.52 dCaa
GPU temp excluding TIM: 4.43 dCaa

Leaving the fans at 7v, and changing the pump to a MCP655 @ 5, you'd only get a half degree or less improvement.

When you assemble, please do the bucket fill flow test and let me know how close I am with my approximations

Just got done with my loop and ran a bucket test for you. I came up with 3.1 LPM, wonder what accounts for the difference. Perhaps the Rev.2 MP-05 Accelerator plate does a better job than originally thought or the MCW60 is a SERIOUSLY low resistance block.

In any case, that's probably at least partially responsible for the temps I'm seeing:

Flow order:
Laing DDC1/Swiftech MCP350 (pump)->D-Tek/Cooltechnica MP-05 SP LE(CPU)->Swiftech MCR120 Quiet Power (Rad1)->Swiftech MCW60 (GPU)->Swiftech MCR120 Quiet Power (Rad2)->Swiftech MCRES1000 (res.)-> Pump

Fans:
1 Scythe S-Flex 1200RPM @8v (800-920 RPM)
1 Scythe S-Flex 1600RPM @6v (860-1040 RPM)

Temps (on-board Sensors):
CPU (opteron 170 stock): 40c load (2x Prime95, 1 small, 1 large)
GPU (ATi x1900xt stock): 39c load (ATi Tool + RTHDRIBL)
Room Temp: 26c

So, CPU (load)= room + 14c; GPU (load)= room + 13c

Hmmm...