Cooling Opterons

[Brad4321]

While waiting for my dual opterons to arrive, I am wondering about my cooling options. Overclocking is virtually nonexistant, so I am mainly looking for something that will keep both of them running nice and cool on my cramped thunder k8w.

I was (kinda) wanting to go with good water cooling system. Keeps noise down (not a major concern since I was running 5 smart case fans 2 at full speed, basically like delta's, but good when I can get it) and everything running nice and cool. However, I have no experience with water cooling, and I don't like taking chances on a motherboard worth more than my soul, not counting everything else in the system.

I am currently running a 5900 ultra. I will be upgrading during the next generation of video cards. If I get WC, I will then have to retube everything when I get the new video card and block for it. Not something I would look forward to.

I was also considering Zalman CNPS7000A-Cu for their performance and kick ass looks. However, they are heavy little bitches and I have read warnings about moving your system. Even though this is all going into my huge lian-li pc70, I still move it around a lot. I am usually careful, as I am in deep love with the case, but I don't want to take chances.

So, I am basically looking for advice. I am pretty sure I could rig up WC leak proof as I am not completely incompetent, but I still really hate to take the chance on these components. I can't overclock at all, so the extra cooling would go to waste.

I am basically looking for something that cools well without being insanely heavy. I am still interested in water cooling if someone can spec me out some good parts to use, but I would go air also. I am fine with not using a kit, but I will probably only get a heatercore prebarbed and designed for PC's.

 

[TekieB]

look at the zalman 7000alcu's, they are lighter, and about the same performance wise.

I've herd some people oc'in opeterons, so it might be worth a shot. But watercooling is "easy" just read up and know what you are doing, and then you really can't screw it up.

 

[Brad4321]

All the (dual) boards out now that I know of don't have any OC options at all, outside of the hypertransport, which really doesn't change that much.

7000alcu? I will look into those.

 

[Ice Czar]

Phase Change Spray Cooled Tech
Jet Cool Modular
is what Im angling towards, but Ive yet to fire off an email for a qoute
both simpler and far more complex than traditional phase change
(process is simpler, component selection far more complex)

currently Im running the stock AMD heatsinks with beefed up fans in a very well aspirated 4U, force fed AC air

water cooling is nice but not quite as easy as its being made out and Id get some experience on a lesser system first,

Im in exactly the same predicament with having sold my soul too (sig)
Personally if Im going to water cool it, I'll be running a chip chiller, something I have no reservations with a cheaper system, but reliability is very important, and automatic shutdown (or failover) just easier off the mobo monitoring
http://www.procooling.com/articles/ < for some monitoring\failover control leads
water cooling adds complexity, and that has to be addressed from a reliability\survivability standpoint
and the system efficiency as well, since the very best your able to do is the ambient without at least evaporative cooling or active phase change, Id opt for a chip chiller if your going to bother at all

cut and paste basics (air\H2O\other)
-----------------------------------------------------------------------------------------------------------------------------


thermal efficiency decreases (thermal resistance increases) the less the temperature differential
which is why thermal solutions are rated with a °C/W rating
(Thermal Resistance (Rht) in °C/W actually)

Heat transfer is all about temperature differentials. Conduction through materials and convection away from surfaces is proportional to the temperature differential that exists.

Delta = CPU temp - Ambient Temp
C/W = Delta / CPU Watts

Interpreting C/W: For every watt (CPUw) that the CPU consumes, the HSF will limit the CPU's temperature rise to (C/W x CPUw) plus the temperature at the HSF's fan inlet. For example, at an ambient temp of 25 C, a C/W of 0.25 with a CPU radiating 50 watts means that CPU temp will increase 50 x 0.25 = 12.5 C over ambient temp, or 37.5 C. The lower the C/W, the better.

any water cooled system employs many heat transfer modes and medias
conduction, convection and radiation (to a far lesser degree)
but temperature differentiual rules the first 2

The heat transfer through the wall follows a simple equation:
Q=k/L(T1-T2)

We can draw some interesting conclusions from this equation. First, heat transfer is proportional to the temperature difference on the object. If the temperature differential doubles, the heat transferred doubles. Second, the conduction coefficient "k" is proportional to heat transfer. If the conduction coefficient doubles, the heat transfer doubles Alternatively, for the same differential temperature, twice as much heat may be transferred. The final observation is "L". As thickness (or distance) increases, heat transfer decreases. Alternatively, to maintain the same heat transfer through a material twice as thick requires twice the temperature differential.

