Project Stupid ( seems appropriate )
For some reason water cooling seemed like a good idea, this describes the road I took. I welcome all feedback and thoughts...I have spent a lot of time haunting your forums, surfing and collecting information and generally forming my own hypothesis about how all this would work.
The beginner information here is great, there are other sources. Like most hobby activities everyone has an opinion regardless of their experience....
Question #1 - Why am I doing this ? ( good to know if you have to explain to your wife. )
My answer(s):
1) I like to over-clock and I can get more extreme with better cooling than air only.
2) I don't want my PC to sound like a vacuum cleaner. ( I built a couple of these ).
3) I am generally bored with normal PC builds and need something different.
4) Because I can and I want to. ( very important )
Question #2 - What are the projects goals ?
My answer(s):
1) With normal ( slightly oc'd ) I want the PC to be quiet(er) than my air-cooled rigs.
2) With heavy OC and fully loaded I want to maintain TJ delta of 30-deg. C. or better.
3) I want a stable 24/7 moderate OC'd rig.
4) Oh yeah, I want both these GTX580's in SLI before they are obsolete. ( 3-slot air cooled ).
Problem #1 - Doing it right the first time.
Complicated greatly by a general lack of hard data.
martinsliquidlab.org, skinneelabs.com and effizienzgurus.de
These great sources of hard data. Unfortunately it is beyond scope for anyone one person or site to have cataloged and tested everything that is available. New stuff comes out all the time.
Heat Loads - good information available, much of the OC'd data not exactly applicable to every use.
CPU [ i7-950 ] TDP is 130W
GTX580 TDP is 244W ( 488W pair )
Minimum cooling is for 374W with a single GPU and 618W for a pair at normal clocking. When OC'ing temps are not linear, 30% OC seems to generate about 4x the heat.
My rule.
So how do we cool this thing ? Which radiator(s) ?
This seems to be a real common question and I sympathize with those asking the question. It seems the more information get find the more questions there are. Lets face it, there are a lot of variables involved. Deciding this only took about a month of research.
120.1 are generally capable of dissipating 100 to 130W of heat.
My rule.
[ this depends on type and design of the radiator, fans used, water flow and so forth ]
General rule #1 - more fins per inch require more air flow.
General rule #2 - thicker cores require more air flow.
With this in mind a dual loop seems the best choice.
I don't want heat and flow problems from the CPU or GPU influencing the other, easier to leave them isolated.
Fluid flow is a key element in the cooling system. Need to move a lot of coolant effeciently to help the water blocks and radiators work as intended. Don't mind the pressure drop through the water blocks, I just don't want to see it anywhere else. Most radiator and water block data demonstrates ( to me ) that nominal performance is around 1.5 GPM. This is where most of the testing seems to occur in any case.
Head pressure is the pumps ability to perform useful work ( over come restriction / loads ). Pressure is a function of flow and restriction. With no flow you can get pretty good pressure out of many things, not much use in a cooling system.
Tubing size ? Yeah I read the comparison articles. 3/8-ID performs pretty close to 1/2-ID.
Cost difference ? Not much if you buy the stuff at the hardward store.
3/8-ID ( 1.09-sq/in cross sectional )
1/2-ID ( 1.94-sq/in cross sectional )
1/2-ID has about 90% more cross sectional area. Can't argue about this, works with exhaust and it works with water. Fuild isn't compressable so it is more important. If everything else is the same 1/2-ID flows more with less resistance than 3/8-ID. If you can make it fit, use it.
Water blocks, OMG! Lots of good choices for CPU's ( lots of lesser performing choices also. ) What is best? Ability to transfer heat into the water at a given flow rate with a minimum of pressure drop.
Keep the water moving.
Reserviors ? I have come to look at the cooling system like a dry-sump setup in a race car. A big reservior is a good thing. ( IMHO ). I understand that everything will eventually arrive at equilibrium within about 2-deg. no mater how much fluid is in the system. In this case I believe I am reducing the work for the water by providing more of it.
So with these thoughts in mind I made the following choices:
Loop #1 ( CPU )
EK-Supremecy Water Block ( works for the next build also )
1x Hardware Labs Black Ice SR-1 360
3x Silverstone Air Penetrator AP121 Fans
1x MCP35X2 Pump
1x EK-MultiOpt 400mm Resv.
Loop #2 ( GPUs )
EK-FC580 ( only full cover block I could find, stupid video cards )
EK-FC Link / Parallel SLI Bridge ( stupid video card & water block, water ports on one side only )
2x Hardware Labs Black Ice SR-1 360 ( series )
6x Silverstone Air Penetrator AP121 Fans
1x MCP35X2 Pump
1x EK-MultiOpt 400mm Resv.