General Heat Transfer Guide < Read this whole guide
for conduction
for convection the corrallary is Newton's Law of Cooling

so much for the basics this gets much more complicated in a water cooling setup since so many additional factors enter into it
flowrate of the whole circuit head, friction loss, pumpheat, heat transfer in the block\turbulence, heat transfer in the rad dependent on the final ambient and the total airflow

areas of common improvement
likely they would be, a less restrictive tubing, and less pump heat
a larger transfer area for the rad or lower ambient

few links
http://www.ocforums.com/showthread.php?s=&threadid=78055&perpage=30&pagenumber=2 (particular attention to RhoXS and BillA)
http://www.procooling.com/articles/html/maximizing_flow_rates_with_h2o.php

advanced thermodynamics studies A Heat Transfer Textbook Lienhard & Lienhard (MIT & U of Houston) PDF

with any basic thermal solution there are 3 main areas to address:

the transfer interface(s) the larger the interface the more heat can be transfered which is why heatsinks have fins, from a massive passively cooled radiator that could be twice as small with a fan assist or a bigger heatsink, to a waterblock with a pelt employed not only as an assist to the transfer but also increasing the heat transfer size from the CPU core through a heatspreader to a much larger interface, which is possible because of the delta a pelt has)

the transfer medium whatever that fluid happens to be, air, water, or phase change refrigerant how well the heat is transfered to it (turbulence being a great assist generally stripping off superheated boundary layers) to the flowrate be it a less restrictive water circuit or a larger fan
more flowrate in and of itself is always better, but its never in and of itself, there is always a penalty in pumpheat (something that is a real concern in a water circuit) TNSTAAFL otherwise known as the 2nd law of thermodynamics

the temperature differential the lower the ambient at the end of whatever set of processes your employing the lower the temperature your able to maintain all the way up the line, in the case of say a phase change setup, the condenser (rad) being cooled by a lower ytemperature air, means the compresser has to do less work to maintain a set temperature (like a room AC) or simply is able to run at a lower temperature over all in a constant on situation like a computer, same applies to an air cooled heatsink or rad in a water cooled system, its able to remove more heat to the air, and thus the water is cooler going through the pump and picking up energy in the form of momentum, and thus able to absorb more heat from the waterblock for a given flowrate and design

design of components is how you address these three issues

 

[Brad4321]

I read all of that, but hardly any made any sense to me.

I was considering some kind of evaporating system too. I am building a case mod that for it to work, has to have water cooling. However, I will be attaching evaporating towers to it, so if I rigged it up for them to work, all that much better. The only side affect is that I leave these systems on all of the time, even when I am not around. If all of the water evaporates, there is a problem.

 

[Ice Czar]

evaporative cooling is a worthwhile experience
but its shortcomings are such that its generally considered more hassle than its worth

1. the better it actually works the more water it looses
(rig a dedicated feed to it like any other swap cooler)

2. actually running the cooolant for both the waterblock circuit and the tower inevitably leads to either serious corrosion or evaporating the additives (to keep that from happening) into the air
the solution is seperate circuits with a heat exchanger, at twice the complexity
and point failures
(advantages, actually creates both cold water and air, and is a nice complementary stage in a hybrid mixed tech solution, are commonly employed as precoolers for AC)

all for 5 to 10 degrees below ambient at best at worse (high humidity) it simply wont work


the previous post outlined some of the physics that influences any thermal solution
review the links and youll have more information then your likely to need
as pointed out maintaining as large a differntial temperature as possible drastically effects whichever technology you employ, the simpler the heat transfer process the more reliable and easier to control it will be

Air Cooled
only one point failure, the fan, which can be monitored by the default mobo protection schemes and shutdown the computer at a set temperature of at fan failure
Heat Transfer Process > Conduction to heatsink > Convection to air > Air removed