1/2-Fatbox Barb fittings
1/2x5/8 Tygon Flexable/Reinforced Tubing
Threaded hose clamps
Distilled water coolant.
2x bigNG controllers (1) Pumps (1) Fans ( not fully tame yet, needs more work )
Big Frigging Case from mountainmods.com. Extended Ascension.
A little bit over kill ( ok a lot, there is a ton of room in this thing ).
Tech tip -
When verifying the case height is acceptable ( fits under the desk ), check the depth also.
( yes - the case is deeper than expected <sigh> forgot about that ) big mother of a case.
Wife told me I had built an end-table.
----------------------------------------------------------------------------------------------------------------------------------------
Results
This CPU has always run hot, I didn't expect water cooling to change this tendancy and it hasn't.
With air cooling it has always idled around 42-deg C. ( core 1 ). at normal clocking.
OC'd at 4.00Ghz on water it idles at 40-deg C.
A couple of hours running Prime95 Large FFT's max tempature is 62-deg. C.
I'm pretty pleased with this.
Next build ( soon ) we will see about the CPU temp issue.
Single card GTX580 on air idled at 38-deg. C. Stressing this card and 70-deg. C is pretty normal.
Single card on water idled at 28-deg C., stressing it we maxed at 38-deg. C.
OC'd the single carg, idle at 31-deg. C, stressing it we hit 39-deg. C.
Thinking another radiator on the CPU loop might be interesting...
GTX580 2xSLI parallel flow cooling OC'd. Primary card idles at 31-deg C and secondary at 33-deg C.
Stressed up we hit 42-deg C. on both cards.
Again I am pretty happy with this.
Baseline is right around 10-deg C. of ambient. My target is 5-deg. C.
Measurement is suspect, I need a direct water temp measurement before I decide on the next change.
I am happy about this also - new computer goodies are fun.
I will either add another radiator to the CPU loop or increase the airflow.
My choice of fans is suspect, they seem to perform well and they are quiet. Only moving an estimated 120-cfm through each radiator. The radiators are optimzed for low airflow, there is room to improve. this amount of air is about 50% of Martins test setup.
Summary
Have I achieved a satisfactory result ? Yes.
Room for improvement ? Absolutely.
Did I have fun doing it ? You bet!
Thanks to everyone here that knowingly ( or unknowingly ) contributed. Your insight, experience, knowledge and stories have been exceptionally helpful.
For some reason water cooling seemed like a good idea, this describes the road I took. I welcome all feedback and thoughts...I have spent a lot of time haunting your forums, surfing and collecting information and generally forming my own hypothesis about how all this would work.
The beginner information here is great, there are other sources. Like most hobby activities everyone has an opinion regardless of their experience....
Question #1 - Why am I doing this ? ( good to know if you have to explain to your wife. )
My answer(s):
1) I like to over-clock and I can get more extreme with better cooling than air only.
2) I don't want my PC to sound like a vacuum cleaner. ( I built a couple of these ).
3) I am generally bored with normal PC builds and need something different.
4) Because I can and I want to. ( very important )
Question #2 - What are the projects goals ?
My answer(s):
1) With normal ( slightly oc'd ) I want the PC to be quiet(er) than my air-cooled rigs.
2) With heavy OC and fully loaded I want to maintain TJ delta of 30-deg. C. or better.
3) I want a stable 24/7 moderate OC'd rig.
4) Oh yeah, I want both these GTX580's in SLI before they are obsolete. ( 3-slot air cooled ).
Problem #1 - Doing it right the first time.
Complicated greatly by a general lack of hard data.
martinsliquidlab.org, skinneelabs.com and effizienzgurus.de
These great sources of hard data. Unfortunately it is beyond scope for anyone one person or site to have cataloged and tested everything that is available. New stuff comes out all the time.
Heat Loads - good information available, much of the OC'd data not exactly applicable to every use.
CPU [ i7-950 ] TDP is 130W
GTX580 TDP is 244W ( 488W pair )
Minimum cooling is for 374W with a single GPU and 618W for a pair at normal clocking. When OC'ing temps are not linear, 30% OC seems to generate about 4x the heat.
My rule.
So how do we cool this thing ? Which radiator(s) ?
This seems to be a real common question and I sympathize with those asking the question. It seems the more information get find the more questions there are. Lets face it, there are a lot of variables involved. Deciding this only took about a month of research.
120.1 are generally capable of dissipating 100 to 130W of heat.
My rule.
[ this depends on type and design of the radiator, fans used, water flow and so forth ]
General rule #1 - more fins per inch require more air flow.
General rule #2 - thicker cores require more air flow.
With this in mind a dual loop seems the best choice.
I don't want heat and flow problems from the CPU or GPU influencing the other, easier to leave them isolated.