Water Cooled
multiple points of failure
leaks throughout the circuit, pump, radiator fan
can only be shutdown by the default aps via temperature
has the potential to rig a failover pump system unlike the air cooled solution
radiator could be large enough to dissapate the heatload (with an overall higher operating temperature) passively if there is a fan failure (or employ multiple fans)
Heat Transfer Process > Conduction to waterblock > convection to water > pump heat added to water > Convection to radiator > conduction through radiator wall > convection from radiator to air > air removed

Extreme section (subambient)

Chip Chiller
An AC employed to refrigerate the coolant instead of the air via a heat exchanger
Points of failure, all the above H2O + the compressor for the refrigerent, and condenser fan
added potential for leaks in the refrigerent circuit
holds the possibility to build a failover (dual AC), or a passively cooled bypass in the event of failure (rad for the water circuit)
-40C
Heat Transfer Process > Conduction to waterblock > convection to water > pump heat added to water > Convection to Liquid to Refrigerant Exchanger, cunduction through wall > convection to refrigerant > convection to condenser > conduction through wall > convection to air > air removal

Phase Change
an AC employed to directly chill a "waterblock" in actuality a refrigerant block
points of failure compressor, condenser fan
failover very limited
-40C
Heat Transfer Process > conduction to refrigerant block >convection to refrigerant > convection to condenser > conduction through wall > convection to air > air removal

Cascade Phase Change
an AC as above cooled by another AC with a different refigerant
able to maintain a much greater temperature differential
twice the point failures
good for -100C or so
Heat Transfer Process > conduction to refrigerant block >convection to refrigerant > convection to condenser > conduction through wall > convection to refrigerant > convection to condenser > conduction through wall > convection to air > air removal

Spray Cooling
a specialized phase change, but there isnt a "compressor" as the phase change is actually initiated by the temperature differential when the coolant impacts the thermal interface, thus its a pump solution
point failures the same as Phase Change
difficult to "home build" special pumps for fluorinert, extremely difficult sealing problems
(fluorinert will leak through a seal that would hold water at 10 atmospheres of pressure)
Heat Transfer Process > Conduction through block (or direct Spray Cooling of component) Convection Phase Change of coolant > Convection to condenser > conduction through wall > convection to air > air removal

 

[Brad4321]

First of all, you are amazing. I think I am going with the safe air method for my opterons. However, after reading all of that again (hopefully understanding more this time ), I am going to do some evaporator for sure. Probably with my athlon 2100 system. I can replace the processor and mobo for $60 if worst comes to worst. The main reason I am interested in this is because I can get a kick ass tower from work for nothing. Blows PVC pipe to hell.

 

[Ice Czar]

if you do that, also run a window AC, use it to force feed the Opterons (lowering its amient)
then duct the exhaust of the cooling tower to the condenser of the AC as a precooler
that will make the compressor have to do less work to maintain a given temperature or allow you to just reach a lower temperature if its constantly on and since the condenser is "outside" any evaporative fumes from the additives is outside as well (make sure the drain pan of the AC is properly set up)
then of course your using the cold water in the other computer

(energy transfer of the heat being used to covert the phase change of water into water vapor, see > latent heat

 

[ryuji]

opterons cant make that much heat do they? ice czar... thats like giving someone the complete webster dictionary in responce to asking how to spell a word.. but you did answer the question your awsone in how much you know tho

 

[dave_graham]

i personally have set up several systems using the Xice bolt 8/6mm waterblocks... with the passive radiator, etc. it works like a charm, is completely silent and frankly, kicks ass.

temps are a little higher than air, but...the silence is amazing...there is also capability for 10/8mm and greater tubing capcity (simply finding the appropriate converter is all that's needed)...there's also capability for video cards to be cooled inline.

cheers,

dave

 

[Ice Czar]

opterons cant make that much heat do they? ice czar... thats like giving someone the complete webster dictionary in responce to asking how to spell a word.. but you did answer the question your awsone in how much you know tho

http://users.erols.com/chare/elec.htm

mine work out to nearly 170 watts\h (2x244s combined) maximum
(612,000 joules or 580 Btu) for just the CPUs
http://www.onlineconversion.com/energy.htm
http://www.members.optusnet.com.au/ncrick/converters/heat.html