Fluid flow is a key element in the cooling system. Need to move a lot of coolant effeciently to help the water blocks and radiators work as intended. Don't mind the pressure drop through the water blocks, I just don't want to see it anywhere else. Most radiator and water block data demonstrates ( to me ) that nominal performance is around 1.5 GPM. This is where most of the testing seems to occur in any case.
Head pressure is the pumps ability to perform useful work ( over come restriction / loads ). Pressure is a function of flow and restriction. With no flow you can get pretty good pressure out of many things, not much use in a cooling system.
Tubing size ? Yeah I read the comparison articles. 3/8-ID performs pretty close to 1/2-ID.
Cost difference ? Not much if you buy the stuff at the hardward store.
3/8-ID ( 1.09-sq/in cross sectional )
1/2-ID ( 1.94-sq/in cross sectional )
1/2-ID has about 90% more cross sectional area. Can't argue about this, works with exhaust and it works with water. Fuild isn't compressable so it is more important. If everything else is the same 1/2-ID flows more with less resistance than 3/8-ID. If you can make it fit, use it.
Water blocks, OMG! Lots of good choices for CPU's ( lots of lesser performing choices also. ) What is best? Ability to transfer heat into the water at a given flow rate with a minimum of pressure drop.
Keep the water moving.
Reserviors ? I have come to look at the cooling system like a dry-sump setup in a race car. A big reservior is a good thing. ( IMHO ). I understand that everything will eventually arrive at equilibrium within about 2-deg. no mater how much fluid is in the system. In this case I believe I am reducing the work for the water by providing more of it.
So with these thoughts in mind I made the following choices:
Loop #1 ( CPU )
EK-Supremecy Water Block ( works for the next build also )
1x Hardware Labs Black Ice SR-1 360
3x Silverstone Air Penetrator AP121 Fans
1x MCP35X2 Pump
1x EK-MultiOpt 400mm Resv.
Loop #2 ( GPUs )
EK-FC580 ( only full cover block I could find, stupid video cards )
EK-FC Link / Parallel SLI Bridge ( stupid video card & water block, water ports on one side only )
2x Hardware Labs Black Ice SR-1 360 ( series )
6x Silverstone Air Penetrator AP121 Fans
1x MCP35X2 Pump
1x EK-MultiOpt 400mm Resv.
1/2-Fatbox Barb fittings
1/2x5/8 Tygon Flexable/Reinforced Tubing
Threaded hose clamps
Distilled water coolant.
2x bigNG controllers (1) Pumps (1) Fans ( not fully tame yet, needs more work )
Big Frigging Case from mountainmods.com. Extended Ascension.
A little bit over kill ( ok a lot, there is a ton of room in this thing ).
Tech tip -
When verifying the case height is acceptable ( fits under the desk ), check the depth also.
( yes - the case is deeper than expected <sigh> forgot about that ) big mother of a case.
Wife told me I had built an end-table.
----------------------------------------------------------------------------------------------------------------------------------------
Results
This CPU has always run hot, I didn't expect water cooling to change this tendancy and it hasn't.
With air cooling it has always idled around 42-deg C. ( core 1 ). at normal clocking.
OC'd at 4.00Ghz on water it idles at 40-deg C.
A couple of hours running Prime95 Large FFT's max tempature is 62-deg. C.
I'm pretty pleased with this.
Next build ( soon ) we will see about the CPU temp issue.
Single card GTX580 on air idled at 38-deg. C. Stressing this card and 70-deg. C is pretty normal.
Single card on water idled at 28-deg C., stressing it we maxed at 38-deg. C.
OC'd the single carg, idle at 31-deg. C, stressing it we hit 39-deg. C.
Thinking another radiator on the CPU loop might be interesting...
GTX580 2xSLI parallel flow cooling OC'd. Primary card idles at 31-deg C and secondary at 33-deg C.
Stressed up we hit 42-deg C. on both cards.
Again I am pretty happy with this.
Baseline is right around 10-deg C. of ambient. My target is 5-deg. C.
Measurement is suspect, I need a direct water temp measurement before I decide on the next change.
I am happy about this also - new computer goodies are fun.
I will either add another radiator to the CPU loop or increase the airflow.
My choice of fans is suspect, they seem to perform well and they are quiet. Only moving an estimated 120-cfm through each radiator. The radiators are optimzed for low airflow, there is room to improve. this amount of air is about 50% of Martins test setup.
Summary
Have I achieved a satisfactory result ? Yes.
Room for improvement ? Absolutely.
Did I have fun doing it ? You bet!
Thanks to everyone here that knowingly ( or unknowingly ) contributed. Your insight, experience, knowledge and stories have been exceptionally helpful.